Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 
THE BREEDING BIOLOGY OF TWO POPULATIONS OF 
THE WHITE-RUMPED SWIFTLET 
(Aerodramus spodiopygius assimilis) IN FIJI and 
(Aerodramus spodiopygius chillagoensis) IN QUEENSLAND, 
WITH SPECIAL REFERENCE TO 
FACTORS THAT REGULATE CLUTCH SIZE IN BIRDS. 
A thesis presented in partial fulfilment. 
of the requirements for the degree 
of Doctor of Philoso\hy 
in Zoology at 
Massey University 
Michael Kenneth Tarburton 
1987 
( 
??SSEY UNIVERSITY 
lnst:ructio,,s: 
(1) Please complete ti-Jo of these forms, by providing title 
detnils nnd striking out one option from eoch of the 
sections 1-3. 
(2) If you select option b of any section you? include a 
specific time period. The mu.ximum permitted period is 
7-4 months. 
(J) Include the ti-Jo forms 1-Jith the copies of your thesis 
submitted to your supervisor. 
(4) The Library must receive the original of your thesis. 
(5) We strongly recommend that you provide the Library 1-Jith 
ti-Jo copies of your thesis if it has colour plates. 
Hassey University Library. Thesis Copyright Form 
Title of thesis: 
(1) 
(2) 
(b) 
R-'<..ro?mu..s Sf"diopx:Jl,<;,S ct.SSrm ?.l,?s iV'i P::(J ( Gtl\c:l 
A<l-r'odro.MIA.S Sft94_( o f?'('{ji1A5 c.b dlo-.jMI'\St s I-I) ?Id:. 
ss ion for nry . thesis to be made available to 
readers in the Massey Un v ? under conditions 
determined by the Librarian. 
I do not ?-Jish my thesis to be made available to readers 
?-Jithout my ?-Jrittcn consent for I? months. 
??--??????at-m?lt?h?e?s?i?s?,?o?r?a?=c?opy, may be sent to another 
institution under conditions e??-? n . 
I do not wish my thesis, or a copy, to be sent ?o another 
institution ?-Jithout my written consent for [ 2,.. months. 
(b) 
(3) 'i(l';a:r')T-?-"I?ang,..,r,...e:,...?"--t.hat my tbesi s oay be copied fo:r Lib:ra-? 
(b) I do not ?-Jish my thesis to be copied for Library use for 
months. {'2, 
S igned 
Date 
The copyright of this thesis belongs to the author. Readers must 
sign their name in the space belo?.J to shoi-J thnt Lhey recognise 
this. They are asked to add their permanent nddress. 
NAME AND ADDRESS 
MASSEY UN!YERSIT'f 
L1 RARY 
DATE 
i i  
ABSTRACT 
White-rumped Swiftlets Aerodramus spodiopygius (Apodidae) build 
nests of vegetable material and cement (from their saliva) in the dark 
sections of caves at Chillagoe in Queensland, Australia, and in Fiji. 
Fijian colonies average 1,762 nests while the colonies at Chillagoe 
contained an average of 77 nests. Breeding takes place between 
September and March in Fiji, and from October to March at Chillagoe. 
There is no sexual dimorphism and both sexes share in incubation and the 
feeding of nestlings. 
At Chillagoe the clutch is one egg whereas in Fiji it is two eggs 
laid three to five days apart. At Chillagoe incubation took 27. 8 days 
in the poor year and 26.6 days in the good year. In Fiji incubation 
averaged 23 days and 58% of eggs hatched compared to 64% of eggs at 
Chillagoe. The Fijian birds successfully fledged an average of 92%, a 
breeding success of 53% or 1.1 young fledged per breeding pair. From 
the two single-chick broods the birds at Chillagoe fledged 6??. a 
breeding success of 44% or 0.9 young fledged per pair in the good 
breeding season. In the poor year at Chillagoe hatching success was 
6??. fledging success was 5??. reducing breeding success to 3??. At 
Chillagoe the fledging period was increased from 46.9 days in the good 
year to 49.8 days in the poor year. 
At both locations most chick mortality resulted from chicks falling 
from their nests. Lost eggs or chicks were normally replaced by eight 
to fourteen days. Chicks in Fiji were fed an average of 2.2 times a 
day, whereas those at Chillagoe were fed an average of 5.2 times a day 
in the good season and 3.0 times a day in the poor season. 
Placing the data for this species with those for other species of 
apodids shows a positive correlation between egg size and adult size and 
a negative correlation between feeding frequency and the nestling 
period. 
iii  
Producing a thi.rd egg would not benef'i t the Fijian Swif'tlet, which 
could not hatch significantly more eggs when given a third egg and could 
not fledge significantly more chicks when given three chicks instead of' 
their normal brood of' two. 
Fijian birds fed the artificially enlarged broods more frequently 
than normal sized broods, but neither the number of' f'eeds per chick nor 
the number of chicks fledged in the larger broods was increased. 
Parents are apparently maximising the number of fledglings that they can 
raise. 
It is suggested that when there is a food shortage in the breeding 
season some passerines will lose more newly fledged chicks than normal 
whereas White-rumped Swiftlets in Fiji will lose more nestlings than 
normal. 
t: 
Nest size is not restricting clutch size as swif')ets at Chillagoe 
did not raise more young when their nests were enlarged, and predators 
cannot be restricting clutch size because their nests are in total 
darkness. The swiftlets at Chillagoe are on the "mainland" yet produce 
a smaller clutch than those on the Fiji Islands. This is the reverse of 
predictions from the theory of "competitive release" on islands, 
therefore this theory cannot be used to explain the smaller clutch size 
of the birds at Chillagoe. 
The remaining factor is the food supply which is controlled by the 
r f . occ?ence o ra1n. 
days when rain fell. 
The abundance of aerial insects was greater during 
iv 
Adult swiftlets gathered less food in the dry season and in the dry 
periods between rain, and chicks put on more weight during rain periods, 
indicating that food was the critical factor restricting chick growth. 
Additionally, artifici.ally enlarged broods grew more slowly and never 
fledged more chicks than single-chick broods. This demonstrates that 
the abundance of food during the breeding season is the factor that not 
only regulates chick growth but also restricts clutch size. 
The food supply at Chillagoe does not last long enough for swiftlets 
to raise two single-chick broods, but it does last long enough for a 
unique strategy to have been developed which allows them to raise two 
chicks without producing a two-chick brood. This strategy involves the 
female laying the second egg after the first chick is fully feathered so 
that the first chick completes most of the incubation of the egg. The 
second egg hatches after the first chick fledges. The timing of laying 
the second egg leaves both parents free to forage for one chick only and 
allows them to raise two chicks in the shortened breeding (rainy) season 
that is characteristic of the savannah. 
V 
INTRODUCTION 
Swifts and swiftlets form the avian family Apodidae. The 84 species 
in the family (Brooke 1971a), are easily identified by their long curved 
wings and characteristic flight. However identification at the level of 
the species has proved much more difficult. This means that one always 
has to be alert to the possibility (particularly in old publications), 
that particular data may not apply to the species under which the data 
were first published. Early methods of identification involved the use 
of relative wing and tail lengths, furcation of the tail (Stresemann 
1931), and the amount of feathering on the tarsus (Oberholser 1906), or 
all of these (Mayr 1937). More recent studies have found these 
characters, along with colour variation, sufficiently diagnostic for the 
swifts (Brooke 1970, Collins & Brooke 1976), but inadequate for the 
swiftlets. The type of nest and the ability or not to echolocate, have 
been used to reduce the confusion in identifying swiftlets (Medway 1966, 
Medway & Pye 1977). 
Swiftlets are also different to the larger Apodidae in that some 
aspects of their biology make them easier to study. This is because the 
swiftlets not only nest and roost in larger colonies than the swifts, 
but they also use the same locations for both activities and are 
therefore in the one cave, every evening throughout the year. This 
means data can be collected more quickly and easily. This advantage 
over the study of swifts has helped in the study of the breeding ecology 
of the Mossy-nest Swiftlet (Aerodramus vanikorensis), the Black-nest 
Swiftlet (Aerodramus maximus) and the Glossy Swiftlet (Collocalia 
esculenta) (Medway 1962 a, b); as well as the Edible-nest Swiftlet 
(Aerodramus fuciphagus) (Langham 1980). These two studies were made in 
Malaysia where a longstanding interest in the culinary use of swiftlet 
nests has developed. 
vi 
It is probably the restricted distribution of swiftlets within the 
poorly studied tropics of the Indo-Pacific region that has delayed the 
study of breeding in other species. 
The majority of swiftlets have been placed in the genus Aerodramus 
because of their ability to orient acoustically within the dark zones of 
caves, or into the cave entrances at night, by means of echolocation. 
This ability is found in only one other bird, (the Oil Bird, Steatornis) 
and so it is not surprising that a number of studies has been made on 
the acuity of this ability (Griffin 1958, Novick 1959, Vincent 1963, 
Medway 1967, Fenton 1975, Roberts et al. 1976, and Smyth & Roberts 
1983). One study has been made on the Syringeal mechanism for producing 
the echolocatory "clicks" (Suthers & Hector 1982, 1985). 
The remaining swiftlets have been placed in the genus Collocalia and 
the monospecific genus Hydrochous. None of these species echolocate and 
hence they cannot leave the roosting site before daylight. This means 
they cannot reach distant foraging locations in time to benefit from 
feeding there during the prey-rich, pre-dawn period. Nor can they 
remain in distant feeding areas through the similarly rewarding period 
of evening twilight. Hence it is likely the colonies of 
non-echolocating species will be smaller than those of Aerodramus and in 
any case will not be found in the relatively predator-free zone of total 
darkness. This is the case in Niah cave, Borneo, where the single 
non-echolocating species, forms a minority of the 4.5 million birds 
using the cave. (Harrisson 1976, Medway 1962a,b). 
The White-rumped Swiftlet (Aerodramus spoctiopygius assimilis) is the 
only swiftlet feeding over the insect rich forests of Fiji. There are 
numerous caves in Fiji, providing more breeding sites than on Borneo and 
perhaps as a consequence the largest known colonies there number only 
tens of thousands. 
vii 
However, these colonies are larger than the breeding colonies of all 
swifts except the White-naped Swift (Streptoprocne semicollaris) (C.T. 
Collins pers. comm.) or even those of this species in Australia, which 
is the only other location where a number of colonies belonging to this 
species have been censused. 
Large colonies should provide a good sample size for determining the 
breeding, feeding, flight and longevity parameters studied in this 
thesis and as these parameters have not been previously determined for 
this species or for any other in the South Pacific, it was useful for 
this study to have such numerous subjects. However it was found that 
having such large numbers can itself create certain problems, for 
example in gathering an adequate sample of recaptures. 
The main object of conducting observations and experiments on 
assimilis in Fiji was to determine which (if any) of the theories on the 
factors regulating clutch size in birds applies to this species in 
particular and possibly to swifts in general. By conducting similar 
observations and experiments on (A. s. chillagoensis) in the 
unpredictable savannah climate of the Chillagoe district, it was 
intended to examine the question of why this species was an exception to 
the rule of savannah birds producing larger clutches than their close 
relatives in tropical rainforest. 
In presenting the results of my work on the White-rumped Swiftlet in 
Fiji and at Chillagoe in Queensland the general breeding data for both 
subspecies are dealt with together but the data from the experimental 
work on the clutch size for both subspecies are considered separately 
for clarity. The discussion and conclusions for both studies are placed 
together, as some aspects involve both studies. 
ACKNOWLEIX?EMENTS 
The help of the fotowing persons has been appreciated and is 
A 
therefore acknowledged. Charles T. Collins has provided much 
viii 
comparative information, both published and unpublished. Mary Lecroy 
kindly provided data on swiftlets and nests in the collections of the 
American Museum and Kimball Garrett provided measurements of swifts in 
the Los Angeles County Museum. Swallows in the National Museum of New 
Zealand were measured by N.H.S. Hyde. H. Rahn kindly sent copies of 
material that is unavailable in New Zealand. 
Valuable field assistance in Fiji was given by: M. & A. Abk0y, E? 
Bolst, A. Curry, P. Hoffman, T. Kabu, R. Li tster, T. Os borne and J. 
Wells. At Chillagoe the help of Queensland National Parks and Wildlife 
staff (particularly J. Barton, D. Flett and L. Little), members of the 
Chillagoe caving Club (particularly A. Cummins, K. Offer, T. Porritt and 
K. Ridgway) and L. Pecotich is gratefully acknowledged. 
The helpful advice and criticism given by supervisors Dr. Edward 
Minot and Prof. Brian Springett has been of inestimable value. Thi.s 
help was most forthcoming in Section 4, as Dr. Minot and I published a 
part of that work after my first season at Chillagoe. 
Financial assistance towards the last visit to Chillagoe from the 
Frank Chapmam Memorial Fund is acknowledged in appreciation. 
ABSTRACT ? ? ?  
INTRODUCTION ? ?  
ACKNOWLEJX}EMENTS. 
TABLE OF CONTENTS 
LIST OF FIGURES 
LIST OF TABLES. 
Section 1 
TABLE OF CONTENTS 
. .ii 
. v  
. . .  viii 
. ?  ix 
xii 
. . .xiii 
BREEDING BIOLOGY OF THE WHITE-ROMPED SWIFTLET IN FIJI AND CHILLAGOE 
Abstract ? ? ?  
Introduction. 
Description of study areas . .  
Methods ? ? ? ? ? ? ? ? ? ? . 
Results & Discussion 
Distribution in Fiji & AustraHa. 
The nests ? ?  
The nest site ? 
The breeding season ? ? . 
The eggs. ? ? ? ? ? . . 
Moult in the breeding season. 
Incubation. ? ? ? ? ? 
Sex determination 
Egg loss & replacement ? ?  
Extra parental 'co-operation' or 'egg dumping'. 
Nestling development ? ?  
The nestling period ? ? ? ? ? . . 
Fledging success ? ?  
Chick mortality ? ? ? 
1 
2 
3 
6 
. 11 
? ? ? . 13 
14 
. ? ? ? 18 
? ? ? 22 
30 
. ? . ? 31 
. . 35 
? ? ? ? 36 
. 38 
39 
41 
? ? ? ? 42 
. 43 
TABLE OF CONTENTS Cont. 
Adult mortality at nesting caves. . 46 
Nest sanitation & ectoparasites 
Feeding of young. ? ? ? ? ? ? 
A morphological comparison of six sub-species 
Conclusions ? ? 
Section 2 
48 
52 
. 56 
? 60 
AN EXPERIMENTAL MANIPULATION OF CLUTCH AND BROOD SIZE TO DETERMINE 
WHETHER LONG-LIVED TROPICAL SPECIES ARE MAXIMISING THEIR FOOD SUPPLY. 
Introduction. 
Methods 
Results 
Hatching success. 
Chick growth ? ? ?  
Fledging success ? ? ?  
Feeding rate ? ?  
Discussion. ? 
Conclusion. ? 
Section 3 
CLIMATE & INTRA-TROPICAL VARIATION IN CLUTCH SIZE 
Introduction. ? 
Methods 
Results 
Hatching success ? ? ?  
Chick growth. ? ? ? ? 
Fledging success ? ?  
Feeding rate ? ? ?  
Available food supply ? . 
? ? ? ? 61 
? ? ? ? 63 
. 64 
65 
. . . . 65 
. 69 
? .  72 
? ?  74 
? ?  75 
. 77 
. . 78 
. 78 
. ? 84 
87 
? ?  88 
TABLE OF CONTENTS Cont. 
Discussion. ? ? 
Clutch size and 'competitive' release on islands ? .  
The theory that nest size influences clutch size . .  
The theory that relates clutch size to predation. ? . 
A non clutch-size strategy ? . ? . ?  
Regulation of clutch size by stability of food supply . 
Conclusions ? 
Section 4 
A NOVEL STRATEGY IN INCUBATION - CHICK INCUBATION 
Introduction. 
Methods ? ?  
Results ? 
Conclusions 
Section 5 
88 
. 89 
90 
? ? 94 
? ? 97 
98 
? ?  103 
? ?  105 
. 106 
. 107 
? . 108 
DISCUSSION. ? ? ? ? ? ? ? ? ? . . . . ? ? ? . . ? . . . . . ? ? ? . . 109 
REFEREN'CES ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? . ? ? . ? ? ? ? ? ? ? ? . 119 
APPENDICES 
Climatological summary table for Fiji & other breeding locations ? ? ?  130 
Daily rainfall data for experimental period in Fiji ? 
Climatological summary table for Chillagoe, Qld ? ? . 
Daily rainfall 1985, Chillagoe. . ? . ? ? ?  
Daily rainfall 1986/87 Chillagoe . ?  
Maximum temperatures at Chillagoe 1985/86 ? 
The food of the White-rumped Swiftlet in Fiji . 
The food of the White-rumped Swiftlet in Queensland ? 
. . 131 
? . 132 
. 134 
. 135 
. 136 
? . 137 
. . 153 
TABLE OF CONTENTS Cont. 
An experimental manipulation of clutch and brood size of White-rumped 
Swiftlets Aerodramus spodiopygius of Fiji ? ? ? . . . . ? ? ? ? ?  164 
A comparison of the flight behaviour of the White-rumped Swiftlet and 
the Welcome Swallow ? ? ? ? . ? ? ? ? ? ? ? ? ? . ? . ? ? ? ? ? ? 172 
The population status, longevity & mortality of the White-rumped 
Swiftlet in Fiji ? ? ? ? ? ? ? ? ?  
The significance of the vocalizations of the White-rumped Swiftlet 
in Fiji 
Breeding of the White-rumped Swiftlet in Fiji 
? .185 
? ?  216 
? ?  225 
LIST OF FIGURES 
Location of White-rumped Swiftlet colonies studied in Fiji 
Distribution of chillagoensis and terraereginae . .  
8 
? ? 9 
Location of chillagoensis colonj_es visited in this study ? . 10 
Relationship between adult weight and egg weight in swifts & swallows 28 
Relationship between feeding rate and nestling periods in swifts . . .  57 
Mean daily increase in the wing length of chicks in different sized 
broods ? . . ? ? . . . . . . . . . . . 66 
Mean daily increase in chick weight. 
Mean daily increase in tarsal length 
. . 67 
? .  68 
Wing growth in individuals from broods of all three sizes. . 70 
Weight increase in individuals from all three sized broods . . . 71 
Mean daily increase in wing length of chicks at Chillagoe 1985/6 . ? .  79 
Mean daily increase in weight of chicks at Chillagoe 1985/86 ? ? ? ? .  80 
Mean daily increase in wing length of chicks at Chillagoe 1986/87 ? . ?  82 
Mean daily increase in weight of chicks at Chillagoe 1986/87 ? . . ? .  83 
Average daily change in relative chick weights (3 Dec. 1986-23Jan. 1987)85 
Average weight increase in chicks 1986/87 ? ? . ?  
Daily weight change in two single broods 1986/87 . 
. ? .  86 
. 86 
Daily weight change in a pair of manipulated chicks 1986/87. ? . 86 
LIST OF TABLES 
The size of breeding swiftlet colonies at Chillagoe. 
Various apodid and hirundine egg and clutch weights. ? 
Progress of moult in the primary flight feathers ? ?  
. 19 
. . 24 
? . 30 
Length and success of incubation and fledging periods in various swifts 
ar1d s-wallows ? ? . ? . . ? . . . . ? . . . . . ? . . . ? ? . ? 33 
The numbers of louse-flies on chicks at 10 day age intervals ? 
Frequency of chick feeding in the Apodidae ? ? ? . ? ? ? ? ? 
Morphology of four sub-species of the White-rumped Swiftlet ? ?  
Synchronization of moult and breeding in Apodidae. 
. . 51 
. . 54 
? ? 58 
? 101 
Section 1 
BREEDING BIOLOGY OF THE 
WHITE-RUMPED SWIFTLET IN FIJI AND CHILLAGOE 
ABSTRACT 
1 
The White-rumped Swiftlet (Aerodramus spodiopZgius Peale) was 
studied in Fiji (A. ?? assirnilis) and Chillagoe, Queensland, Australia 
(A. s. chillagoensis) ? At the majority of both assirnilis and 
chillagoensis colonies, nests were in totally dru?k sections of caves. 
Local habitat determined which vegetable materials were cemented into 
the nests. Colonies in Fiji contained from 49 to 7,370 nests and in 
Chillagoe from 1 to 264 nests. The colonies in Fiji are the largest yet 
recorded for the species. Eggs were found in Fijian nests only between 
September and March. The clutch, which is one in chillagoensis and 
almost invariably two eggs laid three to five days apart in assirnilis, 
was incubated for an average of 26.6 days during the better season in 
the former subspecies and an average of 23 days in the latter. 
Fifty-eight percent of Fijian eggs hatched successfully compared to 64% 
of eggs at Chillagoe. If the total clutch or young brood was lost, it 
was usually replaced within 14 days. The fledging success of 92% gave a 
breeding success of 53% or 1. 1 young fledged per breeding pair of 
assimilis. In chillagoensis the fledging success for the better season 
was 69% giving a breeding success of 44% or, 0. 9 young fledged from the 
two broods of a breeding pair. In the poor year at Chillagoe incubation 
took 27.8 days, hatching success was 6ry/o, fledging success was 5ry/o and 
the nestling period had increased from 46. 9 days in the better year to 
49. 8 days in the poor year. 
2 
Most chick mortality resulted from the chicks (in both subspecies) 
falling from their nests. There is no sexual dimorphism and both sexes 
share in incubation and feeding nestlings. 
New data are compared with those for other Apodidae and 
Hirundinidae. They show a positive correlation between egg size and 
adult size and a negative correlation between feeding frequency and the 
nestling period. Each Fijian chick was fed on average only 2. 2 times 
each day, making them one of the least frequently fed of the Apodidae. 
Chillagoe chicks were fed an average of 5. 2 times a day. The major 
ectoparasites observed on the birds or their nests were Louse-flies 
(Hippoboscidae). Myophthiria fijiarum was being hosted by assimilis and 
Myophthiria spp was found parasitising chillagoensis. 
INTRODUCTION 
Chillagoensis inhabits inland peninsular Queensland, and has only 
recently been separated from terraereginae which inhabits coastal 
Queensland (Pecotich 1982) . No details on the breeding of either 
subspecies haYebeen previously published.dsassimilis inhabits many of 
the islands of Fiji, and although it has been known since 1912 has never 
been studied previously. In this section I describe the basic breeding 
biology of the two subspecies. Egg size, incubation period, hatching 
success, nestling development, nestling development period, frequency of 
chick feeding and fledging success for both sub-species are discussed 
and compared. The relationships between egg weight and adult weight in 
the Hirundinidae and Apodidae are defined as is the relationship between 
incubation period and nestling period in the apodids. Because many of 
the data in this section are new, it serves as a basis for the 
experimental work on clutch size in the following section. 
3 
DESCRIPTION OF STUDY AREAS 
FIJI ISLANDS 
Fiji is an archipelago of islands situated at latitude 18?S and 
178?E in the South Pacific. Much of Fiji is composed of sedimentary 
rocks. The commonest are limestone and soapstone and so caves have 
formed in many areas including some small islands as well as the larger 
islands. Piercing through this layer of sediments are numerous volcanic 
outcrops which give a rugged appearance to some regions. Viti Levu is 
the largest and tallest island, having 32 peaks that exceed 1, 000 
metres. Vi ti Levu and the other large islands are covered with 
rainforest and where that has been recently cleared, tall grasses, 
creepers and shrubs still give the countryside a tangled green look. 
The green look of the dense forest surrounds the caves on Vi ti Levu and 
is quite different to the sparse savannah vegetation of Chillagoe even 
though both areas are about 17 degrees south of the equator. 
Many of the smaller islands are coral atolls or volcanic peaks. 
Cikobia-i-Lau in Eastern Fiji is small, but having both coral and 
volcanic origins is provided with a mountain and surrounding reefs. 
Enough sediment had been deposited before its uplift to provide 
sufficient limestone for the islands only cave. This cave contains a 
swiftlet colony and is one of those studied in this thesis. There is 
little forest on such islands, as the village g??dens and coconut trees 
occupy most of the areas with soil. 
Village gardens also comprise a reasonable portion of the area on 
Vi ti Levu, where the other swift let colonies studied in this thesis are 
located. The commonest plants in these g??dens are cassava, dalo and 
bananas, while the less common plants include pineapples, pawpaws, bele, 
daruka, corn, pumpkin, egg plant, yams and sweet potatoes. Swift lets 
were seen feeding over these gardens and over regrowth areas of grass, 
sedge and guavas but they collected most of their food over the forests. 
4 
Mosses, lichens and filmy ferns from forest trees are commonly used for 
nest construction but in the m?ier parts fibres from the crown of the 
coconut trees and grasses a:?e used in conjunction with the birds' saliva 
for nesting materials. 
CHILLAGOE REGION 
The town of Chillagoe is adjacent to a series of National Parks that 
stretch through two cattle stations at 17?8 and 144?E 130 km due west of 
Cairns, Queensland, Australia. The Chillagoe region consists of a 
parallel array of chert ridges interspersed with granite outcrops. Both 
of these elevated areas as well as the intervening low country are 
covered with ironbark savannah woodland. More impressive, though more 
confined is the three to five km wide belt of abrupt grey towers of 
sculptured limestone extending from south of Chillagoe, 45 km north 
through Chillagoe and Mungana to the Walsh River, where the limestone 
belt subsides beneath the plain. These towers project up to 70 m from 
the surrounding plain and ru?e often partly covered by vine thicket. 
Both the savannah woodland and vine thicket communities contain 
species that sustain high insect populations which in turn support a 
number of species of insectivorous birds. Because many of the insects 
are aerial or have alate stages in their life cycle, the insectivorous 
birds include several aerial feeders. Two of these, the Rainbow 
Bee-eater (merops ornatus) ancl the Fork-ta:lled Swlft arc migratory and 
time their arrival into this district to coincide with that of the rainy 
season. The Whi te-rumped Swiftlet and the Black-faced Woodswallow 
(Artamus cinereus) are the major resident aerial insect eating birds. 
There is no swallow or martin that is common :Ln the district. 
5 
Amongst the commonest of the trees :Ln the woodland over which the 
swiftlets do most of their feeding ??e the Cullens Ironbark (EucalyPtus 
cullenii), Ghost Gum (E. papuana), the Variable-barked Bloodwood (? 
dichromophloia), Cocky Apple (Planchonia careya) and the Cooktown 
Ironwood (Erythrophloeum chlorostachys). Two of the eight plants from 
the poaceae that grow beneath and between these trees are the Black 
Spear Grass (Heteropogon contortus) and Kangaroo Grass ( Therneda 
australis). These two grasses are the rTBjor vegetable component in the 
swiftlets' nests. By regularly burning these grasses the stockmen of 
Chillagoe and Rookwood stations dwnage other flora as well. However, 
the limestone towers offer protection from most fires. In these towers 
the Broad-leafed Kurajong (Brachychi ton australe), the Red Bauhin..'ia 
(Lysiphyllum cunninghamii), the White Bauhinia (L. hookerii) and the 
Helicopter Tree (Gyrocarpus americanus) find protection from fire. Six 
figs, including the Sandpaper Fig (Ficus coronata) and the Cluster Fig 
(F. racemosa) send their roots down the grikes and weak joints in the 
limestone to aquifers that otherwise would be out of reach. These trees 
support several plants such as the Jasmine Vine (Jasminium recemosum), 
the Pisonia Vine (Pisonia aculeata) and the mistletoes (LyPeranthus 
?). Both these supportive and climbing plants support, directly and 
indirectly, a wide variety of birds and insects. Some of the more 
abundant of the latter include the jumping plant lice, termites, flower 
wasps and ants. 
Many of the grikes and collapses in the limestone towers open into 
caves that fracture large areas of the limestone. About 27 of the 400 
caves explored and tagged by the Chillagoe Caving Club, contain swiftlet 
colonies. Swiftlets use these caves for roosting each night and for 
breeding each wet season (October - March). 
Many of the caves have phreatic passages (formed by flowing water) 
suggesting that the past rTBy have been wetter than the present. 
The remnant presence of plants such as the Stinging Tree 
(Dendrocnide moroides), the False Stinger (Pipturus argenteus), Maiden 
Hair Fern (Adiantum spp), Fishbone Fern (Blechnum spp), and the Black 
Orchid (Cymbidium calycalatum) as well as rainforest birds such as the 
Australian Brush Turkey (Alectura lathami) , the Cassowary ( Casuarius 
casuarius) and the Figbird (Sphecotheres vieilloti) also suggest recent 
desiccation. The Chillagoe form of the White-rumped Swiftlet is the 
only one of eight subspecies to inhabit such a dry area and so it too 
may be viewed as part of Chillagoe's relict tropical rainforest 
community. 
METHODS 
The nesting (and roosting) colonies of assimilis most frequently 
visited were those of Dry and Waterfall Caves at Nasinu 9 Mile, located 
nine miles north of Suva, Fiji ? Visits lasted two to four hours and 
were made in most months during 1974 to 1976 inclusive. In addition, 
frequent visits were made during the early part of the 1976 breedin..g 
season to determine the incubation periods for individually marked 
nests. From 1977 until 1986 these colonies were visited in most years 
to collect longevity data from banded birds. Day-long visits involving 
clutch and brood manipulation experiments (Tarburton 1987a, Appendix 4 
p.164.) were made throughout December 1981 and again in December 1983. 
6 
In October-November 1976 the insertion of triangular pieces of 
galvanized iron with individual numbers punched into them proved quite 
satisfactory for identification of nests, and incubation periods were 
determined. Because the galvanized labels might weaken nests, paint was 
used to mark a number beside the nests used in the manipulation 
experiments of December 1981 and December 1983. This method of nest 
identification proved successful and so has also been used at Chillagoe. 
7 
Ten visits were rrade during 1974 and 1975 to Ono Cave near Wailotua 
Village in the Wainibuka Valley, 45 miles NNW of Suva. 
Waiyala Cave in the Sigatoka Valley 80 miles west of Suva was 
visited throughout 13 January 1975. The only cave on Cikobia-i-Lau in 
the east of Fiji was visited on 7 January 1976. 
The 1981 and 1983 experiments were conducted in the Waterfall Cave, 
which is less accessible and rarely visited by other people. However, 
because rrany of the 150 marked nests were difficult to reach and some 
Common Swifts desert easily (Lack 1956a), I visited nests at two or 
three day intervals. 
Data for chillagoensis were obtained in the main from colonies in 
Gord.ale Scar Pot (CH 187) and Guano Pot (CH 146). These were visited 
six days a week 28 November 1985 - 27 January 1986 and 2 December 1986 -
23 January 1987. Other sites were visited on one or more occasions. 
Two all-day chick watches were rrade in Gordale Scar Pot, to determine 
the number of times a day each chick was fed. The first one was made 
from 0600 - 1300 hrs on 18 December and from 1300 - 1800 hrs on 19 
December 1985 and the second was from 0615 - 1715 hrs on 22 December 
1986. 
The first season at Chillagoe was a normal wet season in which 
chicks grew more quickly and fledged in higher numbers than in the 
following season. The second breeding season was markedly dry and for 
the first time I found nestlings that had starved to death. Because of 
this the 1985/86 season is refe{ed to as the good season and the 1986/87 
season is refered to as the poor season. 
The positions of the Fijian sites are shown in Figure 1 and those of 
Chillagoe sites in Figures 2 and 3. 
Permission from the Fiji Department of Forestry and Fisheries was gained 
to trap and band wild birds. Formal permission was obtained from 
Queensland National Parks and Wildlife Se1?ice before entering any of 
the Chillagoe sites as they are within Chillagoe National Park. 
Clt:J> 
""' 
fl 
0 
()0 
0 
.,.,i 
Figure 1. Location of White-rumped Swiftlet colonies studied in Fiji 
q 
?, 
'W v- I 
? 
0 no-. ?u 
. 
? 
0 (.::; 
?' ?0 
" 
LEVU 
/1 t 
D 
Cl ? 
? 
"' 
( /(ovv- ?? 
v?? 
0 180 E 
-f-1sos 
I 
0 50 
q 
? 
? ., 
.,. 
""' 
100 km 
? tJ 
<1 
1:1 
? 
<) 
D 
:c. 
"" 
t?c?
? 
() Cikobia--i-Lau 
q 
i!. 
0 
Q 
() 
0 Qo 
c3 (X) 
Figure 2 
York 
Iron Range 
0 200km 
Islands 
KEY 
.. .... 
? . . 20?S-.,\.:?.?\. terraereginae ? ,>:.:/:?: . . . ?-; . , . ... ?.?;;; ?\.c hilla ?nsis Finch H atton Gorge ?0/ , 
? breeding site _L___ ?1acka'/?, 
9 
Figure  3 Loca t i on  of chillagoe nsis colon i es  visi t ed in thi s s tu dy 
Rookwood S t n .  
312 
Mungana 
ea  ttle yards 
t 
NORTH 
tt++tt Railway  Li ne 
? Lime stone Bluff 
"-"M a in Road 
0 5 km 
24 
322, 362, 374 
CH 398 
CH 81, 176 
CH 144,167,169 
CH 146 
C Et=#=-::.. Fte5is'tereJ.. N" o? Ca.ve 
W<Th C.hillco.?oe.. C<ltllii!J CI..J:l. 
I 
I I I 
I 
I 
I 
I 
,n 
:? 
.::= 
tOJ 
tl.O 
\0 
?I'D 
' 
I I 
'n . ...., 
1(0 
1(0 
'A 
( 
I 
I 
I 
I 
_L!) 
1 1  
Individual chicks were identified with daubs of fast drying model 
paint placed on their head, shoulder or rump. -ance the chicks were old 
enough both those in Fiji and Queensland had an individually numbered 
aluminium band from the Australian Bird and Bat Banding Scheme, 
Canberra, Australia, placed on one leg. Wing growth, which Lack & Lack 
( 1951) demonstrated to vary with the quantity of food supplied, was 
determined by measuring from the carpal joint to the tip of the longest 
primary when the wing was held flat ?n a ruler. Weights were measured 
on 5, 10 and 50 g Pesola spring balances. The differential growth in 
the wing, and weight, for experimental broods of one, two and three are 
dealt with in section 3. In this section only the normal sized broods 
of one from Chillagoe and one and two from Fiji are considered. 
Nest volume was not determined exactly because that was not possible 
using the published average measurements of nests with which I wanted to 
make comparisons, so I?used an approximation. This approximation, the 
volume index, is derived by multiplying the average length, breadth and 
depth measurements together. 
RESULTS & DISCUSSION 
Breeding Distribution 
Swiftlets nest in caves more frequently than the larger swifts and 
the White-rumped Swiftlet is no exception. All colonies of 
chillagoensis and assimilis visited in this study were located in caves, 
where the majority of nests were in total darkness. In addition to the 
islands of Viti Levu and Cikobia-i-Lau where I describe nesting (Figure 
1), I have seen assimilis on the following islands in the numbers 
indicated: Vanua Levu 20, Qamea 5, Taveuni 7, Rarani tiga 3, (all August 
1976); Vanua Balavu 40, (there is a nesting cave near Moalevu) ,  Naitauba 
8, Laucala 3, (all January 1976); Ono 2, (May 1975). 
I did not see them on Vatu-i-Ra, Sovu, Nanuku Levu, Matagi or 
Mabulau. Although these last five islands are lower than any of the 
others they are larger than Raranitiga on which swiftlets were found. 
Beck and Correia found them nesting on Ovalau and Katavanga (American 
Museum). Layard (1876) recorded the birds from Waikaia, Mokani, 
Taveuni , Mango and Kandavu. 
1 2  
While it is clear that assimilis is widespread in Fiji it is also 
clear that Mayr's (1945) statement that this swiftlet is found on all 
Fijian islands needs qualification. In addition to the islands on which 
I did not find it, Clu?e (1984) records their absence from most of the 
Lomaiviti group, though Watling (1982) refers only to their being 
generally distributed. 
That I found it on hilly islands which were smaller than flat 
islands on which it was not found may demonstrate that the birds require 
rocky chasms or caves for nesting. However, this swiftlet is capable of 
flying daily the distance to some of these islands (Appendix 6 p. 193) 
that it is not found on. So, assuming that assimilis is not inhibited 
by the ocean, it is likely that these islands do not have a broad enough 
range of habitat to provide an attractive food supply. 
Another area of Fiji that is unattractive to swiftlets is the 
rainshadow (NW) side of the larger Fijian islands such as Vi ti Levu, 
where assimilis is much less common than in the areas of rainforest. It 
may be that such a climate is less suitable. This interpretation is 
consistent with chillagoensis producing only single egg clutches and 
being unable to raise a manipulated two-chick brood at Chillagoe which 
has a climate similar to that of the rainshadow area of Fiji. 
Although terraereginae is not studied in this thesis, its 
distribution is close to that of chillagoensis and so its distribution 
is covered here. 
1 3  
It has a patchy distribution along coastal Queensland from Iron 
Range in the north to near Mackay in the south; though it has only been 
found breeding between Mount Peter and Finch Hatton Gorge (Pecotich 
1982). The breeding of chillagoensis is more restricted, being found 
inland around the Chillagoe and Mi tchell-Palmer districts. The two 
subspecies are probably geographically separated, for, at intervening 
savannah districts such as Mareeba, swiftlets are never seen (Bill Long, 
pers. comm.). Swiftlets are present on the edge of the rainforest 
where, for example, near Atherton I have seen feeding flocks of up to 48 
birds. Sub-As Figure 2 shows, the distribution of bothAspecies is still 
unclear in the northern portions of their ranges. 
The positions of those breeding colonies of chillagoensis that have 
been visited in the course of this study are shown in Figure 3. 
The nests 
Except at Veidrala, the nests on Viti Levu were composed of foliose 
and crustiose lichens, liverworts, filmy fern and club moss, held 
together and to the wall with copious amounts of nest-cement (the birds' 
saliva). Each of two nests in the Waterfall Cave had an 80 mm long 
angiosperm leaf built into their structure. Pecotich (pers. comm.) 
described those at Veidrala as being made of local grass material. 
Those in the cave on Cikobia-i-Lau were composed mainly of crown-sheath 
fibre, mid-rib and leaf lamina from coconut trees. 
Internal nest size of 36 nests from Nasinu averaged 50.0 + 4.4 mm (x 
+ SD) left to right; 49. 7 ? 4. 4 mm front to rear; 21. 1 ? 6.0 mm, depth 
from lowest part of front rim. The average volume index of these nests 
is 52 cm3. The weight of four nests ranged between 9.2 g and 22.3 g, 
averaging 15.3 g. 
Twenty-one nests from Veidrala, measured by L. Pecotich ( pers. 
comm.) averaged 46 mm x 56 mm x 29 mm, giving an average volume index of 
75 cm3, which is larger than the Nasinu nests. 
The nests of chillagoensis have similar construction to those in 
Fiji except that they are composed of dried grasses. The most common 
grass species used are Kangaroo Grass ( Themeda austral is) and Black 
Spear Grass (Heteropogon contortus). 
1 4  
The internal dimensions of 100 nests averaged 4.'1.'7. + 0.? mm across, 4'2..7 
+ 0. S mm front to back, and 11. <r .::!::. 0. 4 mm in depth. The depth of the 
Chillagoe nests is significantly (t = '2.0.3. ?< 0.001, df = 1 3?) shallower 
than the Fijian nests (x = 21. 1 mm). The front to back measure of 
Chillagoe nests ( 4'2.. 7 + 0. s- mm) was also significantly smaller ( t = 
I fr. it-, P< 0.001, df = /3fr). The volume index of the Chillagoe nests (2S 
cm3) is consequently smaller than that of the Fijian nests (52 cm3).  
This is to be  expected as the Fijian nests hold two young at a time, 
whereas Chillagoe nests have to hold only one young. 
The nest site and colony size 
All assimilis and chillagoensis colonies visited in this study were 
in totally dark sites, though some assimilis nests at the edge of 
colonies were found in the twilight zone of the cave. The ma.jori ty of 
Fijian nests were located high on the walls and ceilings. of the caves. 
In Ono Cave, Wailotua, some nests were 30 m from the floor. A few nests 
were supported on ledges or protuberances from the wall but the majority 
were fully self supporting. Few nests fell and most nests were used in 
successive years. None was within 2 m of the cave floor, except for one 
or two in Dry Cave, Nasinu. 
The typical colony site at Chillagoe is on a smooth concave wall 2 -
20 m above the cave floor. However small protrusions such as solution 
furrows or fault cracks in the limestone are often favow?ed within the 
colony. Thus rows of nests appear in the general random scatter, with 
those alongside closer than those above and below. 
Some assimilis nests sat a metre or two or in one case 70 m away 
from other nests. 
1 5  
Most nests, however, were much closer together and many adjoined 
neighbouring nests; sometimes many did so, forming clusters so that some 
nests were not attached to the wall at all but only to other nests. The 
average distance between nests at Chillagoe was greater than that 
between Fijian nests. This may result from chillagoensis having more 
caves in which to nest, though the situation is not clear for even in 
colonies of assimilis where many nests were joined, there appeared to be 
large areas of sui table cave wall and ceiling unused. The average 
distances between Chillagoe nests were 7. 9 cm (n = 20) at Gordale Scar 
Pot, 17. 4 cm (n = 36) at Guano Pot, 29 cm (n = 41) at Crack Pot, 29. 4 cm 
(n = 11) at Keef's Cavern, an estimated 10 - 40 cm at Squeeze Pot and an 
estimated 3 - 4 cm at Mudlark Cave (n = 31). The greatest distances 
between nests were found in Flow Cavern, where nests were 10 cm to 5 m 
apart (n = 8). A new site with just one nest was established in Swiss 
Cottage Chamber of Royal Arch Cave (CH 9) during October 1986. This 
site is about 90 m from the other site in this cave. 
Unlike most other swifts, most swiftlets, whether in the 
echolocating genus Aerodramus or the non-echolocating genus Collocalia, 
usually build their nests on the walls and ceilings of caves. Recorded 
exceptions are not numerous and because they can help us derive common 
generic nesting requirements, are related here. The Glossy Swiftlet (? 
esculenta) has been found nesting in overhanging cliffs and the roots or 
cavities of large Banyan and Erythrina trees in the south-west Pacific 
(Layard & Layard 1882; Ma.yr 1945; Parker 1967) , old buildings in Java 
(Medway 1962d) and man-made tunnels in Sumatra (Wells 1975). The 
Himalayan Swiftlet (A. brevirostris) nests :Ln fissures in the walls of 
volcanic craters on Java (Medway 1962d) and the Edible-nest Swiftlet (!:::.:._ 
fuciphaguS) frequently breeds colonially in old buildings in Malaysia 
(Langham 1980). 
1 6  
In Java the mono-generic Giant Swiftlet (Hydrochous gigas) nests behind 
or adjacent to waterfalls, where it and its nest are wet by continual 
spray (Stresemann 1928, Somadikarta 1968, Becking 1971); and the 
Mossy-nest Swiftlet (A. vanikorensis) has also been recorded nesting in 
the man-made tunnels on Sumatra (Wells 1975). 
However there are three nest sites used by larger swifts for which 
there is no record of use by swiftlets. The first is the nest of 
another species; for example the feather-lined mud nests of two swallow 
species that are taken over in East Africa by White-rumped Swifts (Apus 
caffer) , (Moreau 1942b, Brooke 1957). There is one record of 
White-rumped Swifts using the nest of a House Swift (Apus affinis) 
(Brooke 1963). The second site is amongst the foliage of a tree. The 
Fork-tailed Palm Swift (Reinarda sguamata) of Brazil (Sick 1948) and the 
African Palm Swift (eypselus parvus) of tropical Africa (Moreau 1941) 
both build their nests on the frond of a palm. The eggs of both palm 
swifts are glued to the nest and palm frond to prevent their falling to 
the ground. The last site is amongst the sand on the floor, or on 
-ledges, in caves, where the White-naped Swifts (Streptoprocne 
semicollaris) of Mexico, lay their eggs in depressions, Without nesting 
material. This may be the only swift that does not use saliva in 
constructing a nest (Rowley & Orr 1962). In summary Aerodramus requires 
a solid vertical object with some overhang, preferably in a dark 
location, for the location of its nests. 
While there are no records of assimilis nesting in man-made 
environments in Fiji, there are records of them nesting in other than 
caves. They have been reported nesting against overhanging rocks in 
deep gullies (Mercer 1966; Sibson, in Belcher 1972; Clunie 1984). There 
are similarly situated colonies of terraereginae in coastal Queensland, 
such as at Finch Hatton Gorge (Crouther 1983). 
1 7  
This last subspecies also exists in a colony of three nests located 
along with colonies of several species of bats in the Twins Mine Shaft, 
Mount Peter (Smyth et al. 1980). Another non-cave site used by this 
species is the underside of a fallen tree, used by reichenowi in the 
Solomon Islands (Haddon 1981) . In New Caledonia leucopygia nests in 
caves, leaving the Glossy Swiftlet to use more open situations 
(Vuilleumier & Gochfeld 1976). An earlier record from New Caledonia 
(Leach 1928), of White-rumped Swiftlets nesting in the basement of a 
hotel and in a fowl house, as well as in a cave entrance, must be 
suspect, even though one nest had two eggs (the Glossy Swiftlet rarely 
has a clutch of two); as the author grouped the Glossy Swiftlet with the 
White-rumped Swiftlet. Whether Hannecart & Letocart (1980) repeat 
Leach's record or have personal knowledge of the White-rumped Swiftlet 
nesting in abandoned houses and cellars is unclear. Last century most 
leucopygia nested in caves, though some used the underside of sloping 
rocks (Layard & Layard 1878, 1882). In Samoa, where there are no 
congeners, spodiopygius, prefers cave locations for nesting (Armstrong 
1932). 
t. 
Swifjets nest in larger colonies than most swifts and it is not 
unusual for several species to share the same cave. The largest 
recorded colony contains 4 . 5  million birds of three species that share 
Niah Cave in Borneo, with half a million bats (Harrisson 1976) ? However 
colonies as small as one to three nests do occur and the Whi te-rumped 
Swiftlet itself shows a considerable range in size of breeding colonies. 
In Fiji it also shares its caves with bats. 
1 8  
The number of nests in the Fijian colonies visited were : Veidrala 
One, 34 nests ; Veidrala Two, 78 nests, (both Feb. 1976 ) ; Cikobia-i-Lau, 
70 nests, ( 3  Jan. 1976 ) ; West-end colony Ono Cave, Wailotua, 115 nests 
( 1975 ) ; Dry Cave, Nasinu, 163 nests (Jan. 1975 ) ; Waiyala Cave, 2 , 810 
nests (Feb. 1975 ) ; East-end colony, Ono Cave, 3 , 455 nests (June 1975 ) ; 
Waterfall Cave, Nasinu 7 , 370 nests (Dec. 1976 ) . 
The population size of chillagoensis colonies visited, as determined 
by the number of active or fresh nests, is given in Table I. 
With the average colony size of chillagoensis being 77 nests, it is 
clear that this subspecies breeds in much smaller colonies than the 
Fijian subspecies which has an average colony size of 1 , 762 nests. 
However it must also be noted that colonies of chillagoensis are much 
closer to one another than those of assimilis. This means the 
population density per hectare for the two subspecies might not be as 
different as the difference in colony size might at first indicate. 
The breeding season 
In Fiji the annual feather moult is complete by August and nest 
building or repair begins soon after. The first eggs ar'e usually laid 
in September or October. In 1976 the first eggs were seen on 7 
September, the same day that gonads of two other birds were found to be 
enlarged and an egg was felt inside another. Working from the oldest 
chick found on 1 December 1981 (which had a 48 mm wing), the first egg 
was laid about 21  October. The oldest chick found on 11  December 1983 
(wing 81 mm ) gives a laying date for that year of approximately 5 
October. Most chicks at the time of the two latter visits were much 
younger than the largest ch:Lcks and as eggs soon appeared in the few 
empty nests and no juveniles were seen sleeeping in the cave at night, 
it was clear that the ] Arp;er ch i.cks were 8mongst thC' C'arl:lC'st , if not 
the earliest, for the season. 
ccc 
No 
CH5 
CH9 
CH24 
CH26 
CH30 
CH46 
CH52 
CH81 
CH124 
CH133 
CH138 
CH144 
CH146 
CH167 
CH169 
CH176 
CH187 
CH221 
CH227 
CH306 
CH312 
CH322 
CH379 
CH380 
CH381 
CH382 
CH397 
CH398 
1 9  
Table I 
The Size of Breeding Swiftlet Colonies at Chillagoe 
Cave 
Tower of London Cave : site A 
site B 
site C 
Royal Arch Cave : Swiftlet CaveJ?n 
Swiss Cottage 
Keef ' s  Cavern 
Clam Cavern 
Stop Press Cave 
Snakey Cavern 
Swiftlet Cave 
New Southlander Cave 
Flow Cavern 
But Good Cave 
Chinese Cavern 
Christmas Pot (a)  
(b )  
Guano Pot 
Crack Pot 
Squeeze Pot 
Capricorn Cave 
Gordale Scar Pot 
September Cave 
Pope John Paul I 
Mudlark Cave 
Project 31 
Swiftlet Scallops Cave 
Entrance CH 37 4 
Hercules Cave CH 362 
Tarby' s Swiftlet Pot 
Golgotha Cave (a)  
(b)  
Swiftrimlet Cave 
Otobeaswiftlet Cave 
Shirl ' s  Triple Twirl 
Swiftlet Swallet 
Number of currently used nests 
previous ' 85/86 ' 86/87 
80* 
122# 
0 
p-
23# 
p-
46* 
p-
p-
13# 
p-
p?
p?
p?
p?
p?
p-
47# 
500?# 
p-
p?
p-
60+-
0 
0 
15 
0 
109 
11 
22 
29 
44 
81 
194 
135+ 
0 
169 
31 
180 
48 
69 
25 
p 
1 
0 
23 
26 
1 
11 
4 
0 
129 
14 
8 
0 
0 
88 
188 
162 
118 
0 
28 
253 
8 
49 
23 
179 
29 
p 
21 
264 
Note An * indicates that only the marks left on the walls by previous 
nests were counted and so that this number existed at the one time cannot 
be guaranteed . + indicates that there were possibly a few more nests on a 
small shelf that was not fully visible . ? indicates that I have some 
doubt that this number of nests were recently found in this cave ? # 
refers to data from Smyth et al . - indicates a figure taken from the 
records of the Chillagoe Caving Club . P indicates swiftlets present . 
2 0  
An examination of 104 nests on 2 2  October 1976 revealed that 6?/o of 
nests were empty , 16% had one egg and 24% had two eggs . One week later 
45% of 149 nests were empty , 16% had one egg , 3?/o had two eggs and 1% 
contained a chick . By November only 1?/o of nests remained empty. 
On 14 November 197 4 ,  5?/o of Nasinu nests had eggs and 5?/o had young. 
These nests had similar percentages the following year and were closely 
synchronized with the Wailotua population , which had chicks in 52% of 
their nests on 25  November 1974 . 
In years such as 1974 and 1976 , when laying started in September , 
the first chicks were found flying in the last week of December . 
However , in each year studied , the majority of chicks fledged during 
January. Because lost clutches and sometimes lost broods are replaced , 
the breeding season is extended into February or even March . On 2 
February 1975 about 8?/o of Nasinu nests were empty , with large chicks 
occupying the remainder . However , there is some geographic variation 
because a few days later Pecotich (pers . comm. ) found a few eggs at 
Veidrala. Veidrala is in a drier region and such a climate may delay 
the breeding season . It is also possible that human or other 
disturbance may have caused more than normal losses and thus more than 
normal re-lays . 
Turning to the Chillagoe situation and working from the oldest chick 
(which had a 51  mm wing ) , found in Gordale Scar Pot and Guano Pot at 
Chillagoe on 29 November 1985 , the first egg was laid about 6 October . 
The oldest chick found on 3 December 1986 (wing 57 mm) gives an 
estimated laying date of 4 October for the first egg that season. In 
both seasons the first few eggs were followed two weeks later by a heavy 
bout of laying. 
The importance of discovering the proximate physical cause for 
triggering laying is that it may link into a major ultimate factor for 
regulating clutch size in the life history of the species . 
To discover this for the White-rumped Swiftlet is the function of 
Section Two and so is worth pursuing. 
2 1  
Various external stimuli may initiate or inhibit the culmination 
phase of the reproductive cycle (Marshall 1961 ) . In many birds , the 
presence of a mate , nesting site , nesting material and finally  the nest 
itself would be necessary stimuli for egg laying . All of these may be 
operative on the White-rumped Swiftlet , but they do not affect the 
timing of egg laying. 
Adult White-rumped Swiftlets roost with their mate each night of the 
year at their nest , and will  add nesting material to it <JVI{,... a.l'l e_)[Tefld.Qc(. 
bre?c:l(Y\? sea..son if it is needed . 
We are left then with food supply and climatic factors as possible 
stimuli for egg laying . While I do not have food samples for the 
non-breeding period , it is likely that if food is a proximate factor 
that climate wil l  still be the ultimate factor , as the abundance of 
aerial insects is closely linked to climatic factors (Lack & Owen 1955 , 
Hespenheide 1975 ) .  It is clear that egg laying in assimilis corresponds 
with the increase in rainfall that occurs in September and October 
(Appendix la. p . 130) . However egg laying in chillagoensis precedes the 
increase in rainfall during November (Appendix lb . p . l31 ) . In 1985 
there had been no rain for 28 days (Appendix le . p . l32 ) prior to the 
laying of the first egg for the season and there had been only small 
showers prior to the peak laying period . So rain does not stimulate 
laying in chillagoensis even though rain ( as is shown later ) is  
essential to  raising chicks . 
The average date of first layings in the Great Tit (Lack 1966 ) has a 
high correlation with temperatures for March and April and thus 
temperature might have a direct effect on the tit . This is the more 
likely because food levels do not increase until after egg laying (Lack , 
Gibb & Owen 1957 ) . 
2 2  
Because chick mortality rate is lowest in those years that the Alpine 
Swift (Apus melba) produces larger clutches than normal it was 
postulated that the swifts were responding to some unknown factor during 
the laying time that is correlated with the food level three weeks later 
(Lack & Am 1947 ) . Further evidence that the unknown factor in the 
Apodidae might be temperature is that the years when the Common Swift 
(Apus apus) produces larger than average clutches are characterized by 
the period immediately prior to laying being warm and sunny (Lack 
1956a) . If the weather turns cold in central Europe after the Common 
Swift has laid , it will throw one , two or three eggs ( "the worse the 
weather the more eggs" ) out of the nest (Koskimies 1950 ) . 
In considering temperature as a possible trigger for laying it is 
clear that maximum temperatures are more likely to affect feeding 
White-rumped Swiftlets or their food , than minimum temperatures which 
occur when the bird is in the insulated environment of the nest and the 
cave . Such a relationship has been found between air temperature and 
egg laying in the House Swift (Naik & Razack 1967 ) . While it was found 
that the minimum air temperature had little effect , eggs were laid 
following a week where the average maximum temperature was between 34?C 
and 36?C .  
The maximum temperatures for 1985 indicate that the first 
White-rurnped Swiftlet egg was laid after a week of temperatures above 
about 32?C .  However in 1986 a week of similar temperatures occurred in 
early Septembe? a month before the first recorded egg appeared . Hence 
Pl"o?a,IJ!J..: 
maximum temperatures?ao not stimulate laying in this swiftlet even 
though they do in the House Swift . 
The eggs 
As in all swifts but one , the eggs are white and without gloss . The 
exception , the Cape Verde Swift (Apus alexandri ) , has small 
reddish-brown spots on its ?ggs (Alexander 1898 , Brooke 1971a ,b ) . 
2 3  
The average dimensions of 118 chillagoensis eggs (from Gordale Scar Pot 
and Guano Pot ) were 19 . 81 + 0 .06 x 13 .04 + 0 .04 mm (x .::!::. S . E. ) .  Both the 
length ( t  = 12 . 3 ,  P< 0 .001 )  and the width ( t  = 10 . 1 ,  ?< 0 .001 )  of these 
eggs are significantly larger than 50 assimilis eggs from Nasinu Caves 
in Fiji ( 18 . 37 + 0 . 1  x 12 . 14 + 0 .08 mm (x .::!::. SE?. However both the 
length ( t  = 0 . 2 ,  ns ) and the width ( t  = 1 . 0 ,  ns ) are not significantly 
smaller than eight assimilis eggs from Waiyala ,  which measured 19 . 87 .::!::. 
0 . 32 x 13 . 3  + 0 . 27 mm . As might be expected ,  both the length ( t  = 4.47 ,  
P <0 .001 )  and the width ( t  = 4 . 12 ;  P <0 .001 ) of assimilis eggs from 
Nasinu are significantly smalleJ? than the assimilis eggs from Waiyala.  
The eggs from Dry Cave were not significantly different from 52  slightly 
smaller eggs that came from replacement clutches in the same cave . 
These averaged 18 . 11 + 0 . 12 x 11 . 99 .::!::. 0 . 05 mm. However these 
replacement eggs ( 1 . 53 + 0 .02 g)  were heavier than the larger original 
eggs ( 1 . 42 ? 0 . 02 g ) , probably because they were newly laid when 
weighed , whereas some of the original eggs were not . 
{\ = t0'2. 
The fresh egg of the Chillagoe Swiftlet ( 1 . 83 + 0 .01 g)  is  - A. 
significantly heavier ( t  = 13. 4 ,  P< 0 .001 )  than 52 freshly laid eggs 
from Nasinu Caves Fiji ( 1 . 53 .::!::. 0 . 02 g) . The average weight of the 
Chillagoe eggs had declined to 1 . 66 + 0 .03 g a day or two before 
hatching . The average fresh weight of the Chillagoe eggs represents 
19 . 7% of the average adult weight . This is slightly more than the 
18 . 7% for eggs and birds at Nasinu , Fiji . 
The proportion of the weight of the eggs to the weight of the five 
swiftlets ( 19 . 5% ? 2 . 15 ;  x .::!::. S . E . ) shown in Table II is significantly 
greater than the proportion of the eggs of the 25 swifts ( 10 .?/o .::!::. 0 . 77 )  
and that of the six swallows and one martin ( 10 . 8% .::!::. 0 . 65 ) . The 
comparison with hirundines is included because they are similar sized 
birds that have had to cope with similar ecological problems as the 
swiftlets . 
TABLE II 
Various Ap?did and Hirundine egg and clutch weights 
----- ----------------------------- ---------------------------------------------------
Species Egg Female Egg Clutch Clutch 
mean mean (% ad size (% ad 
wt(g) wt(g)  wt) wt) Source 
'? 
----- ----------------------- ----------------------- -------------------------
Edible-nest Swiftlet 
Aerodramus 
fuciph?gus 2 . 16 8 . 1  26 . 7  1 . 96 52 . 3  Becking 1971 ; Langham 1980 
Grey-rumped Swift 
Chaetura 
cinereiventris 3 . 06 13 . 8  22 . 2  5 .00  110 . 9  Schonwetter 1970 ; Snow 1962 ; 
Mossy-nest Swiftlet Sick 1948 
Aerodramus 
vanikorensis 2 . 42 11 . 3  2 1 . 4  2 .00  42 . 8  Becking 1971 
White-rumped Swiftlet 
Aerodramus 
s2odio2ygius 1 . 53 8 . 2  18 . 7  2 .00 37 .4  This paper 
White-collared Swift 
Stre2to2rocne 
zonaris 12 .00  65 . 8  18 . 2  1 .00 18 . 2  Schonwetter 1970 ; Snow 1962 
Glossy Swiftlet Sick 1948 
Collocalia 
esculenta 1 . 21 7 . 1  17 . 0  1 . 95 33 . 2  This paper ; Medway 1962a 
Giant Swiftlet 
Hydrochous gigas 5 .06 37 . 0  13 . 7  1 . 00 13 . 7  Becking 1971 ; Somadik??ta 1968 
Band-rumped Swift 
Chaetura s2inicauda 1 . 88 14 . 2  13 . 2  - - Schonwetter 1970 ; Snow 1962 
Wire-tailed Swallow 
Hirundo smi thi 1 . 65 12 . 7  13 . 0  4 .00  52 .0  Schonwetter 1970 ; Donnelly 1974 
Chestnut-collared Swift 
Cypseloides rutilus 2 . 67 21 . 0  12 . 7  2 .00  25 . 4  Schonwetter 1970 ; Snow 1962 
Lack 1956b 
N 
..,.. 
Table II cont ? .  
Species Egg Fen:ale Egg Clutch Clutch 
mean mean (% ad size (% ad 
wt(g)  wt (g)  wt ) wt ) Source 
-------- ---------------- -----------------------------------------------------------
Welcome Swallow 
Hirundo neoxena 
Whi te-rumped Swift 
Apus caffer 
Black Swift 
Cypseloides niger 
Lesser Striped Swallow 
Cecro2is abysinica 
White-tipped Swift 
Aeronautes 
montivagus 
Bat-like Spinetail 
Neafra2us boehmi 
African Black Swift 
Apus barbatus 
Horus Swift 
AEus horus 
House Swift 
Apus affinis 
Pacific Swallow 
Hirundo tahi tica 
Mosque Swallow 
1 .  75 
1 . 52 
2 . 75 
5 . 57 
2 . 02 
2 . 19 
1 . 61 
4 . 40 
2 . 77 
2 . 75 
1 . 45 
Hirundo senegalensis 4 . 00 
Barn Swallow 
Hirundo rustica 1 . 90 
Fork-tailed Palm Swift 
Reinarda squamata 1 .01 
13 . 9  
13 . 9  
22 . 1  
45 . 6  
17 . 0  
19 . 7  
15 . 3  
42 . 3  
27 . 4  
27 . 2  
14 . 7  
40 . 7  
19 . 5  
10 . 3  
12 . 6  3 . 90 49 . 0  This paper ; Schonwetter 1970 
10 . 9  3 . 50 48 . 8  This paper 
12 . 4  2 . 50 31 . 1  Schonwetter 1970 ; Brooke 1971d 
12 . 2  1 .00  12 . 2  Schonwetter 1970 ; Collins unpub. 
11 . 9  3 .00  35 . 6  McLachlan & Liversidge 1978 ; Schonwetter 1970 
Walters 1980 ; Donnelly 1974 
11 . 0  2 .00 22 . 0  Schonwetter 1970 ; Collins 1972 
Collins pers . comm. 
10 . 5  3 .00  31 . 6  Schonwetter 1970 ; Lack 1956b 
Donnelly 1974 
10 . 4  2 .00  20 . 8  Brooke 1971a ; Schonwetter 1970 
Brooke 1970 
10 . 1  2 . 00 20 . 2  Schonwetter 1970 ; Walters 1980 
10 . 1  3 . 70 (a) 37 . 4  Chih-Tung 1973 
2 . 50 (b)  25 . 3  Chih-Tung 1973 
9 . 8  2 . 98 29 . 2  Hails 1984 
9 . 8- c . 3 . 50 34 . 4  Schonwetter 1970 ; Donnelly 1974 
9 . 7  4 . 40 42 . 9  McLachlan & Liversidge 1978 ; Adams 1957 
9 . 7  3 .00  29 . 3  Schonwetter 1970 ; Lack 1956b 
N Vl 
Table II cont ? ?  
Species Egg Ferm.le Egg Clutch Clutch 
mean mean (% ad size (% ad 
wt; (g)  wt; (g)  wt;) wt;) Source 
-------------------------------------------- ------------------------- --------
House Martin Collins unpub . 
De lichen urbica 1 . 90 19 . 8  9 . 6  c .4 .00  38 . 4  Mclachlan & Liversidge 1978 ; Walters 1980 
Vaux ' s Swift Schonwetter 1970 
Chaetura vauxi 1 . 73 18 . 1  9 . 6 c .4 .40 42 . 0  Schonwetter 1970 ; lack 1956b 
Pallid Swift Baldwin & Zaczkowski 1963 
A:[2US pallidus 3 . 50 36 . 7  9 . 5  c . 2 .00  19 . 1  Schonwetter 1970 ; Walters 1980 
Fork-tailed Swift Palomeque et al 1980 
A:[2US 2acificus 2 . 75 29 .0 (d )  9 . 3 2 . 50 23 . 2  lack 1956b ; Gibson 1981 
2 . 75 43 . 5 ( e )  6 . 3  2 . 50 15 . 8  Dementiev et al 1951 
4 . 40(g) 35 . 5 (f )  12 . 4  2 . 50 30 . 1  Shaw 1936 ; Bent 1940 
Sabines Spinetail Dementiev et al 1951 
Chaetura sabini 1 . 42 16 . 6  8 . 6  c . 2 . 50 c . 1 . 4 Lack 1956b ; Brooke 19718  
-
Common Swift 
A2us apus 3 . 62 42 . 7  8 . 5  2 . 27 19 . 2  Lack & Lack 1951 
Chimney Swift 
Chaetura pelagica 1 . 88 24 . 1  7 . 8  4 . 20 32 . 8  Fischer 1958 
White-throated Needletail 
Hirunda:[2US 
caudacutus 8 . 45 113 . 7  7 . 4 5 . 25 38 . 9  Neufeldt & Ivanov 1960 ; 
White-naped Swift Lobko-Lobanovski 1956 
Stre2to2rocne 
semicollaris 12 . 50 c . 175 .0  7 . 1  2 .00 14 . 3  Schonwetter 1970 ; Collins 1968b 
African Palm Swift Rowley & Orr 1962 
Cypsiurus parvus 1 . 00 14 . 6  7 . 0 2 .00 14 . 0  Brooke 1971a 
N Q\ 
Species 
Chapmans Swift 
Chaetura chapmani 
White-throated Swift 
Egg 
mean 
wt (g)  
1 .  76 
Aeronautes saxatalis 2 . 10 
Alpine Swift 
Apus melba 6 . 05 
Brown Needletail 
Chaetura gigantea 8 . 35 
Female 
mean 
wt (g)  
25 . 5  
32 . 3  
98 .0  
137 . 0  
Table II cont ? ?  
Egg 
(% ad 
wt )  
6 . 9  
6 . 5  
6 . 2  
6 . 1  
Clutch 
size 
c . 2 . 50 
c . 4 . 30 
3 . 50 
c . 3 . 50 
Clutch 
(% ad 
wt ) 
17 . 3  
28 . 0  
21 . 6  
21 . 3  
Source 
Schonwetter 1970 ; Snow 1962 
Collins 1968b 
Schonwetter 1970 ; Walters 1980 
Behle 1973 ; Bent 1940 
Lack & Lack 1951 ; Schonwetter 1970 
Lack & Arn 1947 
Schonwetter 1970 ; Lack 1956b 
Collins & Brooke 1976 
NOTES : The species are arranged in descending order of the ratio between egg weight and female weight . 
(a)  size of first brood , (b )  size of second brood . ( c . ) circa , indicates an approximate figure taken from another 
source & used because original studies did not specify this parameter . (d )  weight of an emaciated bird from the 
Aleutian Islands . (e ) weight of two birds from Russia . (f)  median weight of birds from China . (g)  egg weight for 
Japanese birds . 
N -.....) 
1 0  
--..... 
? 5 
4-
..c: 
01 
QJ 3 
3 
? 
01 ? 
01 ? ?  
w 
0 
" , , " , " , , 
? 
"' " 
? ? " " 
.. ... " " " " " " 
" Ill ? , " . . 
, "? 
," 11 ? 
? 
? 
3 0  
we ig ht 
? 
? 
? 
50  1 00  200  
(g) 
Figure 4 .  Relationship between adult weight and egg weight in the swifts 
( circles) and swallows ( squares )  ? 
The solid line is the line of best fit for the Apodidae (Y  = 0 . 28 x0 ? 71 , 
R2 = 76%, P< 0 .001 ) . The dotted line is the line of best fit for the 
. o .  79 2 Hirt.mdinidae (Y = o.  20 X , R = sgo,b , 
P< 0 .001 ) . In both equations Y = egg weight , X =  adult weight , 
2 8  
R2 = coefficient of determination. The White-rumped Swiftlet is designated 
by the hollow circle . 
2 9  
The relationship between adult weight and egg weight for those species 
shown in Table II is shown in Figure 4 .  The lines of best fit for the 
Apodidae and the Hirundinidae both agree with the general trend of large 
birds laying smaller eggs in relation to the parents ' size than do small 
birds (Heinroth 1922 ) . While the White-rumped Swiftlet , represented by 
the hollow circle in Figure 4 ,  lays a proportionately large egg ,  the egg 
is not extreme for swiftlets . 
It is unfortunate that there is some discrepancy over the weight of 
the egg of the Edible-nest Swiftlet . Langham ( 1980) gives the average 
weight of 15  eggs as about 1 .  2 g or 10';b of adult weight . This contrasts 
with the figure ( 26 .  7%) used above (Becking 1971 ) .  The greater weight 
has been accepted because the eggs of the Edible-nest Swiftlet have 
larger dimensions (Morgan 1875 , Gibson-Hill 1948) , than those of 
assimilis and so could be expected to weigh more than assimilis eggs do 
( 1 . 53 g) . 
One of 57 assimilis clutches had one egg - the rest had two eggs . 
This gives a mean clutch size of 1 . 98 ,  which is greater than that of the 
four other swiftlets whose clutches have been recorded . As assimilis 
persists in laying a second egg even when provided with a second one , 
and as it will lay no more than two if its second egg is removed the day 
after laying , this species is a determinate layer ( Cole 1917 ) . Other 
swifts vary their clutch size . Chimney Swifts ( Chaetura pe lagica) 
increase average clutch size from 3 . 5  in their first year to 4 . 1 in 
successive years (Dexter 1981 ) and House Swifts , which are double 
brooded , produce a mean of only 2 .  5 eggs in their second brood compared 
to 3 . 7  in their first (Chih-Tung 1973 ) . These changes in clutch size 
are , however , proportionately much smaller than that which the 
White-rumped Swiftlet would have to make in altering its clutch size by 
just one . 
The laying interval for assirnilis in ten instances was found to 
average four days , with a range of three to five days . This interval 
and that of the Edible-nest Swiftlet ( three days , Langham 1980) is 
larger than that of the Mossy-nest Swiftlet ( two days) , the Glossy 
Swiftlet ( two days , both Medway 1962a) and most Common Swifts ( two or 
30  
three days , Lack & Lack 1951 ) . Similar sized Passerines usually lay on 
successive days (Welty 1975 , Marchant 1984) as does the Welcome Swallow 
(Hirundo neoxena) (Marchant & Fullagar 1983) . The reason for a longer 
laying interval in the swiftlets is very likely the greater proportion 
of the weight of their egg to that of the adult producing it . 
Moult in the Breeding Season 
The synchrony of breeding and moulting in this species is discussed 
more fully in Appendix 5 (pp . 181 , 182 . )  where possible benefits of this 
unconmon combination are suggested . Data in Table III show that moult 
of the primaries begins for most Chillagoe birds in late November or 
early December . 
Table III 
Progress of Moult in the Primary Flight Feathers 
No . examined 
No . not in moult 
No . in moult 
P1 
P2 
P3 
P4 
P5 
P6 
P7 
PS 
P9 
P10 
29 Nov 
-8 Dec 
108 
23 
58 
20 
6 
1 
9-18 
Dec 
97 
9 
17 
29 
14 
18 
7 
2 
19-28 29 Dec 8-17 
Dec -7 Jan Jan 
87 106 50 
1 0 0 
2 3 
30 10 2 
27 36 10 
14 33 20 
13 13 16 
3 1 2 
18-27 
Jan 
55 
0 
2 
10 
31 
10 
2 
NOTE: Pn = the number of the most recently moulted ( smallest or missing) 
flight feather. 
Moult ?ollows immediately upon egg laying and because o? the 
difficulty o? sexing individuals it is not certain whether one sex 
starts be?ore the other . Male and ?enEle House Swi?ts moult their 
primary ?eathers together and also begin immediately ?ter egg laying 
(Naik & Shi vanarayan 1969 ) ? 
3 1  
I? the rate o? moult in the primaries continues ?or another two 
months as it does during December and January (Table III ) , then moult o? 
the primaries takes ?our months to complete . 
This length o? time equals that taken by assimilis in Fiji (Appendix 5 
p . 181 . ) and that taken by the Chimney Swi?t ( Johnston 1958 ) , but is 
shorter than the seven months taken by both the Black-nest Swi?tlet 
(Aerodramus maxi!TR1S ) (Medway 1962c ) and the Alpine Swi?t ( Stresemann 
pers comm in Naik & Naik 1965 ) ? The time required ?or primary moult is 
apparently less in passerines than in swi?ts .  Some passerine examples 
and the time taken ?or them to achieve a complete moult are : Redpolls , 
(Acanthis ?lanmea) , 48 - 56 days ( Salomonsen 1972 ) ; House Sparrows 
(Passer domesticus ) ,  64 - 83 days (Haukioja & Reponen 1969 ) ; Eurasian 
Yellow-breasted Buntings , (Emberiza aureola) , 42 - 63 days ( Stresemann & 
Stresemann 1966 ) ; Steller Jays (eyanocitta stelleri ) ,  60 - 70 days 
(Pitelka 1958 ) ; Andean Sparrows ( Zonotrichia capensis) , about 60 days 
(Miller 1961 )  ? 
Incubation 
The incubation period in chillagoensis was ?ound to average 26 . 6 .?. 
0? 23 days (n  = 27 ) and to range between 25 and 29 days . This is three 
and a hal? days longer than incubation (from the laying to the hatching 
o? the second egg) in assimilis (Range = 22 - 25 days , n = 7 ) . From the 
relationship between adult weight and incubation period shown below, I 
would have predicted that chillagoensis had a shorter incubation period 
than assimilis . 
3 2  
However , as apodid chicks are poikilothermic (Collins 1968a , 1 973b) , the 
cooler temperature in the Chillagoe caves may be at least a contributing 
cause . During incubation the temperature in Gordale Scar Pot never 
varied more than half a degree from 23?C whereas Fi jian cave 
temperatures ranged between 24. 8  and 26 . 4  ?c .  Temperatures outside the 
Fijian cave ranged between 21 . 7  and 31 . 8  ?C .  
Table IV shows the incubation period to be similar to that of the 
Edible-nest Swiftlet , the Mossy-nest Swiftlet , the Chestnut-collared 
Swift , the White-rumped Swift and the House Swift ; less than that of the 
Black-nest Swiftlet ; and longer than that of the other apodids shown in 
the table . 
Using the length of the incubation period of the 12 apodids for 
which egg length and adult weight was also known , no significant 
correlation was shown between egg size and the length of the incubation 
period ( Spearman Rank Correlation , rs = 0 . 14) . However , there was a 
significant inverse correlation between adult weight and incubation 
period (r  = -0. 55 ;  P <0 . 01 ) ; and a significant correlation between s 
adult weight and egg size (r  = 0 . 71 ;  P <0 .001 ) . These results suggest s 
that , although larger swifts tend to lay larger eggs , it is larger adult 
size rather than larger egg size that accompanies a reduction in the 
incubation period. 
The observation of marked birds in Fiji during four full days of 
nest watching revealed that , as in all swifts so far studied , incubation 
was shared by both parents and that they changed duties at about 24 h 
intervals . Once an incubating bird at Chillagoe was caught it would 
tend to fly from the nest on successive approaches and it subsequently 
became easier to catch its mate when it  was incubating . Catching and 
banding such birds did show that , like the Fijian subspecies , both sexes 
share in incubation.  
PLATE 1 Adult White-rumped Swiftlet (Aerodramus spoctiopygius 
assimilis)  incubating. 
Brooding adults of both subspecies may be identified by the dark tipped 
primaries which extend well beyond the tail . Note also that in 
assimilis nests are often joined to others . This means that as 
additional saliva (visible in the photo ) is  applied to strengthen the 
nest ' s  attachment to the wall during incubation , the wings of sleeping 
birds in adjoining nests are sometimes glued to the nest resulting in 
their death. 

TABLE IV 
Length and Success of Incubation and Fledging Period in 
Various Swifts and Swallows 
Incubation Mean Hatching Fledging Fledging Breeding Young Fledged 
Period Clutch Success Period Success Success /Breeding 
Species (days ) Size (%) (days ) (%) (%) Effort Source 
-
White-rumped Swiftlet 
A.  s .  assimilis 23. 0  1 . 98 58 . 0  46 . 0  92 .0  53 . 0  1 . 1  Tarburton 1986b 
A.  s .  chillagoensis 26 . 6  1 . 00 64. 0  46 . 9  69 . 0  44. 0  0 . 9  This thesis 
Glossy Swiftlet 
Collocalia esculenta 21 . 5  1 . 95 76 . 1  38 . 5  75 . 0  57 . 0  1 . 1  Medway 1962a, Medway 
Black-nest Swiftlet & Pye 1977 
Aerodrarrus maximus 28 . 0  1 .00  29 . 3  60 . 0  57 . 3  17 . 0  0 . 2  Medway 1962a, Medway 
Mossy-nest Swiftlet & Pye 1977 
Aerodrarrus vanikorensis 23 . 0  1 . 70 51 . 5  48 . 0  74 . 1  38 . 0  0 . 8  Medway 1962a , Medway 
Edible-nest Swiftlet & Pye 1977 
Aerodrarrus fuciEhagus 23 . 0  1 . 96 79 . 4  43 . 0  61 . 9  49 . 0  0 . 9  Langham 1980 
Chimney Swift 
Chaetura Eelagica 19 . 0  4 . 20 89 . 5  30 . 0  96 . 1  86 . 0  3 . 6  Fischer 1958 
Short-tailed Swift 
Chaetura brach? 17 . 5  3 . 80 57 . 7  c . 33 . 0  53 . 5  27 . 0  1 . 1  Collins 1968a 
17 . 5  3 . 60 - 35 . 0  - - - Snow 1962 
Chapman 1 s Swift 
Chaetura chapnani 17 . 5  2/3 - - - - - Collins 1968b 
vaux 1 s Swift 
Chaetura vauxi 18 . 5  6 . 30 100 . 0  28 . 0  67 . 0  67 . 0  4 . 0  Baldwin & Zaczkowski 
Chestnut-collared Swift 1963 
Cypseloides rutilus 22 . 8  1 . 98 - 63 . 0  39 . 6  63 . 0  40 . 0  1 . 3  Snow 1962 
22 . 5  1 . 90 65 . 8  40 . 0  68 . 4  36 . 0  0 . 9  Collins 1968a 
Common Swift 
Apus apus (Britain) 19 . 6  2 . 27 78 . 0  42 . 0  74 . 5  58 . 0  1 . 3  Lack & Lack 1951 
(Switzerland) 20. 0  2 . 70 76 . 0  42 . 0  74 . 5  65 . 0  1 . 7  Weitnauer 1947 
(Moravia) 20 . 1  2 . 71 70 . 7  42 . 3  88 . 6  63. 0  1 . 7  Pellantova 1975 
w 
w 
TABLE Dl cont ? ?  
Incubation Mean Hatching Fledging Fledging Breeding Young Fledged 
Period Clutch Success Period Success Success /Breeding 
Species (days ) Size (%) (days ) (%) (%) Effort Source 
Alpine Swift 
Apus melba 21 . 0  3 .00  c . 59 . 0  81 . 0  70 . 0  1 . 8  Lack & Arn 1947 
4 . 00 86 . 0  - 60. 0  - c . 2 . 4  Lack & Arn 1947 
White-rumped Swift 
Apus caffer (Kenya) 23 . 0  2 .00  88 . 7  42 .0  86 .0  75 . 0  1 . 5  Moreau 1942b 
( South Africa) - - 81 . 0  46 . 0  70 . 3  57 . 0  - Schmidt 1965 
(Rhodesia) 
House Swift 
Apus affini_? ( India) 
(Tanzania) 
(Malaysia) 
Fork-tailed Swift 
Apus_l2?.Q:i:ficus (nat . cliffs ) 
(walls ) 
African Palm Swift 
Cypsiurus parvus 
Barn Swallow 
Hirundo rustica 
Pacific Swallow 
Hirundo tahitica 
Welcome Swallow 
Hirundo neoxena 
Mangrove Swallow 
Tachycineta albilinea 
Grey-breasted Martin 
Prog:Qe chalybea 
24 . 5  
22 . 0  
23 . 0  
21 . 1  
16 . 9  
16 . 9  
20 . 0  
15 . 0  
16 . 2  
15 . 8  
17 . 0  
16 . 7  
2 .00  9 1 . 0  42 . 0  
c . 3 .00  - 40 .0  
2 . 50 89 . 2  38 . 5  
3 . 70 89 .4  43 . 6  
2 . 10 58 . 8  40. 4  
2 . 30 75 . 7  40 .4  
1 . 90 66 . 0  32 . 9  
2 . 00 - -
4 . 40 78 . 2  18 . 5  
2 . 98 54 . 0  19 . 8  
3 . 90 71 .4  c . 21 . 0  
4 .00  62 . 5  25 .0  
3 . 30 94 .0  26 . 5  
60 .0  55 . 0  1 . 0  Brooke 1957 
- - - Ali & Dillon 1970 
- - - Moreau 1942a 
- - - Chih-Tung 1973 
63 . 6  37 .4  0 . 8  Yuren & Ben.xiang 1985 
63 . 6  48 . 1  1 . 1  
26 .0  17 . 0  0 . 3  Moreau 1941 
- - 1 . 9  Brooke 1963 
91 . 7  71 . 9  3 . 1  Adams 1957 , Bent 1942 
69 .0  37 . 6  1 . 1  Hails 1984 
73 . 8  52 . 7  2 . 0  Marchant & Fullagar 
1983 
- - 1 . 0 Dyrcz 1984 
- - 2 . 4  Dyrcz 1984 
-------------------------------------------------------------------------------------------------------------------------? w 
-!>-
3 5  
Both birds usually brooded, or at least sat on the nest, at night. By 
contrast, in some hirundines, for example, the Pacific Swallow (Hirundo 
tahitica), males do not help incubate (Hails 1984) . 
Sex determination 
Being able to identify the sex of individual breeders can help in 
understanding breeding behaviour but sexing swifts externally is 
difficult (Brooke 1971b) . The following characters were measured and 
analysed to detect any sexual dimorphism in unsexed birds : exposed 
culmen, wing length, body length, tail length, outer rectrix length, 
emargination of the outer rectrix, weight and intensity of plumage 
colour. None of these demonstrated any dimorphism that might be 
sexually based. As the following examples show this situation is normal 
in apodids. An examination of the Chimney Swift for 12 physical and 
four behavioural characters that may vary diagnostically between the 
sexes found that none was reliable. (Fischer 1958 , Zammuto & Franks 
1979 ) ? Brooke ( 1971b )  found in a number of swifts that both sexes had 
similar ranges for the standard measurements, though the females 
averaged heavier. Collins ( 1968a , 1972 ) , however, found males to 
average heavier in the Chestnut-collared Swift and the White-tipped 
Swift. It has been claimed that the emarginate tail of the male Black 
Swift (Cypseloides niger) sometimes allows sexual identification even in 
the field (Rathbun 1925 ) . However, Brooke ( 1969 ) and Bent ( 1940) use 
the elongated and emarginate fifth rectrix of this species to separate 
non-juveniles from juveniles. The point is that there is no external 
character that can be used to differentiate the sex of all individuals 
in any species of swift . 
This lack of variation in coloUl? and size has even been considered 
fortunate by taxonomists (Oberholser 1906 ) who appreciated that a lack 
of sexual variation reduced the complicati?ns of what was already a very 
difficult group to identify. 
3 6  
When I graphed the wing-span of 54 assimilis known to be older than 
18 months , a slightly bimodal distribution appeared , but when I found a 
dead male and female ( sex determined by examination of gonads ) each with 
wing-spans in the upper quartile it became clear that this method could 
not be used to sex this species . The wing-span measurement has been 
useful in sexing Kestrels (Cam & Cam 1975 ) and Lewin Honeyeaters 
(Smedley 1977 ) . In graphing 176 wing-span measures from free-flying 
birds of unknown age no bimodal form was evident . Three more birds with 
wing-spans above the average ( 261 . 4  mm) proved by gonad inspection to be 
two females and one male . 
Egg loss and replacement 
In assimilis , the first egg to replace lost clutches or even lost 
broods of up to 22 days of age was laid within 10 to 17 days of the 
loss . The average time taken was 12 . 6 days ( n = 22)  . It was assumed 
that replacements were made by the original pair using the nest . One 
pair was not seen to re-lay in the 27 days after the loss of their 
clutch - the time when observations ceased . But as two birds were found 
on the nest 10 days after the loss , it is possible that ? replacement 
egg was lost before I could see it . 
In chillagoensis twelve out of fourteen ( 8??) first clutches were 
replaced an average of 9 . 4  days (range = 6 - 14 days ) after they had 
been lost or removed . The other two clutches were not known to have 
been replaced . As this average time of replacement was not 
significantly different ( t  = 1 . 8 ,  df = 15 , ns ) to the average elapsed 
time for replacement of lost replacement clutches nor to the average 
time for replacement of lost broods ( t  = 0 . 63 ,  df = 12 , ns ) ,  all three 
averages have been pooled . The average time for the replacement of all 
lost clutches ( that were known to have been replaced) ,  and broods less 
than ten days old , was 10 . 4  ? 0 . 4  days (Range = 8 - 18 days , n = 41 ) .  
3 7  
This range is similar to that of Fijian birds ( 10 - 17 days ) although 
the lower minimum time may indicate that chillagoensis has larger body 
reserves , is able to gain the required food more quickly , or has a more 
active ovary than assimilis . 
When chillagoensis lost chicks older than 20 days , the nest either 
already contained the second egg , or the replacement egg appeared in 
less than six days , probably as a result of previous development of a 
follicle in the ovary in preparation for the first chick to incubate it . 
Of 58 eggs produced by chillagoensis in 1985/86 , 69% (40) were 
successfully hatched . In the dry season of 1986/87 , 6?? (40) of 67 eggs 
were hatched . The average hatching success for both years was 64% (n = 
125 ) . This compares with 58% (n  = 159 ) of eggs successfully hatched by 
Fijian swiftlets . Of those that failed to hatch , 78% disappeared ,  
compared with 88% in Fiji . Both these figures are a little higher than 
the 64% similarly lost by the Edible-nest Swiftlet (Langham 1980) . 
However if we take account of the 17% of Edible-nest Swiftlets lost 
through nest harvesting and nest fall , the results are very close . 
Another 11% were infertile ( 1?? in Fiji ) , and 11% broke ( 2% in Fiji ) . 
By comparison 14% of Edible-nest Swiftlet eggs failed to hatch (Langham 
1980) . 
That some of the eggs that disappeared were infertile , cracked or addled 
is possibl? for Lack & Lack ( 1952 ) found that Common Swifts usually 
ejected cracked eggs and sometimes ejected infertile eggs . 
None of the manipulated clutches of two or three was included in the 
above egg losses , but it is interesting to note that one manipulated egg 
hatched five to ten days after the first , while others remained six , 
ten , eleven and fifteen days after the hatching of the first chick 
before disappearing . 
3 8  
Extra parental ' co-operation ' or ' egg dumping' ?  
In Fiji several events have suggested that occasionally more than 
one female may lay in one nest . Eleven days after one of two eggs were 
taken from nest 108 , a replacement was laid . The first egg hatched six 
days later and the second egg disappeared ,  suggesting that the first 
female had been incubating for some time prior to the laying of the 
third egg . In the second event , tru?ee eggs were laid in nest 86 , one on 
the 20th + 1 ,  one on the 23rd ? 1 ,  and the last on the 27th ? 1 ,  
December 1981 . In that one egg disappeared four days later , another two 
days after that and the third within another three days , it is probable 
that a second female was involved and squabbling at the nest meant all 
the eggs were lost . 
Such incidents are not without precedent in the Apodidae . Nine eggs 
laid in 14 days by a Short-tailed Swift (Chaetura brachyura) were 
assumed to be the product of two females , as one female normally lays 
every second day . All nine eggs were successfully hatched even though 
they were two or three deep in the nest ( Snow 1962 ) . Collins ( 1968a)  
once found two eggs had been added to  a nest of this spepies somewhere 
between 4 and 16  days after the first female had commenced incubation of 
four eggs . This meant four chicks had considerable advantage over the 
other two , which died within a day of hatching . That evening three 
adults were found roosting near the nest , one of which was a yearling 
raised in that same nest cavity . Three eggs laid in the short period of 
three days by the Ashy-tailed Swift ( Chaetura andrei )  and found broken 
later on the third day , were also attributed to two females ( Sick 1959 ) . 
Clutches of eight and nine produced by the C?tmney Swift and 
attributed to two females were found by Fischer ( 1958 ) . However , extra 
parental co-operation is known to be well  established in the Chimney 
Swift (Dexter 1952 ) . 
3 9  
Snow ( 1962 ) proposed that the reason for multiple female use of the 
one nest by Short-tailed Swifts was a shortage ? of holes suitable for 
nesting . This is not applicable to the White-rumped Swiftlet in Fiji 
( or Chillagoe ) ,  as there is still considerable room left in each cave . 
Three other causes are possible . As nests take one or two months to 
construct , a fallen nest or the failure of a pair to construct one may 
pressure a bird to lay in the nest of another . Secondly , as these nests 
are in total darkness , the normal ability to recognise nests may fail . 
Thirdly , birds may purposely ' dump ' their egg in another bird ' s  nest . 
There is only one other record of a swiftlet nest having an extra 
' egg ' . One of the eight nests in a Glossy Swiftlet colony in New Guinea 
had two eggs and a newly hatched young (Mackay 1968) . Whatever the 
origin, a third egg in the nest of any swiftlet so far studied is 
uncommon and , when it does occur , there is no clear evidence to suggest 
that one female was responsible for laying all three eggs . 
The situation of extra eggs in the nest at Chillagoe is unique and 
so is developed as a separate topic in Section 3 .  
Nestling development 
In Fiji the second chick usually hatched within a day and a half of 
the first , even though the laying interval averaged four days . 
This is because the first egg is left unattended during the day until 
the second is laid . One exception to this occurred where I had 
introduced a second egg into a nest . It hatched five to ten days after 
the first and grew normally until a much larger chick moved into the 
nest and the first two starved to death . 
The average weight of newly hatched assimilis chicks was 1 . 14 + 0 .03 
g (range = 0 . 9  to 1 . 4  g ,  n = 23 ) , while that of chillagoensis was 1 . 31 + 
0.01 g ,  (range = 1 . 09 to 1 . 60 g ,  n = 52 ) . This difference is 
significant ( t = 5 . 1 ,  f<O .  001 , df = 73 ) ? One remaln:lng egg she 11  from 
assimilis weighed 0 . 2  g .  
4 0  
Usually egg shells rapidly fragmented with the pieces remaining in the 
nest for one to three days . Whether it was the adults , American 
Cockroaches ( Blattaria americana) or chicks that removed the pieces is 
uncertain . 
All swifts are naked upon hatching and most remain naked until their 
flight and contour feathers emerge . However , the three tree swifts 
(Hemiprocnidae ) and the African Palm Swift (CYPsiurus parvus) develop 
nestling down, presurrably for warmth and camouflage as they nest in 
exposed sites (Lack 1956a) . Down also develops in the Black Swift (Bent 
1940) and the Chestnut-collared Swift (Snow 1962 ; Collins 1963 , 1968a) . 
Snow ( 1962 ) and Collins ( 1968a) suggest that the down is needed in these 
two swifts because the nest-sites selected by them are cooler ( 18 . 8  to 
26 . 2  ?c) than the surrounding habitat . That the parents consistently 
brood the young until the down grows , supports their suggestion . The 
young of the House Swift (Moreau 1942a) , the Whi te-rumped Swift (Moreau 
1942b) and the Pallid Swift (Apus pallidus ) (Kainady 1976 ) , are also 
brooded for at least the first week ,  by which time downy semi-plumes 
appear on restricted parts of the chick ' s body . The temperatures near 
the nests of the African swifts range between 15 ?C and 
.
41 ?C (Moreau 
1942a; Taylor 1942 ) . This is more extreme than the range of 24 . 8  ?C to 
26 . 4  ?c in the caves at Nasinu , Fiji . Hence downy plumage grows on 
those species nesting in locations that experience cooler temperatures 
than do the swiftlets . 
In assimilis the developing feathers appear as dark dots along the 
feather tracts a few days after hatching . The first to break through 
the skin are the remiges , which do so on the eleventh day (range = 8th 
? 15th day) ; though in one manipulated brood they never broke through in 
the third chick until the 17th day. 
4 1  
The average wing length at the time of emergence was 15 . 6  .::!:. 0 . 42 mm (x .::!:. 
SE , range = 12 - 20mm) and even the manipulated late sprouter in the 
exceptional brood mentioned above had a wing of 14 mm when the pins 
broke through. The eruption of the remiges was only one day later than 
the single-brooded Black-nest Swiftlet (Medway 1962a) , but seven or 
eight days later than in the Short-tailed and Chestnut-collared Swifts 
(Collins 1968a) . 
In assimilis the average time taken for the first flight feathers to 
break out of their sheathswas 18 . 6  + 0 . 31 days , (range = 15 - 27 days , 
n=61 ) , when the wing measurement was between 25 and 39 mm ( 32 . 1  .::!:. 0 .  46 
mm, n=61 ) .  In chillagoensis the average time for this event?ds 18 . 7  + 
0 . 4  days (range = 16 - 24 days , n = 24) , when the wing measurementllqS 
between 25 and 36 mm ( 29 . 3 .::!:. 0 . 6 , ) .  It is interesting that although the 
newly hatched chicks at Chillagoe were significantly heavier than newly 
hatched Fijian chicks , the time taken for the first flight feathers to 
eruptwas not significantly different ( t  = 0 . 2 ,  df = 52 ) . The eruption 
of flight feathers in both subspecieswds six days later than in 
Short-tailed and Chestnut-collared Swifts (Collins 1968a) . 
Swiftlets use their feet to hang on to the small nest during the 
nestling period and most chick deaths occur because the chick had fallen 
from its nest . This probably explains why a grasping reflex was 
developed in day-old chicks and why in assimilis the tarsus reached full 
adult size by the 14th day . By contrast , adult weightwas not reached 
until the 23rd day ( 21st day in chillagoensis )  ; and adult wing size in 
both subspeciesw?s only reached after fledging . 
Chick begging calls are described in Tarburton (Appendix 7 p . 218 , in 
prep . ) .  
The nestling period 
Twenty one assimilis chicks fledged at an average of 46 days (range 
= 41 - 51 days ) .  
42 
Eighteen chillagoensis chicks fledged at an average age of 47 . 9  + 0 .7 
days (range = 43 - 51 days ) , which is similar to the age at which Fijian 
birds fledge . The 21  Fijian chicks had at last measurement a wing 
length of from 99 to 107 rrm . The average wing length of Chillagoe 
chicks at the time of fledging was 101 . 6  mm which is 95% of adult wing 
length. The minimum wing length at fledging (range = 91 . 3  - 108 mm) , 
was 85% of adult wing length. This is just 3% below the minimum I found 
in Fijian birds and that which Lack & Lack ( 1951 ) found in the Common 
Swift . As the wing was growing from 2-3 . 5 rrm a day in the nestling ' s  
last week , it would take only two to five days for the fledgling ' s  wing 
to approximate the average adult length . 
The nestling periods of 46 days for assimilis and 46 . 9  days for 
chillagoensis are similar to those of other Apodidae having similar 
incubation periods . Most other apodids and all the hirundines shown in 
Table IV have shorter nestling periods . The nestling period varies with 
brood size as it does in the African Palm Swift (Moreau 1941 ) , and so 
this is discussed more fully in Section Two . 
The increased safety from nesting in inaccessible places has relaxed 
the selec?ion pressure for raising young as quickly as possible (Snow 
1962 ) . The White-rumped Swiftlet and other cave-nesting species have 
relatively long incubation and fledging periods compared to those 
species , such as the Common and House Swift , that do not nest in caves 
(Table IV) . This slow egg and nestling development has possibly been 
reinforced by the swiftlets ' practice of breeding in large colonies , 
forcing birds to forage further afield and to feed the chicks less 
frequently than do the swifts . 
Fledging success 
Despite the low rate of hatching in assimilis compared to other 
Apodidae , their fledging success of 92% was surpassed only by the 
fledging success of the Chimney Swift ( 96 . 1%) . 
4 3  
The success of other Apodidae (Table IV) ranged between 26% and 8?h. 
The fledging success of 69% (n = 29 ) for chillagoensis in the 1985/86 
season was well below that for Fijian birds . However , because of the 
high hatching rate the breeding success was 44% , which is only 9% below 
that of the Fijian birds . Chillagoe birds raise a second brood whereas 
the Fijian birds do not . This means the Chillagoe birds raise an 
average of 0 . 9  chicks per season which is very close to the average of 
1 . 1  chicks that Fijian birds raise per season . However , in poor seasons 
such as the 1986/87 season , fledging success in chillagoensis was only 
500h (n = 18) and with hatching success reduced from 69% to 600h ,  the 
breeding success achieved was only 300h a brood and therefore only 0 . 6  
chicks per season. 
In producing 1 . 1  fledglings per clutch (breeding effort ) ,  the 
production by assimilis equals that of the Short-tailed Swift , the 
Glossy Swiftlet and the Pacific Swallow (Table IV) . Although the Glossy 
Swiftlet has three breeding peaks per year there is no evidence that the 
same birds produced any more than one clutch (Medway 1962a) . Of the 
three swiftlets producing fewer fledged young per brood than the 
White-rumped Swiftlet , the Black-nest and Mossy-nest Swiftlets are known 
to produce second clutches as well as replacement clutches (Medway 
1962a) and the Edible-nest Swiftlet is presumed to lay second and third 
clutches (Langham 1980) . 
Chick mortality 
While the majority of chicks were in their nests at each visit and 
stayed there during four separate days of nest watching in Fiji , some 
occasionally fell from the cup and spent considerable time clinging to 
the underside of the nest . When small , such chicks were unable to 
return to the nest cup and fell to the floor . Chicks presumably fell 
out of the nest during or shortly after feeding visits or other periods 
of activity. 
4 4  
My chick measuring visits certainly increased the numbers ' falling ' out 
of the nest during the first few minutes after my visit . This 
necessitated my keeping an eye on each nest for two or three minutes 
after I had returned the chicks . 
In Fiji only three chicks were found dead in the nest . Two of these 
had been starved and trampled by a larger chick that intruded into their 
nest . The disappearance of most chicks must have resulted from falling 
out of the nest , as there was no evidence of predation in the colonies 
which were in totally dark sectors of the caves . Chicks were seen 
dangling below nests and numerous dead or dying chicks found on the cave 
floor . Some chicks seemed prone to falling or being knocked out of 
their nest . For example one chick was replaced three times in one day 
and on three other days . One chick , rescued from the ground three 
metres below its nest , grew to fledging , but normally most fallen birds 
disappeared within a day . 
An analysis of the age of 69 chicks that disappeared from the nest 
revealed that younger chicks were most susceptible . Thirty-three 
percent were less than 10 days old , 4ry? were between 10 and 20 days old . 
The question arises , were those that fe 11  from nests with broods of two 
or three more likely to be the younger chick in the brood? The answer 
is no . Twenty-eight percent of the 69 chicks that disappeared were sole 
occupants , so siblings could not have caused their fall . Thirty-four 
percent were the smaller of the two chicks (including 1% that were 
older ) . Twenty-six percent were the larger of the siblings ( including 
1% that were younger ) .  The remaining 12% were equal to their sibling at 
the time they were last measured before their disappearance . So , while 
the first chick to hatch usually remains the larger , being older gives 
it no clear protection against falling from the nest . 
During nest inspection , most sitting and brooding birds flew from 
the nest at my approach. 
4 5  
A few birds sat tight , even while I felt under them for eggs , but even 
these flew if I took the eggs or chicks out of their nest . At one nest 
a parent was often found in attendance , even brooding the chicks on five 
occasions . On the last of such occasions the older chick was 24 days 
old . 
Most chicks that died at Chillagoe , did so , as did the Fi jian 
chicks , from falling out of the nest . Because of the chicks ' resistance 
to starvation, many of the chicks picked up from the ground and placed 
in nests recovered . However , contrary to the Fijian situation , most of 
the chicks which fell  to the ground had disappeared by my next visit . 
It is presumed that Childrens Pythons (Liasis childreni ) and Brown Tree 
Snakes (Boiga irregu1aris )  were responsible for their disappearance as 
these were seen near nest sites in some of the caves . Rats may also 
prey on fallen chicks for John Barton of Queensland National Parks and 
Wildlife Service has reported seeing rats in several caves . There was 
no evidence of predation upon chicks in the nests , almost all of which 
were in totally dark sectors of the caves . Confirmation that snakes 
cannot reach most nests , which are high up on smooth concave surfaces ,  
comes from Vince Kinnear (pers comm) , who has watched Childrens Pythons 
fall off the smooth wall as they have tried to reach nests . 
Subsequently , I have observed a Childrens Python on the wall adjacent to 
the nests in Mudlark Cave , but as all the nests still  contained their 
egg or chick , it was evident that the python could not cross the smooth 
overhanging wall that surrounded the colony. I have also seen a 1 .  5 m 
Brown Tree Snake fall from the smooth wall above the nesting colony in 
the New Southlander Cave . Evidence that the tree snake could only cross 
the smooth wall where it did not overhang was that there were only two 
empty nests and these were both on vertical wall . In Fiji fallen chicks 
were still to be found on the cave floor far more fequently than at 
Chillagoe . 
4 6  
This is  probably because Fijian Pythons Enygrus bibronii (presumably due 
to the activities of the Mongoose ) and rats ar? uncorrrnon in caves. I 
only ever saw one python and one rat in Fijian caves. 
Adult mortality at the nesting caves 
Although snakes cannot reach most swiftlet nests adult birds are not 
free from predation by snakes. I have watched Childrens Pythons 
position themselves on the wall adjacent to narrow sections of cave 
passage through which swiftlets must fly to reach their nests . Numerous 
elongated seats containing swiftlet flight feathers in the vicinity of 
such pythons in Golgotha Cave, Project 31 Cave, September Cave and in 
the tube connecting Christmas Pot and Squeeze Pot, indicated that the 
python ' s  ambush method was successful. 
There are also other predators at Chillagoe. The entrance to 
Swiftlet Cavern is a low crawl from a daylight chamber of Royal Arch 
Cave and the ground is littered with swift let feathers. John Barton and 
Lionel Leafe ( cave guides with QNPAWS ) had seen cats in the vicinity and 
were convinced they were preying on the swiftlets. A pile of feathers 
but no seats is indicative of cat predation. That the subsequent laying 
of baits for the cats was successful is indicated by the reduction of 
swiftlet feathers at the cave entrance and the increase in the size of 
the breeding colony in the second season ( 1986/87) .  Similarly fresh 
feather piles were observed at Clam Cavern, Swiftrimlet cave, Hercules 
Cave and Project 31 Cave. A decaying pile of feathers was seen in the 
entrance to September Cave indicating that cats had discontinued killing 
swiftlets at that site. 
On 25 January 1986 I was watching about 50 swiftlets feeding very 
close to some rocky pinnacles, above Otobeaswiftlet Cave, which I had 
discovered just that day. There had been an exchange of birds through 
the cave entrance when a Brown Goshawk (Accipiter fasciatus ) landed in a 
nearby tree . 
4 7  
The birds responded suddenly and in unison by rapidly dispersing in all 
directions just as noddies ( pers . obs . ) and terns ( Cull en & Ashrnole 
1963) do in their "dread" or "panic" behaviour . The noise of their 
wings made a loud swish , which I have only heard when much closer to 
individual birds making their power dives into or from cave entrances .  
However in this event I could hear the noise from most birds , many of 
which were 20 to 40 m away . About five to ten minutes later the 
swiftlets began to reassemble overhead at an altitude of about 130 m 
from where they made a call (a  high pitched 1 shree 1 ) I had not heard 
before . 
A Brown Goshawk had been seen outside the swiftlet entrance to Guano 
Pot on 20 January 1986 , where after sitting in a tree it appeared to 
make a stoop at a swiftlet that was flying steeply and slowly out of the 
cave entrance . The next day another Brown Goshawk sat outside the 
entrance to Tarby1 s Swiftlet Pot . All three observations of Brown 
Goshawks near entrances to Swiftlet breeding caves were at the time when 
most first brood chicks were taking their first flights . The young 
swiftlets are much slower than adults when leaving cave entrances and it 
is likely that Goshawks are able to take young birds in this situation. 
On 4 January 1987 I saw a Brown Goshawk put the swiftlets over Gordale 
Scar Pot and Tar by 1 s Swift let Pot into 1 dread 1 behavour when it caught 
an individual out of the flock. Four days later an unidentified falcon 
circling approximately 80 m above the same two caves caused abnormal 
behaviour (mainly erratic flying) in the swiftlets and so it is possible 
that falcons as well as owls and goshawks take swiftlets . 
Birds of prey have been seen taking two other species of swifts in 
similar situations and the same hesitancy shown by the swiftlets to 
enter small openings was displayed by the swifts . 
48  
The first instance involved a flock of 1 , 500 Chimney Swifts entering a 
chimney for roosting purposes , when a Sharp-shinned H:awk (Accipiter 
striatus) flew into the flock at the point where they slowed down to 
drop into the chimney and carried away a squealing swift (Musse lrran in 
Bent 1940 ) . The second instance occured when a flock of Vaux ' s  Swifts 
were entering a chimney and a Merlin (Falco colurnbarius) rrade a stoop 
which due to the swifts scattering was unsuccessful (Rathbun in Bent 
1940) . It appears that adult colonial swifts are most vulnerable while 
entering or leaving their nesting and roosting sites . It is only in 
this situation that the birds show hesitancy by flying circuits just 
inside or outside the entrance . That the entrance or exit is performed 
at high speed once initiated is further evidence of the swiftlets ' 
vulnerability to predators at this point . By timing chillagoensis over 
a measured eight metres at the entrance to Tar by ' s Swift let Pot , the 
average speed was determined to be 37 km/hr . The maximum speed recorded 
was 111 km/hr which is similar to that ( 106 km/hr) recorded for 
assimilis entering Waterfall Cave ( Tarburton 1986b , Appendix 5 p . 180) . 
Nest sanitation and ectoparasites 
From the first day , nestling White-rurnped Swiftlets defaecate over 
the low front rim of their nest , thus leaving the nest free of faecal 
contamination. Similar behaviour has been described in the Common Swift 
(Lack 1956a ) , the Chimney Swift (Fischer 1958) , the Chestnut-collared 
and Short-tailed Swifts ( Collins 1968a) . When newly hatched Chimney 
Swifts did not void their faeces over the front of the nest the parents 
either ate the faeces or threw them out of the nest ( Fischer 1958) . 
Common Swifts were not so clean , as the adults would not remove old 
faeces from their nests and towards the end of the nesting period the 
parents also became less regular in removing fresh faeces (Lack & Lack 
1952 ) . 
4 9  
The only fouling found in Whi te-rumped Swift let nests was the broken 
remains of' hatched egg-shells but even these were either eaten or 
removed in many nests . It is possible that the American Cockroaches 
were responsible for removing some eggshell .  The eggshell was not 
removed from the nests of' the three species of swif'tlet that nest in 
Niah Cave , Borneo (Medway 1962a) . Neither in Fiji nor Borneo were dead 
eggs left in the nest for very long and Weitnauer ( 1947 ) , Lack & Lack 
( 1952 ) and Lack ( 1956c ) record the ejection of' eggs by the Common Swift . 
While the chick ' s  practice of' def'aecating over the f'ront of' the nest 
provides nutrient for the guanophile community on the cave floor , or for 
the aquatic community downstream, it deprives those ectoparasites that 
thrive in fouled nests , of' an important source of' food . 
Both assimilis and chillagoensis are parasitised by louse-flies 
(Hippoboscidae ) .  When adults left their nests these flies were often 
lef't behind and they moved rapidly over the nests , cave wall and the 
chick. They also moved rapidly over and through the feathers of' adults 
in the hand . In repeatedly examining chicks from more than lOO Fijian 
nests through December 1981 , the few chicks in nests isolated by more 
than 0 . 5  m f'rom other nests had no louse-flies . Birds in nests having 
contact with surrounding nests had the most flies . 
The fly found in Fiji was separated as a new species (Myophtheria 
f'ijiarum) and placed in the new sub-genus Myophthiria (Maa 1980 ) . This 
sub-genus was used to group 11 species of' flies hosted by several 
species of' swif'tlets distributed on islands or archipelagoes between 
Sri-lanka and Fiji and at the same time separating them from two louse 
fly species hosted by three new world swifts (Black Swift , 
White-throated Swift (Aeronautes montivagus) and White-tipped Swift 
(Aeronautes saxatalis ) . Maa ( 1980 )  quotes Rondani 1878 as cause for 
listing M .  f'ijiarum as being hosted by the "Mossy Swif'tlet" as well as 
the White-rumped Swif'tlet on Viti Levu. 
s o  
However , the Mossy-nest Swiftlet does not occur in Fiji . Consequently 
this species of louse-fly is hosted only by the White-rumped Swiftlet in 
Fiji . Only one of the other 10 species of louse-flies named by Maa is 
hosted by another subspecies of the White-rumped Swiftlet . This fly is 
M.  neocaledonica, hosted by A. s. leucopygius on New Caledonia .  The 
louse-fly M.  queenslandae which Maa suggests is hosted by either the 
White-rumped Swiftlet or the Glossy Swiftlet , is probably the species 
found on terraereginae (Pecotich 1974) . This means three species of 
,-.'.;r 
louse-fly have each been found on one of three separate subspecies of 
the White-rumped Swiftlet . 
Fifty-six louse-flies collected from Nasinu Caves in 1974 and 
December 1981 , were five to seven mm long and weighed 0 .016 - 0 . 044 g.  
Their wings were only one mm long. Having such short wings probably 
explains why the flies were found in the nests as well as on the chicks 
and adult birds . When moving between nests and over the birds the flies 
moved very rapidly . When they stopped they were usually well hidden in 
a crevice or under some feathers . However those on birds that were 
being handled were easy to flush and so easy to catch. 
Although a search for other ectoparasites was not made it is 
possible they exist as a number of others have been found on other 
apodids , some of which move through or inhabit the south-west Pacific . 
Russian workers have found ticks ( Ixodidae ) on the Fork-tailed Swift 
(Apus pacificus) which winters in the south-west Pacific (Bolotin & 
Kolonin 1979 , Kolonin & Bolotin 1977 ) . Ticks have also been found on 
the Glossy Swiftlet in Java (Hoogstraal et al 1974) and the Alpine Swift 
(Apus melba) in Eurasia (Tovornik & Cerny 1974 ; Filippova & Panova 
1975 ) . The ticks on the Glossy Swiftlet may be significant as the 
parasitic louse-fly from the swiftlets in Java are the most closely 
related to those on the Fijian swiftlets (Maa 1980) . 
5 1  
Feather-mites (A carina) have been recorded from African swifts 
including the House Swift (Gaud & Atyeo 1976)  and the African Palm Swift 
( Cort 1975) ? Mites have also been found on the Short-tailed Swift and 
the Chestnut-collared Swift in Trinidad (Collins 1968a) . Such a wide 
distribution on apodids suggests they could be expected on swiftlets in 
the south Pacific . Feather-lice (Mallophaga) have been reported on the 
two swifts studied by Collins ( 1968a) in Trinidad as well as on the 
Chimney Swift , along with parasitic bugs ( Cimicidae , Hemiptera) in North 
America (Fischer 1958 ) . Lice are common ectoparasites on birds and 
being small could have been easily overlooked on the Fijian swiftlets . 
A variety of insects inhabit the swift let nests at Chillagoe . Apart 
from the Louse-fly , none was abundant except towards the end of January 
when a tiny feather louse covered some chicks . The largest ectoparasite 
seen was the Louse-fly Myophthiria spp. rvraa ( 1980)  stated that flies in 
this genus are never found in high densities ,  hence are very difficult 
to collect . However this Louse-fly was common on adults and chicks at 
Chillagoe . By collecting and recording the number of Louse-flies found 
on the chicks or in their nests , the data in Table V were accumulated . 
Table V 
The numbers of Louse-flies found on Chicks at 10 day age intervals .  
Period (days) 0-10 11-20 21-30 31-40 41-50 
Total from 142 Fijian nests 0 . 0  1 . 0  12 . 0  6 . 0  1 . 0  
x/nest at Chillagoe 1 . 0  1 . 5  5 . 9  4 .0  3 . 1  
n 46 . 0  38 . 0  26 . 0  20. 0  10 . 0  
SE 0 . 21 0 . 23 0 . 78 0 . 74 0 . 78 
5 2  
These data indicate that this ectoparasite is least corrmon on 1 - 10 
day-old chicks and not significantly more commcin on 11 - 20 day-old 
chicks (t  = 1 . 5 , ) .  The flies are significantly more common on 21 - 30 
day-old chicks than on 11 - 20 day-old chicks (t  = 5 . 5 ,  P<0 .001 ) . The 
decline in numbers found on 31 - 40 day old-chicks , and the further 
decline on 41 - 50 day-old chicks , are not significant (t  = 1 . 8  & 0 . 8  
respectively) . In spite of this decline , the average number of 
Louse-flies on 41 - 50 day-old chicks was still significantly greater (t 
= 2 . 029 , P<0. 05 )  than those found on 11 - 20 day-old chicks . This peak 
of density on 21 - 30 day-old chicks coincided with that for the 
Louse-fly cogener in Fiji . However , the fly was much less corrmon in 
Fiji where the numbers shown in Table V constitute the total for 142 
nests rather than the average per nest as shown for the Chillagoe birds . 
The average distance between nests at Chillagoe was greater than 
that between Fijian nests and should have proved a greater deterent to 
the spread of ectoparasites such as the Louse-flies . Because many nests 
were joined together in Fijian colonies , the spread of Louse-flies 
should have been facilitated , so finding them to be less common on 
Fijian chicks is surprising , although they were not as uncommon on adult 
Fijian swiftlets (though unfortunately uncounted) as on Fijian chicks . 
Feeding of young 
From observations on 20 assimilis nests for one full day in December 
1981 a daily feeding rate of 2 . 8  ? 0 . 26 (x ? S . E . ) was determined . 
Although six of these nests contained only single chicks the number of 
visits to these nests was not significantly different to the number of 
visits to those nests with two chicks . Each chick received an average 
of 1 . 7  feeds per day. 
5 3  
From observations on chillagoensis nestlings for one full day during 
both the good season (December 1985 , n = 20) and the poor season ( 22 
December 1986 , n = 9) rates of 5 . 2  ? 0 . 28 (x  ? S .E . ) for the good season 
and 2 . 9  ? 0 . 5 for the poor season were determined . The rate for the 
good season is significantly ( t = 6 .  3 ,  P< 0 .  001 , df = 38) above that for 
Fijian birds ( 2 . 8  ? 0 . 26 ) . The rate for the poor season was not 
significantly different ( t = 0.  4 ,  ns , df = 25 ) ? That Swift lets at 
Chillagoe have an easier time providing for their single brood in good 
seasons , than Fijian birds in providing for their brood of two , is also 
suggested by the shorter daily period the Chillagoe birds spend foraging 
(Section 2 ) . However there are two factors which prevent chillagoensis 
from raising two chicks . The first is that the density of available 
prey is much below the density of that available in Fiji (Appendix 3 
p . 158) . The second is that the irregularity of extreme wet and dry 
periods throughout the breeding season at Chillagoe means that prey 
species such as termites are not flying each day for they only fly a day 
or two after each period of rain ends . 
The number of feeding visits other Apodidae make to their chicks 
varies within and between species ( Table VI )  ? Some species such as the 
Alpine Swift (Arn-Willi 1960) , the White-rumped Swift , the Palm Swift 
(both Moreau 1942b) , and the House Swift (Moreau 1942a) , increase their 
feeding rate for larger broods . Some Apodidae such as the Chimney Swift 
(Kendeigh 1952 , Fischer 1958 ) and the Black Swift (Michael 1927 ) , vary 
their feeding rate with the age of the chicks . Other Apodidae such as 
the White-rumped Swift (Moreau 1942b) and the Chimney Swift ( Fischer 
1958 ) , increase their feeding rate prior to sunset , though the amount of 
food brought at each trip decreases ( Fischer 1958 ) . As has been shown 
in the Black Swift (Michael 1927 , Collins & Landy 1968) , more than. one 
bolus of food ma.y be brought to the chick in one visit . Changes in the 
weather may cause variations in the feeding rate as Zammuto et al . 
( 1981) have shown in the Chimney Swift . 
Species 
TABLE VI 
Frequency of Chick Feeding 
Mean 
Mean time Range 
number between between 
Brood visits visits visits 
size I day (rnin) (rnin) 
Mean 
feeds 
I Chick 
I Day Source 
----------------------------------------------------------------------------
White-rumped Swiftlet 
Aerodrarrn.ls 
spodiopygius (Fiji )  1 . 7  2 . 8  1 . 7  This study 
(Chillagoe wet yr) 1 . 0  5 . 2  5 . 2  This study 
(Chillagoe dry yr) 1 . 0  2 . 7  2 . 7  This study 
Black-nest Swiftlet 
Aerodrarrn.ls maxirrus 1 . 0  2 . 0+ Medway 1962a 
Mossy-nest Swiftlet 
Aerodrarrus 
vanikorensis 2 .0 c . 3 .0 30-? Medway 1962a 
Chimney Swift 
Chaetura pelagica c .4 .0  c .49 . 0  14. 6  12 . 3  Zammuto 
et al .  1981 
4 .0  1-28 c . l1 . 8  Bent 1940 
1st week c .4 .0  c . 30 . 0  
-c . 7 . 7  Fischer 1958 
2nd week c .4 .0  c. 45 . o  c. 7 . o  Fischer 1958 
day 1-4 20 . 0  -} 
day 5-14 15 . 0  -}  11 . 7  Kendeigh 1952 
day 16-20 60 . 0  -} 
Short-tailed Swift 
Chaetura brachyura 3 . 8  c . 36 . 0  20 . 5  3-45 c . 9 . 5  Collins 1968a 
Vaux Swift 
Chaetura vauxi 4.0  c . 48 . 0  15 . 0  c . 12 . 0  Bent 1940 
c .4 .0  18 . 0  c. 11 . o Baldwin & Za-
ckowski 1963 
4. 0 15-20 c . 13 . 0  Finley & F .  1924 
4 .0  c . 69 . 0  10 . 0  1-21 17 . 3  Davis 1937 
Chestnut-collared Swift 
Cypseloides rutilus 1 . 9  36-100++ Collins 1968a 
Black Swift 
Cypseloides niger 1 . 0  1 . 0  1 .0 Michael 1927 
Corrmon Swift 
Apus a2us wet 2 . 0  7 . 0 3 . 5  Lack & Owen 
fine 2 . 0  14. 0  7 .0 1955 
Pallid Swift 
Apus pallidus 2 .0 c . 17 . 8  c . 8 . 9  Affre & Affre 
Alpine Swift 1967 
Apus melba c .3 . 0  24 .0  8 . 0  Bartels 1931 
2 . 0  7 . 5  3 . 8  Arn-Willi 1960 
3 . 0  8 . 5  2 . 8 Arn-Willi 1960 
White-rumped Swift 
Apus caffer 1 . 0  8 . 8  8 . 8  Moreau 1942b 
2 . 0  13 . 2  6 . 6  Moreau 1942b 
5 4  
Species 
House Swift 
Apus affinis 
White-throated Swift 
Aeronautes am 
saxatalis dusk 
African Palm Swift 
Cypsiurus parvus 
Brood 
size 
1 . 0  
2 .0 
3 . 0  
3 . 1  
4 .0  
4. 0 
1 . 0  
2 . 0  
TABLE VI cont . . . .  
Mean 
number 
visits 
I day 
6 . 6  
10 . 2  
13 . 8  
25 . 8  
Mean 
time 
between 
visits 
(min ) 
118 . 0  
73 . 5  
54. 3  
35 . 3  
Range 
between 
visits 
(min ) 
8-254 
<8-120+ 
<8-120+ 
21 . 5  3 .36 
11 . 5  2 . 27 
12 . 4  c . 60 .0  <15->90 
19 . 3  c.40.o  
Mean 
feeds 
I Chick 
I Day Source 
6 . 6  Moreau 1942a 
5 . 1  Moreau 1942a 
4 . 6  Moreau 1942a 
8 . 3  Chih Tung 1973 
c . 9 . 0  Collins un?
c. 9 . 0  pub . 
12 .4  Moreau 1941 
9 . 7  Moreau 1941 
NOTE: c .  indicates an approxirration derived from data , that did not cover 
the full diurnal period , or that was not precise . 
s s  
5 6  
Interspecific comparisons are most valid when the size of the brood 
is taken into account . This is done in the final column of Table VI 
where the average number of feeding trips lffide per chick per day is 
shown. In spite of the perturbations caused by the above factors , a 
strong inverse correlation (r = -0. 80 ;  n = 18 ; P< 0 .001 )  exists between 
the length of the nestling period and the number of feeds a chick 
receives in a day (Figure 5) . This correlation should not be used to 
suggest that the number of feeds per chick , per day , determines the 
length of the nestling period . That would ignore the role of genetic 
constraints on the growth rates of chicks . Some species develop at a 
slow rate no matter how much they are fed . Ultimately , this may be 
related to selection of growth rates based on food supply. Genetic 
constraints are clearly shown by the ability of the Black Swift to raise 
its chick successfully on one feed (visit ) per day , whereas the 
Black-nest Swiftlet is unable to do so on less than two feeds a day 
(Medway 1962a) ? 
Adult Morphology 
A sample of adults of both sub-species was measured in order to 
compare the wing length and weight of nestlings with that of their 
parents and with other subspecies . These data are summarized in Table 
VII . The average wing length of 300 adult chillagoensis was 107 . 1  .:!:. 0 . 1 
mm (range = 101 - 113 mm) .  This was significantly (t  = 14.4 ,  P< 0 .001 ,  
df = 400) ,  smaller than the wing of Fijian birds . That my average 
measurement is also significantly ( t = 12 . 8 ,  P< 0 .  001 , df = 350) larger 
than that taken by Pecotich ( 1982 ) on the same population is less easily 
explained.? The average weight of 300 adult chillagoensis was 9 . 3 .:!:. 0 . 04 
g (rarige = 7 . 9 - 12 . 1  g) . This is significantly (t  = 15 . 4 ,  P< 0 .001 ,  df 
= 400) heavier than Fijian birds and significantly ( t = 10. 0 ,  P< 0 .  001 , 
df = 350) lighter than the average weight given by Pecotich ( 1982 ) for 
the same population. 
* lie pl'oba.bly v?.Sec::L e-. d..i ?ent l'l'lethod o? MeasureMQ,i-. 
5 7  
60 
? 
V) 
? 50 
......__ 
LJ ? 
0 
'-
QJ 0 
Cl. ? ? -
01 c ? 40 
., e -4-
V) 
QJ ? 
z e 
Et 
30  
? 
0 5 1 0  15 
Fe ed i ng  ra t e  (feeds/chick/day) 
Figure 5 .  Relationship between feeding rate and nestling period in swifts . 
The White-rumped SWiftlet is designated by the hollow circle . The line is 
the reduced major axis (Clymo 1983) . The correlation is significant 
(r = -0. 80 ,  n = 18 , P<0 .001) . 
TABLE VII 
Morphology of four subspecies of the White-rumped Swiftlet 
A.s . spoctiozy?ius A.s .assimilis A.s .terraere?inae A.s . chill?oensis A.s . chillagoensis 
x+S .E .  a x+S .E .  (b) x+S .E .  (c x+S .E .  c) x+S .E .  (b) 
Length (nm) 
Wing (nm) 
Wing Span ( nm) 
Outer Rectrix(mm) 
Tarsus (mm) 
Exposed Cul . (mm) 
Gape (rrrn) 
Mid. Toe (rrrn) 
Mid .Claw (mn ) 
Weight (g) 
n 
102 (d) 
111 . 7+0 . 8  
52 . 8+0 . 6  
8 . 5  ( e) 
4 . 5  (e) 
8 . 1+0 . 4  
8 
103 . 1+0. 4  
112 . 0+0. 3  
261 . 0+0 . 6 
44. 7+0 . 2  
10. 6+0 . 1 
3 . 8+0 . 03 
11 . 2+0. 08 
4 . 7+0 .05 
3 . 8+0 . 03 
8 . 2+0 . 06 
102 
, ____________ 
109 . 3+0. 3  97 . 2+0.4  105 . 8+0 . 2  
110 . 6+0 . 2  102 . 5+0 .3  107 . 1+0. 1  
249 . 9+0 . 5  
52 . 3+0 .3  47 . 3+0 . 2  47 . 4+0 . 2  
10 . 0+0 . 1  9 . 1+0 . 2  
4 . 5+0.01 4 . 5+0 .01 
12 . 2+0 .07 10 . 5+0 . 1  9 . 3+0 .04 
50 50 300 
NOTE : (a) United States National Museum; (b) This paper; (c) Pecotich 1982 ; (d) Armstrong 1932 ; 
( e) Oberholser 1906 . 
lJ1 
(X) 
5 9  
That chillagoensis weighs more than assimilis and yet has a shorter 
wing suggests that selection pressures to decrease wing length have been 
operating on chillagoensis .  Alternatively ,  selection pressures might 
have increased wing length in assimilis as it flies further and longer 
each day than chillagoensis (Appendix 3 ,  p . 159 ) ? It might also be that 
it is to the advantage of chillagoensis in its erratic climate with long 
dry winters to have a good body store of energy. 
Even though a statistically significant difference was shown between 
the size of eggs in the Waiyala and Nasinu colonies , Student ' s  ! tests 
on the wing , weight and wing span of adult birds indicated no 
significant differences . Furthermore , there was no significant 
difference between either of these populations and the sample measured 
from the Ono Cave population. Most of the data from these colonies have 
therefore been pooled , to provide a sample of 102 birds from which the 
standard morphological measurements have been taken for this subspecies . 
Table VII compares these measurements with those of the two Queensland 
subspecies , and those of the Samoan subspecies . 
The Fijian subspecies has the largest wing and tarsus , but the 
second lightest weight of the four subspecies shown. The wing range of 
115-120 mm given for the species by Medway ( 1966 ) , is too high for any 
of the above subspecies . It is also too high for A. s . leucopygius on New 
Caledonia ,  which has a wing range of 95 - 99 . 5  mm (Oberholser 1906) . 
Two individual birds alone with wings of 116 . 5  mm reach the range given 
by Medway. One is a specimen of A.  s .  townsendi from Tonga and the other 
is A. s .  spodiopygius from Samoa ( Oberholser 1906 ) ? The range of wing 
length for the species should be 95 - 117 mm . 
6 0  
CONCLUSIONS 
As swiftlets are smaller than swifts , the prediction that swiftlets 
will lay heavier eggs than swifts , in proportion to adult weight (Lack 
1968) is supported by this study . While larger Apodidae lay larger 
eggs , it is larger adult size that accompanies a reduction in the 
incubation period. The cause for this is probably that larger birds are 
better able to maintain their proportionately smaller eggs at the 
optimum temperature for reducing the incubation period. This effect is 
probably reinforced by swiftlets gaining the increased safety of nesting 
in places that are inaccessible to predators and which Snow ( 1962 ) 
maintains relaxes selection pressures to hatch and raise a brood as 
quickly as possible . By nesting in larger colonies than swifts and on 
"cliff" sites that have greater safety because of their dark location , 
we would correctly conclude that swiftlets will have longer incubation 
periods than swifts . 
Because incubation periods in swiftlets are inversely related to the 
size of the species , and chillagoensis (which is larger than assimilis)  
has a longer incubation period than assimilis , it may be that factors 
such as the cooler temperature in the caves at Chillagoe can affect this 
basic relationship between adult size and incubation period . 
Because the length of fledging periods is related to the length of 
incubation periods and because chillagoensis (which has a longer 
incubation period than assimilis ) , has a similar fledging period to 
a8similis , it is concluded that factors such as the cooler temperature 
in the Chillagoe caves can also affect the basic relationship between 
incubation and fledging periods . If this is so then it is likely that 
the extra feeding visits to chillagoensis chicks ( 1()(Y;6 more than to 
assimilis) , are being used to keep the chicks warm through extra 
feeding. Alternatively , the reason for adults spending more time with 
their chicks at Chillagoe than in Fiji is to keep them warm. 
6 1  
Section 2 
AN EXPERIMENTAL MANIPULATION OF CLUTCH & BROOD SIZE TO DETERMINE 
WHETHER LONG-LIVED TROPICAL SPECIES ARE MAXIMISING THEIR FOOD SUPPLY 
The fact that the number of eggs per clutch generally increases with 
latitude has been confirmed many times (Lack 1954 , 1968) since the 
discovery of the phenomenon early last century (Wagner 1957 ) . However , 
the search for a general theory to explain these observations and the 
exceptions still continues . Hesse ( 1922 ) suggested that the longer day 
experienced in high latitudes increased the amount of food parents could 
make available to nestlings and therefore facilitated the production of 
a larger clutch than related species breeding in the tropics . Lack 
( 1947 : 331 ) pursued this line of reasoning with a study of seasonal and 
regional variations in the number of eggs which birds lay and concluded 
that "in most species clutch-size is considered to be ultimately 
determined by the average maximum number of young for which the parents 
can find food" . 
Such reasoning presumes that tropical birds raise as many young as 
they can nourish. Skutch ( 1949 ) believed this was not the case as some 
tropical birds can feed their brood at a rate faster than nonna.l . This 
occurs where parents are prevented from feeding their brood for some 
time or have laid more eggs than nonna.l . Skutch also found that species 
in which both parents feed the nestlings usually raised no more chicks 
than species where only the female attends the nestlings . Further 
evidence came from birds such as the Senegal Fire Finch (1agonosticta 
senegala) which can raise the large chick of a parasitic weaver in 
addition to and at the same time as raising its own brood (Morel 1967 ) . 
These and other situations have led to several modifications to Lack ' s  
theory. 
6 2  
In one of these modifications , Winkler & Walters ( 1983) state "birds 
lay that number of eggs that result in the parents operating at the 
optimal working capacity" . In stating this , they accept Ashmole ' s  
update ( 1961 , 1963 ) of Lack ' s  hypothesis ( 1947 ) that birds are 
maximising their clutch size in respect to the quantity of food that is 
immediately available during the breeding season . However they also 
allow for parents which may be maximising their reproductive output over 
their entire lifetime rather than within a single breeding attempt . A 
number of authors use the latter view to predict that some brood sizes 
will  be below the maximum that parents can feed in a particular season 
( eg .  Williams 1966 ; Murphy 1968 ; Gadgil & Bossert 1970 ; Charnov & Krebs 
1974; Goodman 1974 ; Pianka & Parker 1975 ) . 
It would be expected that long-lived birds with small clutches would 
be amongst those most likely to be working below their capacity to 
gather food , in order to enhance their lifelong capacity to raise more 
young. Although it has been doubted that a bird presented with an extra 
chick could be expected to attempt to raise it , the more recent view is 
that the manipulative addition of eggs to broods can be useful in 
helping to determine whether a bird is maximising its food supply or not 
at the time of breeding (Klomp 1970) . It has been shown by monitoring 
chick growth in artificially enlarged broods that some birds have been 
able  to gather enough food to raise more chicks than normal at one time . 
However , all such results were obtained with temperate species and no 
such manipulations have been carried out on swiftlets which are 
long-lived and tropical . 
By monitoring hatching success , chick growth and fledging success , 
in normal sized and experimentally enlarged clutches and broods of the 
White-rumped Swiftlet in Fiji , this section of the thesis demonstrates 
the inability of this species to raise more young than is normal . 
6 3  
METHODS 
The subjects of this manipulation experiment were White-rumped 
Swiftlets nesting in ' Dry'  and ' Waterfall '  caves , Nasinu , north of Suva 
(Section 1 ,  page 8 ,  Tarburton 1986) . Daily visits were made to these 
De<eomher 
colonies throughout December 1981 and again inA1983 . On these visits 
eggs were checked for hatching , and measurements were made of each 
chick ' s  wing , tarsus and weight . Six natural single-egg clutches and 37 
natural two-egg clutches were mdnitored as controls along with the 
manipulated clutches . Single-egg clutches were c.,.-ep_-tecl. by removing 
one of the two eggs from 24 normal clutches . The eggs that were removed 
were added, as were another three , to normal clutches to make up 27 
three-egg clutches . Six two-egg clutches were also made up by swapping 
eggs so that none of these birds was incubating its own eggs . The rates 
of growth for 13 one-chick and 39 two-chick natural broods were compared 
with those for 37 one-chick , seven two-chick and 23 three-chick 
manipulated broods . 
Lack ( 1956) reported some parental desertion when the nests of 
Common Swifts were disturbed . To test the effect of disturbance on the 
swiftlets I visited one isolated group of nests less frequently than 
other nests in the study. The chicks at frequently-visited nests grew 
at the same rate as those in nests visited less frequently. Nests were 
approached slowly to allow incubating parents time to leave the nest 
without the haste that caused increased egg loss . After being handled 
most chicks were restless and likely to fall from the nest . This made 
it necessary to keep an eye on a brood for about a minute after 
replacing it . 
Individual chicks within a brood were initially identified by 
placing a daub of dark green quick-drying model paint on the head , 
shoulders or rump. 
6 4  
Once the tibia was large enough to retain a band , individually numbered 
size one aluminium bands from CSIRO , Canberra , ? Australia were used . I 
banded all swiftlets on the tibia instead of the usual tarso-metatarsus . 
Bands applied to the tarsus often slip over the toes (Tarburton , 
unpublished) ? The swiftlets usually stay on the wing all day ( except 
when breeding) and this prevents the band from damaging the thin skin 
and feathers of the tibia . 
The day of hatching was known for only 22 chicks . Their wing , 
tarsus and weight measurements exhibited an even exponential growth form 
through day eight . In order to increase the sample of aged chicks , 
those up to eight days old when found , were aged with a formula derived 
using regression analysis of measurements of known-age chicks . 
Four fUll-day watches were made at a sample of manipulated and 
untouched nests to determine the daily feeding rate of the different 
sized broods . The low beam on a miners lamp was used to make the 
observations . The median test was used to compare the feeding rates 
because it makes no assumptions on distribution and is more conservative 
than the t test used for other comparisons . 
RESULTS 
Hatching Success 
The hatching successes of natural clutches and manipulated clutches 
were not significantly different . Those with a natural clutch size of 
??ss 
one , hatched an average of 0 .43 _+ 0 . 18 (x  + SE , n = 9 ) , while the - " 
manipulated ones hatched an average of 0 . 61 .?. 0 . 12 (n = 20) , (Median 
2 test , X = 0 . 28 ,  n . s . ) .  Pairs with a natural clutch size of two hatched 
on average 1 . 18 ? 0 . 15 (n = 34) and the manipulated twos , 1 . 0  + 0 . 26 (n 
= 6 ) , (Median test , x2 = 0 . 6 ,  n . s . ) .  
The combined hatching success of single-egg clutches was 0. 52 ? 0 . 09 
??q.s 
(n = 29 ) and that of the combined two-egg clutches 1 . 15 + 0 . 14 (n = 40? 
6 5  
showing a clear advantage to the two-egg clutch ( t65 = 3 . 81 ,  P <  0 .001 ) . 
The manipulated three-egg clutches had an average clutch hatching of 
1 . 22 .:!:. 0 . 17 (n = 27 ) , which was not significantly larger than that of 
the two-egg clutches (t55 = 0 . 33 ,  n . s . ) .  
Chick Growth 
Figures 1 ,  2 and 3 ,  show the mean daily increase in length of wing 
and tarsus , and weight for individuals in broods of all three sizes . 
The overlapping standard errors on the wing growth curves indicate no 
significant difference between broods of one and two whereas that of 
three is significantly less than both one and two after the tenth day. 
Average adult wing length was not reached before birds fledged , though 
the minimum adult wing length was reached by most fledglings . 
The weight of chicks in broods of three was also significantly 
lower. After day 12 a significant difference could be detected between 
those of single and two-chick broods , but it was not as great as between 
those of two and three-chick broods . One-chick broods reached adult 
weight on day 19 , twos on day 22 and those of three-chick broods on 
about 30 to 50 days . 
Tarsal growth exhibited the least variation with no consistent 
differences amongst chicks of differing clutch sizes . Of the three 
measures taken, the tarsus achieved adult size the most rapidly . This 
took an average of 13 . 5 days . 
Fledging Success 
In order to determine brood success an arbitrary point of success 
needed to be established due to the difficulty of knowing when the 
nestling period ended . 
Figure 1 .  Mean daily increase in the wing length of chicks in different 
sized broods: 
The means for broods of one are represented by hollow circles , broods of 
two by horizontal bars and broods of three by solid circles . Standard 
errors of the means (vertical lines) for chicks in broods of three 
rarely overlap those for chicks from single or double broods after the 
tenth day. Although n = 118 , not all chicks were measured each day , 
hence the unevenness of the growth curves . 
Figu r e  1 .  M ea n  d a i l y  i n c r e as e  i n  w i ng le ngth of chi cks i n  d iff ere nt  s ized  b r o o ds  
1 00 ? 
80  
--. 60  t:: 
t:: 
....__ 
40 
0"1 
c 
3 2 0  
1 0  
Adult  R - -
A du l t  m i n . - - -o r 
0 0 
+ 
' I - I 0 t . o O + 
0+ - t 
?t 1,_ 0 
+ L.t? T ? ? y? I 0 I I 
o rl+? ? ? ? r 
+<t?<* + ,r , , I 
r+if 9 + t
. . + 
e-Oeq,-. 
? ? 
2 0  
D a y s  
3 0  40  
Lo-o -- . i 
I)-
(J\ 
(J\ 
Figure 2 .  Mean daily increase in chick weight : 
The means for broods of one are represented by hollow circles , broods of 
two by horizontal bars and broods of three by solid circles . Standard 
errors of the means (vertical lines ) for chicks in broods of three 
rarely overlap those for chicks from single or double broods after the 
tenth day . Although n = 118 , not all chicks were measured each day , 
hence the unevenness of the growth curves . 
Fig u re 2 M ea n  d a i l y  i n c r e ase  i n  ch i c k  we i ght  
9 
9 l 0 
A d " l t  ' - - ? 
? 0 9 H? I ' b ? ? ' ? H 
0 
?:1, Tt? t-tH? t?ftJ:?-=- tj 0 +
o ? - _ _ 
9 1? + + + ' T t t + t + + + 1 ? t + I ??? 
?? + , f f . + 
T t 
o 
? . ?  
o-? ? 
? ? 
cP? + + 
1 
ooa. 
? 
1 0  2 0  
D a ys  
3 0  4 0  
0\ 
'-! 
Figure 3 .  Mean daily increase in tarsal length : 
The means for broods of one are represented by hollow circles , broods of 
two by horizontal bars and broods of three by solid circles . The 
consistent overlap between the standard errors (vertical lines ) of 
chicks from all three clutch sizes indicates the small effect food 
supply has on tarsal growth. 
c 
Q.J _. 
c 
Q.J 
V1 
to 
Q.J 
. c... 
u 
c 
c 
to 
Q.J 
::E: 
Q.J 
c... 
::I 
?-
LL. 0 CO '-.0 ...-
(W W} ? i D U al 
0 
m 
(/) 
0 >.... 
N ru 
0 
6 8  
6 9  
The difficulty was caused by the wandering habits of older chicks , which 
although possibly beneficial in exercising their wings as they "walked" 
around on the cave walls , made location of marked birds more difficult . 
Taking a wing length of 90 millimeters as indicating that a nestling is 
likely to fledge , the following determinations were made . Parents with 
a brood of one raised 0.43 + 0 . 11 chicks ( x  ? SE ,  n = 24) . Those with a 
brood of two raised 0 . 92 + 0 . 15 (n = 25 ) and those with a brood of three 
raised 1 . 09 ? 0 . 3  (n = 11) . 
These data show that a pair with a brood of two , rear significantly 
more young than those with a brood of one (t44 = 2 .
57 ,  P< 0 .01 ) , while 
those with a brood of three do not rear significantly more than those 
with broods of two ( t15 = 0. 59 ,  n . s . ) .  
Because average growth curves conceal certain characteristics of the 
individual growth curve and in particular the variation within a brood , 
the progress for a selection of individuals has been plotted in figures 
4 and 5 .  It will be noticed that even in the brood of three (and this 
was true for all other broods of three ) ,  death was not preceded by a 
drop in weight as was found in the Common Swift in wet summers (Lack & 
Lack 1951)  and sometimes in the smaller siblings of the Edible-nest 
Swiftlet (Langharn 1980)  ? 
Feeding Rate 
The four full-day observations on a sample of the manipulated nests 
demonstrated that the average of 2 . 9  + 0 . 34 visits (x ? SE ,  n = 11 )  to 
broods of three was significantly greater (Median test , x2 ::: 4 .  97 , .!:< 
0 .05)  than the 2 . 2  + 0 . 1 visits (n = 18) to broods of two . The 2 . 1  + 
0 . 01 visits (n  = 18) to broods of one was not significantly different 
(Median test , x2 = 0. 8 ,  n . s . ) from the number of trips to broods of two . 
While the increased number of feeding trips made to broods of three 
might be taken to indicate that the chicks in the enlarged broods would 
receive adequate food for normal growth , this is not the case . 
1 00  
80  
6 0  
E 
E 
40 
en 
c 
3: 
2 0  
1 0  2 0  
D a ys 
3 0  
/ 
./ 1  
40  
I 
I 
I 
I /  ./ 
j./  
Figure 4.  Wing growth in individuals from broods of all three sizes . 
Chicks from single broods are represented by continuous lines , those 
from broods of two , by dotted lines and those from broods of three by 
pecked lines . 
7 0  
1 1  
9 
/ 
'1 ?  / / 
/ 
/ 
0"1 
r:? .) 
+-
.s::: 
0"1 
QJ 
:3: 
3 
1 0  2 0  
D a ys 
...--
/ 
3 0  
- -- /  
/ - / / / 
/ 
/ 
- - -
/ 
/ 
40 
7 1  
Figure 5 . Weight increase in individuals from all three sized broods . 
Chicks from single broods are represented by continuous lines , those 
from broods of two , by dotted lines and those from broods of three by 
pecked lines . 
The feeding rate per chick ( 1 . 0  feed each) in broods of three was not 
significantly different ( t27 = 1 . 1 )  from the feeding rate ( 1 . 1  feeds 
each) per chick in broods of two . 
72 
During the fUll-day ' s observation on manipulated nests in 1981 a 
simultaneous collection of data was made on 20 undisturbed nests . These 
were not part of the manipulation experiment and were so high as to be 
out of reach. An average of 2 . 8  ? 0 . 26 feeding visits was made to these 
nests . Thi.s wa.s si?l'li?c.::?,I\Hy .ti?re"i ? 6 ::?- ?? P< O?O?}j -t-o Ut sirs -{-D Tt..e I6Wef" fle..sTs. 
DISCUSSION 
Despite the apparently favourable weather for gathering food , the 
chicks from artificially enlarged broods of three experienced 
significant delays in wing growth and weight increase . These results 
are inconclusive because survival studies following brood manipulation 
on the Puffin ( Fratercula artica) (Harris 1982 ) and the Blue Tit (Parus 
caeruleus) (Nur 1984a,b)  show that delayed fledging does not necessarily 
reduce the future survival of either chicks or parents , and this should 
be particularly true in the Apodidae where chicks are adapted to 
withstand long periods without food . Although post-fledging survival is 
higher for individuals from smaller broods than those from larger broods 
in species such as the Starling ( Sturnus vulgaris )  (Lack , Gibb & Owen 
1957 ) and the Great Tit ( Parus major) ( Perrins 1965 ) , this does not mean 
the same phenomenon applies to the Apodidae . 
The only study on clutch-size and post-fledging survival in an 
apodid is that of Lack & Arn ( 1947 ) , which showed post-fledging 
mortality in the Alpine Swift to be independent of brood size . In both 
the Common Swift (Weitnauer & Lack 1955 , Perrins 1964) and the Alpine 
Swift (Lack & Am 1947 ) , nestling survival decreases with an increase in 
natural brood size . 
7 3  
It may be that the effect of insufficient food for breeding apodids is 
seen in the death of the nestlings rather than the newly fledged young . 
In the final analysis the extended nestling period that all apodids 
experience and which may be considered as a first line of defence 
against starvation in nestlings , appears insufficient to ensure that all 
chicks fledge when there is a food shortage . That assimilis chicks from 
artificially enlarged broods spent longer in the nest ( than those from 
normal sized broods) in order to reach adult weight before departing the 
nest , may indicate why the underfed apodid nestling is more likely to 
die before fledging than soon after . By contrast , Starlings (Lack , Gibb 
& Owen 1957 ) and Great Tits (Perrins 1965 , Moss 1972 , Garnett 1981 ) 
fledged at the normal time but at lighter than normal weights , when 
there was a shortage of food at the time of breeding . The adaptive 
pressure for apodids to remain in the nest is that once they leave it 
they have to gather food for themselves . This is not the case in the 
Starling ( Feare 1984) , Great Tit (Gibb 1960 ) , or even birds that are 
ecologically similar to swiftlets , such as swallows (Gooders 1975 , Frith 
1976 ) , which after leaving the nest are still fed to some extent by 
their parents . 
Because the 20 undisturbed high nests undoubtedly contained both one 
and two-chick broods their visitation rate can be compared to the 
combined average of 2 . 17 ? 0 . 06 visits to manipulated broods of one and 
two chicks . Because the difference is significant ( t34 = 2 . 36 , P< 
0 .05 ) , the possibility arises that my presence was reducing the number 
of feeding visits made by parents whose chicks were being regularly 
handled . Such an effect would be equal on clutches of all sizes and so 
not affect the comparison between different sized clutches . However 
birds nesting at greater heights might be older , have closer feeding 
ranges and/or have some other benefit that will increase their capacity 
to provide for their brood. 
7 4  
CONCLUSIONS 
Because parent assirnilis could not hatch significantly more chicks 
when given a third egg , or fledge significantly more chicks when given a 
third chick , it is clear that the parents would not gain any advantage 
by producing a clutch of three themselves .  
Because broods of three were fed significantly more than broods of 
two it appears that assirnilis is not maximising its harvest of the 
available food supply during the breeding season. Even though the 
feeding rates per chick in broods of two and three were not 
significantly different , parents could not hatch significantly more 
chicks when given a third egg , or fledge significantly more chicks when 
given a third chick. It is clear that parents are maximising the number 
of fledglings they can raise . The inability to rear three chicks may 
mean that food boluses are smaller when the feeding rate increases thus 
preventing significantly more chicks fledging from the larger broods . 
Although this swiftlet is not maximising its harvest of the 
available food supply , this does not mean that they are reducing annual 
production in order to optimise their life-long production. However , 
the effort saved by not working at its maximum food-gathering capacity 
may help explain why assimilis is a long lived species for its size 
(Tarburton 1987b) . 
It is  concluded that whereas a shortage of food in the breeding 
season means some passerines will lose more recently fledged chicks than 
normal , the White-rumped Swiftlet (and probably other apodids in the 
same circumstance ) will lose more nestlings than normal . 
7 5  
Section 3 
INTRA-TROPICAL VARIATION IN CLUTCH SIZE 
INTRODUCTION 
It has long been held that bird species nesting in the tropical 
savannah will have larger clutches than the same or closely related 
species nesting in tropical rainforest (Moreau 1944 ; Ashmole 1961 ; Lack 
& Moreau 1965 ; Skutch 1967 , 1976 ; Lack 1968 ; Ricklefs 1970 , 1980 , ) .  
However , it has taken some time to find an acceptable explanation for 
this phenomenon. While most are agreed that a large clutch is  an 
adaptive compensation for a shortened average life expectancy , there is 
a variety of views as to what controls the increase in clutch size . In 
Lack ' s  theory (particularly as modified by Ashmole ,  1961 , 1963) of 
maximum reproduction by maximum use of food available to breeding birds , 
it is suggested that high mortality during the non-breeding season in 
the savannah, would mean more food per pair for those that survive until 
the next breeding season . Having more food than those in a stable 
environment consequently allows for larger clutches . Alternatively,  
Skutch ( 1949 , 1967 ) reasoned that the reproduction rate is  determined by 
the average annual mortality rather than the other way round and that 
therefore , in places such as the savannah ,  the high mortality of the 
non-breeding season would increase the clutch size . Cody ( 1966 ) 
maintained that birds nesting in stable environments such as tropical 
rainforests would have little need for large clutches as part of high 
reproductive rates and would profit from placing more energy into 
K-selected activities that would prolong their own life . These would be 
activities such as predator avoidance and competition . Alternatively , 
birds living in savannah would do best to concentrate on r-selected 
activities such as increasing clutch size to help replace the large 
numbers lost each winter . 
Ricklefs ( 1980) has also lent support to the general prediction that 
savannah birds will have larger clutches than close relatives in the 
rainforest . 
He has shown that clutch size is determined by the ratio of the food 
supply to its demand by the population of breeding adults . As the 
7 6  
supply depends largely upon the habitat during the breeding season and 
the demand is  regulated by density-dependent factors acting upon the 
population during the non-breeding season , the population density is 
predicted to be regulated primarily by winter mortality and not by 
territorial behaviour during the breeding season as Wynne-Edwards 
( 1962 , 1963) had maintained . 
It has been suggested (Winkler & Walters 1983) that studies of 
exceptions to this widely supported trend should be especially 
instructive . Because A.  s .  chillagoensis from the savannah environment 
of Chillagoe produces a clutch of one , and A .  s .  assimilis from the 
rainforest environment of Fiji produces a clutch of two , it would appear 
that this species is  an exception to the rule and therefore worthy of 
study. 
In comparing the White-rumped Swiftlets of Fiji and Chillagoe 
(Queensland) , several factors that have complicated other studies on 
clutch size are avoided. The variation in day length with a change in 
latitude and hence the unequal time to gather food for breeding purposes 
is avoided. Problems arising from the observation that some savannah 
birds are seed eaters , while their rainforest relatives are insect 
eaters (Lack & Moreau 1965 ) , are also avoided . In fac? very few studies 
comparing clutch size in savannah and rainforest have used the same 
species . 
r 
To these advantages can be added the occ?ence of a good season 
during the 1985/6 breeding period and a poor season during the 1986/7 
breeding season at Chillagoe . 
7 7  
The rain that fell during December and January of the good season 
represented 152 per cent of the average rainfall and this correlated 
with a much higher density of available insects than was found in the 
poor season , when for the most part only 35 per cent of the average 
rainfall was recorded . :Having a good and a poor season has allowed for 
a clearer assessment of the birds ' ability to feed an extra chick under 
both abundant and scarce food supply situations . 
METHODS 
A sample of nests in Gordale Scar Pot (CH 187 ) and Guano Pot (CH 
146) was used for controls and for manipulation experiments . Methods 
used are similar to those used in Fiji ( section 1 ) , except that because 
the only natural clutch is one , it was necessary to manipulate only 
clutches and broods of two . Both control and manipulated nests were 
visited six days a week between 28 November 1985 - 27 January 1986 and 2 
December 1986 - 23 January 1987 . Two additional experiments were run 
simultaneously. The first involved feeding meat based-dog food twice a 
day to seven manipulated twins . The amount given was as much as the 
chicks would take during two to three minutes . Once it 'Was realized 
that the average growth rate of chicks on the supplemental food was 
still falling behind that of those from single-chick broods and these 
chicks were still falling from the nest , another strategy was tried . 
Insects such as winged termites collected from their mounds , and Emperor 
Gum Moths collected from the windows of lit rooms at night , were given 
to the chicks instead of the dog food. That on three occasions chicks 
neeJ for o. 
were seen regurgitating the dog food also validated theAchange in diet . 
The second additional experiment involved enlarging eight nests by 
gluing a length of 6 mm manila rope along the rim of the nest and to the 
wall with "supa glue" . This made nests as deep , as long and as wide as 
Fijian nests which normally accommodate two chicks . 
7 8  
RESULTS 
Hatching Success 
The hatching success of 58 single-egg clutches ( 0 . 69 + 0 .06 ) in the 
good season of 1985/86 at Chillagoe was not significantly different 
(0? 60:!  () ?0(;) 
( t125 = 1 .05 , n .s . ) from the hatching success in the poor seaso?of 
1986/87 and so the results may be pooled . This average (0 . 64 + 0 . 06 )  is 
not significantly different ( t154 = 0 . 15 ,  n . s . ) from the hatching rate 
of single-egg clutches from Fiji (0 . 52 ? 0 . 09 ,  n = 29 ) . The hatching 
success of 10 two-egg clutches in the good season at Chillagoe was 1 . 8  + 
0.09 .  It is obvious that this is significantly better than the hatching 
rate for single egg clutches and is also significantly greater ( t58 = 
3 . 9 , P<0 .001)  than the hatching rate for two-egg clutches in Fiji ( 1 . 15 
? 0 . 14 ,  n = 40) . However , in the poor season at Chillagoe the hatch 
rate of twin eggs was only 0 .  7 ? 0 .3  (n = 6 ) , which is significantly 
less ( t26 = 3 . 5 ,  P< 0 .02 )  than the hatching rate in the good season and 
not significantly different ( t46 = 1 . 36 ,  P> 0 . 1 )  from the Fijian 
hatching rate . 
Chick Growth 
Figures 1 and 2 show the mean daily increase in length of wing and 
weight for individuals in broods of one and two during the favourable 
season of 1985/6 . The standard errors on the wing growth curves 
indicate a significant difference between these broods after the eighth 
day. 
r 
This divergence occ?ed earlier than in assimilis where it  was the 
tenth day before it was evident that the broods with the extra ( third) 
chick were dropping significantly behind those in normal sized broods . 
Average adult wing length was not reached before birds fledged though 
the minimum adult wing length was reached by most fledglings . 
Figure 1 .  Mean daily increase in wing length of chicks at Chillagoe 1985/86 . 
The means for broods of one are represented by solid circles , broods of two by 
hollow circles and those of surviving chicks by addition signs . Standard 
errors of the means rarely extended beyond the symbols used and so are not 
shown. 
1 0 0 ? 
8 0  I E E 
:::r:: 
6 0 1 f-l:l z UJ _J 
l:l 
40 ? z ....... :3 
2 0 1-
0 
F ig u r e  1 M e a n  d a i ly i n c r e a s e  i n  w i ng l e n g t h  of c h i c k s  at C h i l l ag o e  1 98 5/ 86  
x a d u l t w i n g l e n g t h 
m i n. a d u l t  w i ng l e n g t h 
? ? 
? 
? ? 
+ + + ? 
? + 
? ? + + ? + 
? ? + 
? + + ? + 
? 0 ? 0 
? ? + 0 
0 
? + ? + ? 0 
? 0 
? 0 0 
? 0 
? 
? 0 0 
? 0 
? ? 
0 0 
? 0 0 
? 0o o 
eo'o Oo eo oo 'o ? 
1 0  2 0  3 0  
DA Y S  
? 
-=-? . . .
. ? + + i ? 
+ + ? ? + ? + 
+ + 
+ 
4 0  5 0  
-...J 
1.0 
Figure 2 .  Mean daily increase in weight of chicks at Chillagoe 1985/86 . 
The means for broods of one are represented by solid circles , broods of two by 
hollow circles and survivors from broods of two by addition signs . Standard 
errors of the means (vertical lines ) for chicks in broods of two do not overlap 
those for chicks from single broods after the fifth day . 
en 
1-
:c 
0 
w 
3 
1 0  
8 
6 
4 
2 
0 
Fi g u r e  2 M e an dai ly i ncr ease i n  w e ight of c h i c ks at C h i l l a g o e  1 985/86 
eo 
eo 
eo 
+ +  + + + + + + t +  f ? d t t. d  + +  + q  + ? ? 
j T , t , f i ? 
' ?  
X a d u l t  We ig h t ----<>- t 1 f t + r + + + + + + + 
m i n. a d u lt we i ght - ? f , !' f!' j j j ! } l 
.L + 6 T + + ? I 
+ + + t ? 
+ 
+ t ? t 
? 9 
? ? ? 9 + 
? 0 
+ 
? 0 
+ 
+ ? 0 
1 0  
+ 
2 0  3 0  4 0  
DAY S 
5 0  
(X) 
0 
8 1  
In the poor season of 1986/7 the significant difference between the 
average wing length of the two broods did not appear until the tenth day 
(Figure 3 ) . Figure 3 also shows that the wings of chicks from single 
broods grew significantly more slowly in the poor year than in the good 
year. Not only did the chicks from manipulated two-chick broods grow 
significantly more slowly than those from single-chick broods but in all 
IIVI"'5S 
cases one died and the survivorl-1\took an average of 10 days longer to 
reach minimum adult length. 
The weight of chicks in broods of two was also significantly lower 
than the weight of those from single broods . On the fourth day a 
significant difference could be detected between the weight of those 
chicks in single-chick and two-chick broods during the good season. The 
performance in the poor season (Figure 4) was even worse for it was the 
sixth day before the weight of the single chicks increased significantly 
above that of the two-chick broods . Comparing these measurements with 
those for assimilis chicks from one and two-chick broods , which were not 
significantly different until the 12th day , it is clear that 
chillagoensis is rruch less able to cope with an extra chick. Chicks 
from single broods reached adult weight by the 17th day whereas 
surviving chicks from two-chick broods did not reach adult weight until 
the 23rd day. These times are respectively only two and one day earlier 
than those from the same sized broods in Fiji . In the good season 
chicks from single broods did not attain average adult weight until the 
36th day . In the poor season it was the 42nd day before the average 
chick reached average adult weight . 
The long-lasting effect of insufficient food for chicks in two-chick 
broods is shown by the longer time taken by them to reach maxirrum weight 
when compared with single-brood chicks . During the good season 
single-brood chicks took 23 days to reach an asymptote whereas the 
survivors of the two-chick broods took 35 days to reach the same level . 
Figure 3 .  Mean daily increase in wing length of chicks at Chillagoe 1986/87 . 
The means for broods of one are represented by solid circles , broods of two by 
hollow circles and survivors from broods of two by addition sign . As this was 
the poor season and the previous season was a good one , the mean daily growth 
curve (using solid squares )  for the preceding season has also been plotted . 
8 2  
+ 
+ 
+ 
+ 
+ 
"+ 
? .+ 
? + 
?? + 
? ?  + 0 lJ") 
... + 
... + 
? ? + 
[-... ? ? + CO 
? ? + -Cl 
CO ? ? + "' ..-- ? ? + 
OJ ? ? + 
0 ? 0 + en 
ro ? 0 + 0 ...:t 
...c:: ? ? + 
u ? ? + 
4- ? ? + 
ro 11 ? + 0  
V) Ill ? + 0 ..X: u ? ? + 0 
...c:: 1111 ? + 0 u 
..__ ? ? + 0 
0 ? ? 0 
...c:: 11 ? 0 0 4- m en ? ? 0 c 
OJ ? ? 0 (/) 
1111 0 0 >-
en <( 
c 1111 0 0 0 
3: 1111 0 0 
c 11 0 0 
1111 ? 0 
OJ Ill 0 0 V) 
ro ? ? 0 OJ 
'- ? ? 0 0 u N c ? ? 0 
>- 11 ? 0 
ro ? .0 
"0 ? ? 0 
c ? 0 0  ro 
OJ ? 00 
::E a tP  
? et> m -? 
OJ ? 0 <-::J 
en ? 
lL ? 
? 
0 
0 0 0 0 0 0 CO -Cl ...:t N ..- ( w w ) H l9 N  ::Jl 9 N l M  
Figure 4 .  Mean daily increase in chick weight at Chillagoe 1986/87 . 
The means for broods of one are represented by solid circles , broods of two by 
hollow circles and survivors from broods of two by addition signs . The solid 
squares represent the growth curve for chicks from the same colony in the 
previous season which was a good season in terms of growth and survival rates . 
F i g u r e  4 M e a n d a i l y  i n c r e as e  i n  c h i c k  w e i g h t  a t  C h i l l a g o e 1 98 6 / 87 
1 0 1-
8 ? 
11 
11 
11 
? 6 1- ? 
Ill ? ? 
? ? J  @ 0  ? ? + ? ? 0 ? ? 11 ? 0 + + 
? ?  0 0 
? ? <;!! 
+ 
? 0 + ? 0 
2 ' 0 
? 
? 
. . . ? ? ? ? ? ?
? ? ? ? ? ? * ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? + + + ? ? ? '1- + + ? ? ? ? ? + 
11 ? ? ? ' + i + 
? + + + + + 
? ? + + ? 0 
0 0 0 
? 0 + 0 
0 + 0 
? 0 + + 0 9 
+ 
11 ? 11 
?? ?? ? ?  . + + 
? ? ' I + ? t. ? '+_ + + + + + 
+ + ? + 
o?----?------?----?------?----?----??----?----??----?------?-1 0  2 0  3 0  40 5 0  
DAY S 
(X) 
w 
84  
Wing growth in these two-chick brood survivors was also delayed , taking 
47 days to reach 100 rrrn, whereas single-brood chicks took only 43 days 
to reach the same wing length . 
By graphing the average daily weight changes in all chicks per 
calendar day it became apparent that at periodic intervals the average 
weight gain was noticeably higher than usual . These days are shown 
(Figure 5 )  to correspond with the first days on which rain fell . 
Flectging Success 
In the good season, chillagoensis parents with a brood of one raised 
0 . 69 + 0 .09 chicks , which is not significantly different ( t38 = 0 .65 ,  
n . s . ) from the 0 . 56 .:!:. 0 . 18 chicks raised from broods of two . The 
fledging success with single broods was also not significantly different 
( t54 = 1 . 83 ,  n . s . )  from the 0 .43 .:!:. 0 . 11 chicks raised from single broods 
in Fiji . However , in not one case in either the poor or good year at 
Chillagoe were both chicks from a two-chick brood successfully fledged 
(n = 34) . In the poor season chillagoensis with broods of one raised 
0 . 31 .:!:. 0 . 12 chicks , which is significantly less ( t45 = ? . 5 , P< 0 .02 )  
than those raised from single-chick broods in the good season. In the 
poor season chillagoensis with two-chick broods raised 0 .71 .:!:. 0 . 11 
chicks . This is significantly more ( t34 = 2 .  46 , P< 0 .  02 ) than those 
raised from single chick broods in the same season . 
Only one chick fledged from each of the two-chick broods that were 
given supplementary food. Similarly only one chick fledged from each of 
the two-chick broods that were provided with enlarged nests . 
Because average growth curves conceal certain characteristics of the 
individual growth curve and in particular the daily variation within a 
brood, the daily increase in weight of a selection of individuals has 
been plotted in Figure 6 .  
Figure 5 .  Average daily change in relative chick weights ( 1986/87 ) . 
The plotted indices are daily means derived from the log ofthe wei,shlot="a. chic.k ofl one day 
divided by the chick' s  weight on the previous day . This method ffi9.kes the index 
sensitive to the amount of food collected but less sensitive to the age of the 
chicks . The distinction is important as larger chicks are normally fed more 
food than sms.ller chicks and because there are more large chicks later in the 
season. Together these factors would distort the index if absolute weight 
changes were used rather than relative weight changes . The index clearly shows 
the above average weight increases that follow rain days ( shown by arrows above 
the date line ) .  
-0'1 
0 . 3 1- F I G U RE 5 Averag e  Da i ly C h a nge  i n  R e l a t i v e  C h i c k  We igh ts  ( 3 De c .  1 986 - 23  Jan .  1 9 87 ) 
+ 
- 0. 2 
VI 
.... 
...!. f t 
I T ? 0.1 + + + L . .  t 
3 Dec  10  
+ + 
20  3 0  1 Jan. 1 0  2 0  
T ime ( days ) 
t 
CX> 
\Jl 
Figure 6a. Average weight increase 1986/1987 . 
This graph shows average weight increase for single-brood chicks of equal age 
during the good season of 1986/87 . It is placed here for comparison with the 
growth curves of the individual chicks shown below.  
Figure 6b. Daily weight change in two single-brood chicks . 
Steps in the growth curve in response to rain ( shown by arrows above the date 
line ) are more marked in chicks when they are large . 
Figure 6c . Daily weight change in a pair of manipulated chicks . 
Rises in the' growth curve in response to rain are not so evident as they_ are in 
chicks from single-chick broods . Growth responses to rain in the surviving 
chick are more pronounced. 
F I G U R E 6 a  Average  We ight I n c r e a s e  ( 1 986 / 87 ) 
-9 
en 
4-?
..c 
7 
?? QJ 
3 
3 
F I G UR E  
1 1  
9 
3 
? 
? ? 
? ? 
? 
? 
? 
A A 
1 D?c 
FIGURE 6c 
9 
? 
? ?  
? 
? 
A g e  ( d ays ) 
e i n  Two S i n  le Brood Ch i cks 
? 0 0 e ? 0  o o 
0 0 
? 
? ? 
? ? 0 ? 0 ? ? ? o o o ? 
? 
? ? 0 0 o e ? ? ? ? ? 
? ? ?  ? 
? ? ? 
0 0
0 
v 
0 
0 0 0  
0 o o 
Oat!'! 
e in a Pa i r  of Man i  ulated Ch i c ks 
0 0 00 0 o o 0 0 0 0 ? o0 0 0 0 0 o 0 ? O o o o o o o o o 
? ? ? ? ? ? ? ? ? 
?
? 
? ? ? 0 0 
O o 
D a te 
? 
? ? ? ? ? -
? 
? 
0 0 
0 
0 0 o o  
o o 0 
0 
8 6  
0 0 
8 7  
The individual growth curves of chicks from single-chick broods (Fig.  
6b) show greater deviation in response to rain than the deviations for 
chicks from two-chick broods ( Fig , 6c ) when compared to the average 
growth curve ( Fig. 6a) . However , the decline in the weight of starving 
chicks is very clear in the individual growth curves for chicks from 
two-chick broods (Fig . 6c ) . Some chicks that died did not show weight 
declines because they fell from their nests . 
Feeding Rate 
The full day ' s  observation on a sample of the manipulated nests of 
chillagoensis in the good season demonstrated that the average of 4 .  7 + 
Per do.r. 
0 . 67 feeding visits (x ? SE ,  n = 3)A to broods of two was not 
significantly greater ( t21 = 0 .43 ,  n . s . ) than the 5 . 2 ? 0 .3  visits (n = 
20) to broods of one . Another full day ' s watch on chillagoensis nests 
in the poor season similarly demonstrated that the average of 2 . 7 + 0 . 3  
visits ( n  = 3 )  to broods of two was not significantly greater ( t9 
0 . 39 ,  n . s . ) than the 3 . 0  ? 0 . 7  visits to broods of one . However , the 
number of visits to broods of one in the poor season was significantly 
( t26 = 2 . 9 ,  P< 0 .01 )  less than during the good season. Similarly , the 
number of visits to two-chick broods in the poor season was 
significantly ( t6 = 2 .  7 ,  P< 0 . 05 )  less than in the good season. The 
number of visits to single broods of chillagoensis in the good season 
was significantly greater (t38 = 6 . 3 ,  P<O . OOl ) than that of feeding 
visits to single broods of assimilis ( 2 . 8  ? 0 . 3 ,  n = 20) . In the poor 
season, however , the number of visits to single broods of chillagoensis 
was not significantly greater ( t26 = 0 . 26 ,  n . s . ) than the number of 
feeding visits to single broods of assimilis . 
88  
Available food supply 
The average number of insects ( 95 ? 29)  caught in the sweep net 
samples of available prey in Fiji was significantly more ( t32 : 3 . 0 ,  f< 
0 .  01) than the average number caught in the sweep net during the good 
year at Chillagoe ( 9 .  7 + 1 .0) . The average number of insects ( 5 . 0  ? 
1 . 1 )  caught in the sweep net at Chillagoe in the poor year was 
significantly less ( t40 : 2 . 1 ,  P< 0 .05 )  than that caught there in the 
good year. 
It was significantly (Median test , x2 ? : 6 . 55 ,  P< 0 .05 )  more likely 
than not , to catch more than the average number of insects caught in the 
sweep net when either rain fell or the irrigation sprinklers on the 
block adjacent to the main sample site had been running in the previous 
24 hours . It was significantly (Median test , x2 ? = 6 . 55 ,  f< 0 . 05 )  more 
likely that swiftlets would be feeding in the vicinity of the sweep net 
site on those occasions when the net gathered more than the average 
number of insects . There was no significant correlation (X? : 2 . 3 ,  
n . s . ) between whether swiftlets were feeding or not in the area adjacent 
to the sweep net sampling site , and whether either rain 'had fallen or 
the irrigation sprinklers had been operating in the previous 24 hours . 
DISCUSSION 
Despite the apparently favourable weather for gathering food in 
three of the four br?eding seasons followed in this study , neither 
chdllagoensis nor assimilis were able to raise significantly more chicks 
from artificially enlarged broods than from normal sized broods . These 
results mean that chillagoensis is not responding to the harsh extremes 
of the savannah climate in the way that a number of theories predict . 
Rather than producing a larger clutch than assimilis , 
chillagoensis produces a smaller clutch. 
8 9  
Even the higher fledging rate of two-chick broods in the poor season at 
Chillagoe can be explained by the supplementary feeding given to most 
twins , but not to chicks in single-chick broods . This situation 
therefore needs to be evaluated from several theoretical standpoints . 
Clutch size and ' competitive release ' on islands 
It is commonly stated that island species of birds have smaller 
clutches than their closest mainland relatives . Lack ( 1954) gives 
evidence of' this for the limicoline birds from the Falkland islands 
compared with South America and for ducks (Lack 1968 ) on a number of 
remote islands . Cody ( 1966 ) cites evidence for smaller clutches in 
passerines on small oceanic islands off' the coast of' New Zealand . These 
examples are all from temperate regions , and when Klomp ( 1970) includes 
the Caribbean examples given by Cody ( 1966 ) as further examples of' 
reduced clutches on islands , he has missed the point Cody was making. 
Cody ( 1966 ) was predicting , from his model relating clutch size to the 
birds ' allocation of' time and energy, that although temperate islands 
should have reduced clutch sizes , tropical island clutch sizes , if 
different at all , will be only slight and not necessarily smaller .  Cody 
( 1966 , 1971 ) believes birds on temperate islands will have smaller 
clutches because they are likely to have fewer predators , a more equable 
climate and larger ecological niches than on the mainland. His reason 
for predicting little difference between island and mainland clutches in 
the tropics is that on tropical islands there is little difference in 
climatic stability and the main deciding factor will be the level of 
predation on the island . 
On the other hand, Murphy ( 1968 ) asserts that predation has nothing 
to do with clutch size in the tropics and that smaller clutches have 
evolved on tropical islands in response to the uncertainty of' survival 
from zygote to maturity resulting from populations at or near 
saturation. 
9 0  
The 1()()';b increase in clutch size that assimilis has over 
c..h i lla.;s<9ens/.s is not insignificant and therefore is not supportive of 
Cody' s prediction. This is not the only example in the Apodidae where 
island subspecies have larger clutches than their mainland counterparts .  
The African Palm Swift has a clutch of two throughout its range on the 
African continent but a clutch of three on Madagascar (Moreau 1941 , 
Brooke 1971a) . That this palm swift also has a more restricted breeding 
season on Madagascar (Rand 1936 ) than it does on the mainland , suggests 
that on islands there are more factors controlling breeding strategies 
such as clutch size , than have so far been accounted for . Other 
evidence that there are more factors is found in the variation in clutch 
size within a swiftlet species that is found only on the oceanic islands 
of Micronesia. The Caroline Swiftlet (Aerodramus inquieta) lays one egg 
on Kusaie and Pona.pe , yet two eggs on Truk Island (Brandt 1966 ) . All 
three islands are in similar latitudes , have similar altitudes and area , 
and are similar distances from the Asian mainland . 
Two of these other factors thought to contribute to the regulation 
of clutch size are nest size per se and predation. Both deserve a 
closer look as data for six subspecies of the White-rumped Swiftlet and 
three subspecies of the Caroline Swiftlet allow for an examination of 
these as possible causes in regulating clutch size . 
The Theory that Nest Size Influences Clutch Size 
Quite separate from the effect that food abundance may have on 
clutch size is the constraint of the nest itself. Snow ( 1978) has made 
the point that the structure and the size of the nest has never been 
properly considered as a factor that limits the maximum clutch-size of a 
species . He begins by taking the extreme example of the nest of the 
Tree Swifts (Hemiprocnidae ) ,  which can hold only one egg. He suggests 
that these nests point to a problematic situation rather common in 
tropical birds : the need for an inconspicuous and inaccessible nest . 
9 1  
However , it is difficult to see how the benefits of an inconspicuous 
nest , made so by its small size , could operate in the case of the 
Whi te-rumped Swiftlet , where the darkness of the cavernicolous nest 
sites it  chooses make the nests invisible to predators and competitors 
alike . Inaccessibility is achieved by the placement of the nests on the 
rock walls and ceilings of the caves , mostly in complete darkness . Such 
sites offer no restriction as there are considerable areas suitable for 
nesting in each of the six caves examined in Fiji and at 21 of the 22 
caves examined at Chillagoe . These vacant nest sites are not confined 
to the entrance areas where predation (by Barn OWls (Tyto alba) in Fi ji , 
and raptors and cats in Queensland) occasionally occurs , but between 
areas used for nesting . 
That some of the swiftlet nests and their chicks in Guano Pot , 
Chillagoe , were washed off the wall by seepage from heavy rains in 
January 1986 suggests another reason why nests may not be built  in what 
appear to be sui table areas of a cave . This also suggests that nest 
size may be affected by water . Nests becoming wet from water flowing 
over rock surfaces is apparently more common in the smaller coastal 
Queensland colonies , where Smyth et al ( 1980) suggest that this wetting 
has contributed to the failure of such vulnerable sites in wet years . 
While it is theoretically possible that building smaller nests would 
reduce the chance of the nest being on a wet portion of a cave wall , 
this suggestion has not been researched . The suggestion may benefit 
from further study as none of the five caves I visited in Fi ji , where 
the birds raise two chicks , were seen to lose nests due to wetting and 
so such a pressure to reduce nest size may not exist there . That a 
number of Apodidae frequently nest behind waterfalls with at least one 
species regularly building in the spray of waterfalls without detriment 
to their breeding effort ( Stresemann 1928 , Somadikarta 1968 & Becking 
1971 )  does not necessarily mean that this species can do likewise . 
9 2  
However , the second season at Chillagoe was very dry and a number of 
nests fell apart as a consequence , so both extremes of wet and dry can 
be detrimental to nest durability.  
The common need for the total brood to be effectively warmed by one 
parent until their feathers grow may place an upper limit on brood size . 
However , this is not a serious restriction to the Apodidae as the young 
are resistant to long periods of cooling and as in all swiftlet species 
the White-rumped Swiftlet is confined to the tropics where cooling is 
less likely to cause death than in temperate regions . 
Even in Blue Tits and House Sparrows the saving in metabolic energy 
made by individuals in larger broods than average when the environment 
was at 15?C was not made at 20?C ( 0 '  Connor 1975 ) ? As assimilis and 
chillagoensis  chicks experience temperatures between 23?C - 26?C it 
appears that larger broods will not benefit significantly from losing 
less heat than smaller broods . 
It may be suggested that the weight of an extra chick causes a 
significant increase in the number of nests that fall from the roof , but 
none of the nests in Fiji or Chillagoe that were given an extra chick 
fell . 
Notwithstanding that there is no apparent environmental selection 
pressure that would favour swiftlets with smaller nests , there is 
evidence that such pressures exist for other birds . For example the 
hole-nesting tits lay larger clutches and raise larger broods when given 
larger nest cavities or larger nest boxes . Ludescher ( 1973) has shown 
this in the Marsh Tit (Parus palustris )  and Willow Tit (P .  montanus) ,  
while Lohrl ( 1973) demonstrated it in the Great Tit (P .  major) . 
In a situation more applicable to the Swiftlets , the Barn Swallow 
(Hirundo rustica) has been shown to lay significantly larger clutches in 
larger nests (Moller 1982 ) . So we may well ask does the size of the 
nest vary in the White-rumped Swiftlet? 
9 3  
Thirty terraereginae nests from the Tully Falls Cave in coastal 
Queensland ( Pecotich 1974) averaged 56 x 45 . 5  mm in length and width and 
3 mm in depth , giving a volume index of 7 .  6 cm3 ? One ht.mdred 
chillagoensis nests from Gordale Scar Pot and Guano Pot had an average 
size (x ? SE) of 49 . 9  ? 0 . 49 x 42 .7  ? 0 .47 x 11 . 9  ? 0 .42 mm and a volume 
index of 25 . 4  cm3 ? Larger than both of these were thirty-six nests from 
Fiji which averaged 50. 0  ? 0 . 7  x 49 . 7  ? 0 . 7  x 21 . 1  ? 1 . 0  with an average 
volume index of 52 . 4  cm3 ? Thus the nests of assimilis attain a much 
larger volume through being much deeper than the nest of either 
terraereginae or chillagoensis .  From this difference Moller ( 1982 ) 
would correctly predict that the Fijian birds would be the ones to 
produce the larger clutch. Why then should the Fijian birds produce the 
larger nests , which Moller suggests leads the laying female to respond 
by producing a larger clutch? 
The answer may depend on the length and therefore the darkness of 
the caves . The three ' caves ' with swiftlets at Tully Falls are very 
short (the longest is 21m) , forcing the birds to nest close to the 
entrance (Pecotich 1974) . It is therefore possible that predatory 
pressures would exceed those of the Fijian situation where the nests 
closest to a cave entrance are at least 30m from it . It is possible 
that greater predatory pressures on terraereginae have led to a 
reduction in nest size and therefore a reduction in clutch size . 
However , if predation (aided by twilight) was such a selective force how 
could it explain a clutch of one in chillagoensis where all but one of 
the 27 colonies I visited were in total darkness? 
Apparent support for the predator effect on nest size comes from 
earlier measurements of chillagoensis nests which were more than 40m 
from the entrance and more than 5m from the floor . These nests averaged 
55 x 45 x 35mm ( Pecotich 1974) giving a volume index of 86 . 6  cm3 , which 
is even greater than that of Fijian nests . 
9 4  
In apparent contradiction to these measurements Smyth et al ( 1980) 
summarize the range of Queensland nests as 40 - 60mm left to right , 30 -
7Cmn front to back and 3 - 30mm deep . This clearly ignores the 35 mm 
deep nests at Chillagoe . The same authors however report considerable 
variation in nest dimensions between colonies in Queensland . In any 
case nest size in the White-rumped Swiftlet does not consistently 
correspond with clutch size . 
The Theory that Relates Clutch Size to Predation 
Following an experimental increase in brood size that showed 
predation to be a greater cause for death than starvation ,  Lill  ( 1975 ) 
has suggested that nest predation ( selecting for small and inconspicuous 
nests ) rather than the ability of the female to feed the young , has been 
the main factor determining the natural clutch size of the White-bearded 
Manakin (Manacus manacus) . Could it be that predation has directly 
reduced the clutch size of the two Queensland subspecies of the 
White-rumped Swiftlet? Predation may be a stronger force in Queensland , 
for the coastal ' caves ' are shorter and therefore the nest sites are 
better lit , allowing predators to see the colony . There are more 
species of predators in Australia and so there is a greater chance that 
at least one is able to utilize swiftlet nest sites . 
As the largest Queensland colonies only have 500 nests ( Smyth et al 
1980) , compared to the average of 2 , 785 nests for my sample of five 
Fijian caves , it might be that predatory pressure is responsible for the 
small colony size in Queensland . However , there are many more species 
of prey in Queensland than in Fiji and the whole concept may not be 
transferable from manakins to swiftlets as the former are not colonial 
whereas swiftlets are. Colonial nesting i s  usually thought of as 
reducing the effects of predation. 
9 5  
My discovery of six new colonies at Chillagoe , including one ( Tarby ' s 
Swiftlet Pot , CH 379 ) only 34 m from the previously largest Chillagoe 
colony ( Gordale Scar Pot , CH 187 ) brings the number of known Chillagoe 
colonies to 27 . This number of colonies ( and there are surely more ) is 
greater than would be found in a similar area of Fiji and so what is 
lost in colony size by chillagoensis is at least partly made up for in 
the greater number of colonies . So then the smaller colony size at 
Chillagoe does not necessarily indicate higher predation than in Fiji . 
However ,  higher predatory pressure might exist in coastal 
Queensland , caused by a lack of long caves suitable for nesting in the 
dark. The consequent increase in predation when compared to that in 
Fiji might have reduced nest and brood size making the nest as 
inconspicuous as possible and the nestling period as short as possible . 
Contrary to this proposal is the theoretical consideration given by 
Perrins ( 1977 ) . He suggests that birds laying very small clutches 
and/or having very long incubation periods (Whi te-rumped Swiftlets would 
be covered by both criteria) will be unlikely to have evolved a 
reduction in clutch size solely as a result of predatory pressure 
because the increased risk involved in laying an extra egg would be 
small ( 5% in examples used by Perrins , probably 15% in the White-rumped 
Swiftlet , which does not lay on consecutive days ) compared with doubling 
the number of young raised . 
The predation proposal looks less likely when one considers how 
little the reduction of the height of the nest cup would contribute to 
hiding a nest from a predator . Recent measurements in Queensland caves 
have shown that nest sizes vary even within a single cave (Pecotich 
pers . conm. ) .  
9 6  
This could mean that the published data may not give a true picture of 
nest size in Queensland - though the greater exposure of terraereginae 
nests to sunlight and so possibly to predation remains real . The 
Caroline Swiftlet (A. inguieta) has one subspecies (A. i .  rukensis )  that 
lays a clutch of two in deep , completely dark caves , and two subspecies 
(A. i .  inguieta) and (A. i .  ponapensis ) that lay single-egg clutches in 
less dark situations (Brandt 1966 ) . An interesting aspect is that the 
subspecies with the larger clutch is more often found nesting singly or 
in small groups as well as in large colonies . As solitary nesters are 
more prone to predation , the finding of the dark-nesting subspecies in 
solitary situations may indicate that nesting in the dark section of 
caves reduces predation to a very low level . One untested possibility 
is  that the Queensland nests that are in total darkness may be larger 
(within the limits imposed by the materials available to build it)  than 
those in the twilight , but if this is so why do they not have two eggs? 
The theory that predation pressure can influence the size of the 
clutch ( Snow 1978) would suggest that White-rurnped Swiftlets nesting in 
the total darkness of a cave will suffer less predation than those 
nesting near the entrance or under overhanging rock . If predation had 
caused the White-rurnped Swiftlet to alter its clutch size we would 
expect those subspecies that use dark caves to be consistent in 
producing a larger clutch than those nesting in lit locations . This , 
however , is not the case . Terraereginae generally nests in lit 
locations (Pecotich 1982 ; Srnyth et al . 1980) , while chillagoensis (pers . 
obs . ) and the Samoan subspecies spoctiopygius (Whitrnee 1875 ) , which nest 
in dark caves , each produce a single-egg clutch . The other subspecies 
known to produce two-egg clutches do so regardless of whether the nests 
are concealed by darkness or not . Known examples follow. On 
Bougainville Island reichenowi has been found nesting in abandoned mine 
shafts and under a dead leaning tree (Haddon 1981 ) . 
9 7  
In Tonga, townsendi produces its two-egg clutches in sea caves where 
some nests are only 3m from the entrance (M.  Potts pers . comm. ) .  In New 
Caledonia leucopygia also appears to nest in twilight situations 
(Hannecart & Letocart 1980) . 
A non clutch-size strategy 
One means that both Queensland subspecies may use to overcome a 
smaller clutch than that of the Fi jian birds is to produce more than one 
clutch, apart from replacements . This suggestion is not new; Banfield 
( 1912 ) suggested that the swiftlets on Dunk Island may rear four 
clutches in a breeding season and Smyth et al ( 1980) give some credence 
to the suggestion , adding that they have been found breeding from July 
to April . However, they found only four eggs in July and only one in 
April , compared with several hundred found in October , November and 
December , the three months that are clearly the peak laying and 
incubation period. 
On average , a pair of chillagoensis take 27 days to incubate their 
eggs and 47 days to fledge their chicks . Thus it takes a pair of 
Queensland swiftlets two and a half months to raise a single brood 
(ignoring any time nest building may take) . To raise the four broods 
suggested by Banfield would take 10 months without any time for building 
a nest . However , the summary (Smyth et al ) of Queensland breeding data 
(which included the colony Banfield wrote about ) in no way indicates 
that the colony is in peak breeding activity for that long . In fact the 
Queensland breeding season seems little longer than the Fiji season and 
the activity of early layers , the production of late replacement 
clutches and annual variation in the commencement and termination dates 
would better explain the extended , though light , tail-end portion of the 
breeding period seen in the Queensland data. 
Three other factors rule against Queensland birds breeding for ten 
months . 
98  
Firstly, the large seasonal variation in  rainfall would indicate a large 
variation in the abundance of aerial insect prey which would not 
therefore be likely to support breeding for so long if food is the 
limiting factor in determining clutch size . Secondly , if predation is 
the mechanism that holds the clutch size to one , surely breeding for ten 
months would make the parents more vulnerable than when raising as many 
in half the time . Thirdly , the smaller size of the Queensland colonies 
tends to indicate that they are not producing twice as many replacements 
as the Fiji birds . 
Regu1ation of clutch size by stability of food supply 
The generalized assumption that where a population and the 
environment are reasonably stable the clutch size will be at an optimum, 
has been extended by Hogstedt ( 1981 ) . He found that the quality 
( largely determined by food quality and quantity) of the territory held 
by the Magpie (Pica pica) determined both clutch size and adult 
survival . He further suggested that territory quality probably explains 
the finding that in many passerines the most productive clutch size is 
larger than that which is most corrrnon (Klomp 1970) . 
In the Apodidae the quality of the territory is correlated with the 
abundance of flying arthropods , which is inversely correlated with 
rainfall (Lack 1956a, Hespenheide 1975 and Emlen 1982 ) . This , in 
conjunction with Ricklefs????ification of Lacks hypothesis (that 
clutch-size is related directly to the resources available during the 
breeding season and inversely to the density of the population) , 
suggests that there should be a correlation between the evenness of the 
year ' s  rainfall and clutch-size . To test this I have expressed the 
average rainfall of the month with the lowest rainfall during the 
non-breeding season as a percentage of that for the month with the 
highest rainfall during the breeding season , for several localities 
where the clutch size for the White-rumped Swiftlet is known. 
9 9  
By following this procedure the resulting figure should be 
comparable between tropical localities . The data from Koronivia (near 
the Nasinu Caves ) gives 3?? while those from Tully and Chillagoe (which 
are near the main breeding caves of the two Australian subspecies ) give 
13% and 2% respectively . It may be argued that the total rainfall for 
the year will  be more important than the variance between the wet and 
dry season. However , total rainfall seems less important than the 
seasonal variation , for , while the Tully district has a higher annual 
rainfall than Koronivia ,  Chillagoe ' s  is much lower yet both Australian 
subspecies produce a clutch of one . Full climatological comparisons are 
given in Appendix lA to 1D . 
Rainfall data from both the dry and wet sides of New caledonia 
indicate low variance ( 3?? and 24% respectively) ,  more like that of Fi ji 
than Queensland. These data are consistent with my finding that low 
variance corresponds with the larger clutch of two . 
Savannah is  not the only example of a harsh climate ( one having 
extremes) and there may be some carry-over from experimental work in 
other situations . For example , it has been suggested ( ?alomonsen 1972 ) 
that because Arctic birds sometimes do not breed at all in inclement 
years , laying a small clutch would be a compromise between breeding and 
not breeding. In short , it would be expected that Arctic birds would 
have smaller clutches than temperate birds . Evidence for this reversal 
of general predictions comes from a study on 15 passerine species 
( Jarvinen 1986 )  that breed in southern Finland as well as at a mountain 
site in Arctic Lapland. Only one of the species produced a larger 
clutch in the more extreme climate . This view had been predicted by 
Kendeigh' s ( 1976 ) suggestion that a species devotes about the same 
amount of energy to reproduction regardless of where it breeds . 
AUV'cl811 
.U.lSLi3AINn A35SVr.? 
l O O 
The idea of similar energy being put into reproduction regardless of 
environment contradicts the theory of r- and K-selection (MacArthur and 
Wilson 1967 ) when it is applied to the same species . The ' bet-hedging ' 
theory (MUrphy 1968 , Schaffer 1974) attempts to solve the discrepancy 
between the r- and K-selection theory and such observations as given 
above . By pointing out that mortality in unstable environments is 
higher for juveniles than in stable environments , ' bet-hedgers ' will 
produce smaller clutches and concentrate on raising a higher percentage 
than they would from a larger clutch . 
Greenslade ( 1983)  suggests that invertebrates respond to not only r-
and K-selection pressures , but also to adversity or A-selection 
pressures , which migtlt be found in predictably unfavourable conditions . 
One of the suggested responses to this third pressure is reduced 
fecundity. 
Another response to A-selection pressures in birds of harsh 
environments might be to moult flight feathers while breeding during the 
short favourable period. Payne ( 1969 ) suggests that the general pattern 
of non-overlap between breeding and moulting means that similar demands 
. .:? 
on the energy requirements of breeding and moulting operate in both the 
tropic and temperate regions . Moulting while breeding would also tend 
to reduce clutch size as both activities take up large amounts of energy 
and rrutrients . Such an effect has been suggested for Arctic birds 
(Haukioja 1971 ) , and it may be that swiftlets have smaller clutches than 
most swifts because they moult while breeding , whereas swifts breed and 
moult at separate times . Data in Table 1 confirm these relationships . 
Swifts commence moult after laying or after the fledging of their 
chicks . However , because swifts are larger than swiftlets and their 
eggs are proportionately smaller than the adult , swifts could be 
expected to produce larger clutches even with partial overlap of moult 
and breeding. 
Species 
Apus apus 
Apus melba 
Apus berliozi 
AEus acuticaudis 
Apus a.ffinis 
AEus myoEtilus 
Chaetura brachyura 
Chaetura chapmani 
Chaetura boehmi 
Chaetura ussheri 
Chaetura sabini 
Chaetura cineriventris 
Chaetura Eelagica 
Chaetura vauxi 
Cypseloides rutilus 
Nea.fraEus cassini 
Aerodramus fuciphagus 
Aerodrarnus maximus 
Aerod.ram.:ls sRQ<iiopygius 
TABLE 1 
Synch-
ronous 
some 
X 
X 
X 
SYNCHRONIZATION OF MOULT & BREEDING 
Discreet Moult after Clutch Climate 
laying size 
X 2 . 3  temp. 
X 3/4 temp . 
2 . 0  temp. 
X temp . 
X 3 .0  trop . 
X trop . 
X 3 . 8  trop . 
X 2/3 trop . 
X 3 . 0  trop . 
X 4.0  trop . 
X 2 . 5  trop . 
X trop . 
X 4 . 2  temp. 
X 4-6 temp. 
X 1 . 9  trop . 
some some trop . 
2 . 0  trop . 
1 . 0  trop. 
1/2 trop. 
Source/Moult 
Lack & Lack 1951 
Lack & Am 1947 
Brooke 1969 
Brooke 1971a 
Naik et al 1969 
Prigogine 1966 
Collins 1968a 
Collins 1968b 
Brooke 1966 
Brooke 1971a 
Brooke 1971e 
Snow 1962 
Zammuto et al 1979 
Bent 1940 
Collins 1968a 
Brooke 1971a 
Langham 1980 
Medway 1962c 
Tarburton 1986b 
,___. 
0 
,___. 
l Ol 
Whether synchronous moult and breeding restricts brood size or not , 
that both assimilis and chillagoensis moult while breeding means that 
the only remaining variable between the subspecies that is likely to 
affect clutch size is the food supply . Because food supply varies  with 
rainfall ( Figure 5 )  and the rainfall total and pattern for the two 
regions differ so rruch , further consideration should be given to that 
aspect . 
It was the generally low level and high variability in food supply 
rather than an inability of the parents to obtain food that restricts 
clutch size to one . When a pair of chillagoensis were provided with a 
second chick they increased the number of feeding trips to the nest on 
days when food was abundant . The number of extra feeds , however , was 
insufficient to raise a second chick , even in a good season. Food 
availability was too variable .  The number of poor days were too 
frequent and even the maximum number of insects caught in the sweep net 
was below the average number caught in Fi ji . 
If the food was available parents were able to collect sufficient 
for two chicks . When rain ended a dry period even in the poor season , 
their chicks gained weight rapidly (up to one third of adult weight in 
one day) . In the good season parents fledged their single broods in the 
same time as Fijian birds with single broods did .  However , with the 
greater variability of food supply in the poor season , parents took 
significantly ( t1 1  = 3 . 2 ,  P < 0.01 )  longer to fledge single chicks in that season than in the good season . 
The significant relationships between the abundance of flying 
insects and either rainfall or watering by irrigation , and between the 
greater likelihood of finding feeding swiftlets overhead on those 
occasions when the sweep net caught more than average numbers of 
insects , indicate the dependence of this bird ' s  food supply upon the 
rainfall .  Whether swiftlets were feeding overhead or not was not 
significantly correlated to whether rain had fallen (or the irrigation 
sprinklers had been used ) in the previous 24 hours or not . This lack of 
correlation is probably due to the birds foraging in other untouched 
areas after several days of rain . 
1 0 3  
CONCLUSIONS 
Most field data and most models concerned with the regulation of 
clutch size have shown or predicted that clutch size in birds inhabiting 
regions with climatic extremes will be larger than that of close 
relatives living in more uniform environments .  However ,  the finding of 
this study , that chillagoensis is  unable in its savannah environment to 
raise an artificially enlarged brood of two which assimilis normally 
does in the more uniform climate of Fijian rainforest , is an exception. 
This section has shown a number of theories to be ineffective in 
explaining this phenomenon . Because enlarging nests did not increase 
the fledging rate I have shown that nest size is  not effective in 
controlling clutch size in chillagoensis . Because predators cannot 
reach the nests of spoctiopygius , the number of chicks in a nest cannot 
influence the rate of predation. Because chiL 'oensis is the mainland 
subspecies and yet has a smaller clutch than ass1 lis , which is  the 
reverse of that predicted by the theory of "competi re release"  on 
islands , that theory cannot explain the smaller clutcl ize of 
chillagoensis . Finally, chillagoensis does not have enc, ' time to 
compensate for its smaller clutch by raising two consecuti\ 'Jroods in 
the normal manner. This is due to the shortness of the wet ? ?on , 
which is  shown to coincide with an abundance in the food supply ?n 
other words , food is the limiting factor that prevents chillagoeru. 
from raising two chicks at one time . A newly discovered response tc. 'e 
shortness of the period when food is abundant has been evolved by 
chillagoensis to enable it to raise two chicks within a season. This is 
examined in the next section . 
Further evidence that food is  the limiting factor restricting the 
clutch size of chillagoensis to one is shown by the following . 
1 0 4  
Broods with an extra chick had significantly shorter wings than those in 
natural broods of one by the eighth day , while the average weight of the 
same chicks fell significantly behind that of the two-chick broods 
earlier than assirnilis chicks from artificially enlarged broods . This 
is taken to indicate the greater difficulty chillagoensis has in 
collecting adequate food for two chicks . 
That chillagoensis made significantly fewer feeding visits to chicks 
in the poor season than in the good season and that two-chick broods 
were not fed significantly more often than single-chick broods in either 
the poor or good season , indicates that this bird is struggling to 
adequately feed one chick in a poor season and cannot feed two chicks 
even in a good season. The correspondence of the days of largest weight 
gain to the first days of each bout of rain further demonstrates the 
need for frequent rains in the maintainance of high food levels . 
Together , these data demonstrate that a lack of available food was the 
major cause preventing chillagoensis from fledging significantly more 
chicks from manipulated two-chick broods than from natural single-chick 
broods . 
Section 4 
A Novel Strategy in Reproduction 
CHICK INCUBATION 
INTRODUCTION 
1 0 5  
The breeding biology of birds has been extensively studied and it 
would seem that every conceivable pattern for accomplishing the 
essential task of incubating the embryo in birds ' eggs has been 
recorded . The basic categories are incubation by : 1 )  both parents , 2 )  
one parent only , 3)  other adult conspecifics , 4)  other species ( e . g . , 
cuckoo) ,  and 5 )  non-animal heat ( e . g . , megapodes ) (Van Tyne & Berger 
1976 ) . All five strategies depend on nearly constant care by an adult .  
With the exception of minor variants within categories all five have 
been known for nearly 150 years (Maunder 1854 , Serventy & Whi ttell 
1962 ) . However , there is a sixth category , incubation by nestlings . 
This incubation pattern was first reported only recently (Tarburton & 
Minot 1987 ) yet appears to be normal practice in chillagoensis . In this 
population a single egg is incubated by the adults and a? second egg is 
incubated by the first nestling before it leaves the nest . Thi s  
behaviour reduces the time between successive nestlings and the time 
spent incubating by adults . Alternative evolutionary strategies , such 
as shorter development period , are constrained by an unreliable food 
supply which necessitates a long nestling period . 
Because assimilis produces two-egg clutches and chillagoensis was 
known to produce only single-egg clutches (Smyth et al 1980) , it  was 
assumed that the Queensland bird was disadvantaged.  
1 0 6 
However , when daily nest inspections designed to gather data for the 
manipulations of Section 2 revealed an egg in the nest of a well  grown 
chick for the second time , I began to question my previous conclusion 
that in the first instance the egg was either mistakenly placed or was a 
case of "dumping" .  Within a few days a number of other cases appeared 
and so a search of nests in other caves was made to determine whether 
the phenomenon was just the result of my disturbance of the two colonies 
I was working each day , or whether it was a normal event in undisturbed 
caves .  The practice was found to be common in all breeding colonies 
inspected. 
METHODS 
During the daily checks of numbered nests in Gordale Scar Pot and 
Guano Pot the appearance of eggs under chicks and their subsequent 
history was recorded . These data demonstrated that the average age of a 
chick at the time it was given an egg to incubate was 32 days . As the 
chicks ' average wing length at this age was almost 70 rrm,  it was decided 
to inspect nests for eggs in other caves when the chicks had wings 
larger than about 70 rrrn . Although this method means that some chicks 
would have been inspected before the second egg was laid , the results 
still show that most birds provide their first chick with an egg to 
incubate .  
It is possible that the second egg was not produced by the mother of 
the first chick. Another female may have produced the egg and "dumped" 
it under any chick of adequate age and not already incubating an egg.  I 
therefore decided to take two eggs from under chicks to see if they were 
replaced before the minimum time taken for an egg to be replaced . 
1 0 7  
RESULTS 
Although chillagoensis was unable to raise artificial broods of two , 
in the good season of 1985/86 , 86% (n=140) of large chicks (wings 
greater than 70 rrm) were found to be incubating an egg. The proportion 
of incubating chicks was reduced to 68% (n=25 ) in the poor season of 
1986/87 . Nestlings exhibited active , responsive incubation behaviour. 
When an egg was removed from under a nestling and placed beside it or 
the nestling was placed at the edge of the nest , the nestling moved to 
resume incubation. Thus the presence of an egg in the nest was 
sufficient external stimulus to release incubation behaviour in the 
nestling. 
The timing of the laying of the second egg was such that it would 
not hatch before the incubating chick fledged . The mean incubation 
period was 26 . 9  ? 1 . 2  days (+  S .D . , n = 40) and the mean nestling period 
was 46.  9 ? 2 .  2 days (n = 13) . The mean age of a chick at the laying of 
a second egg was 33 . 4  + 1 . 3  days (range 23-47 days , n = 24) . When the 
incubating chick fledged , the adults completed any remaining incubation. 
Neither of the chicks that had their eggs removed , and neither of 
two other chicks which lost their eggs through other means , received 
replacements before the minimum period ( eight days ) required for a 
brooding female to replace its own lost egg . In addition , one of the 
chicks to which a manipulated egg was given , was also given a second 
egg , presumably by its mother . 
Despite the differences in incubation strategies the annual 
production of chillagoensis was similar to that of assimilis in good 
years. A breeding pair of assimilis averaged 1 . 1  ? 0 . 11 (x  ? S . E . , n = 
39 ) fledged young and a breeding pair of chillagoensis averaged 0 . 9  + 
0 . 15 (n  = 27 ) fledged young in good seasons . 
PLATE 2 White-rumped Swiftlet (Aerodrarnus spoctiopygius chillagoensis )  
chick incubating its parents ' egg. 
Chicks of both subspecies may be separated from adults by their light 
tipped primaries , which, until their last week in the nest , do not 
extend beyond the tail . However , only chillagoensis chicks will be 
found incubating an egg. The egg in the photo was pushed into view so 
that it could be photographed .  The chick ' s  eyes are shut , as is  most 
often the case until a few days before flying . 

1 0 8 
Although the fledging rate of 0 . 6  chicks per poor season for 
chillagoensis shows that this adaptive behaviour is unable to maintain 
the productivity of good seasons during poor seasons , the rate would be 
only half this without the delayed laying of the second egg. 
CONCLUSIONS 
Without incubation by nestlings at least 148 days would be required 
to rear two nestlings - one at a time - from egg to fledging at a single 
nest . This assumes 27 days incubation , 47 days as a nestling and laying 
on the day the first chick fledges . Incubation of the second egg by the 
first nestling makes it possible to save more than three weeks in the 
time needed to rear two young . (This assumes the second egg is laid 
when the first chick is 23 days old and the two chicks will be fledged 
in 121 days) .  
The incubation of a second egg by the chick from the first egg 
reduces both the time spent incubating by adults and the interval 
between successive nestlings . As nestlings do not incubate in other 
populations of this species , where a single clutch of two is incubated 
by the parents , the behaviour appears to be ecologically facultative . 
1 0 9  
Section 5 
DISCUSSION 
The 21 species of swiftlets generally inhabit the stable environment 
characterized by the tropical rainforests of the Indo-Pacific region. 
The majority of theories concerned with factors controlling clutch size 
predict low annual mortality in such a uniform environment because of a 
reliable year-round food supply . This low mortality results in high 
competition for food during breeding . Thus , although high survival 
requires only low replacement to maintain population levels , i t  is 
thought to be mainly the high competition for food at the time of 
breeding (rather than the low demand for replacement ) that transmits a 
downward pressure on clutch size . An alternative view is that 
long-lived (K-selected) species such as swiftlets may be maximizing 
their output over a lifetime , which may mean producing a smaller clutch 
than the maximum for which parents can find adequate food . 
While in general terms swiftlets conform to the trend of tropical 
species laying smaller clutches (no species lays more than two eggs) 
than temperate relatives (which lay up to six eggs) , they could be 
raising less than they are capable of , so in Section A of this thesis I 
set out to test which theoretical construct applies to the White-rumped 
Swiftlet nesting in tropical rainforest . I have shown that the Fijian 
subspecies (assimilis)  could not raise one chick more than the two they 
normally raise , and this finding supports the general theory that clutch 
size in small land birds is  restricted to the number of chicks that the 
parents can adequately feed . 
The habit  most swiftlets have of nesting in relatively predator free 
cave sites has been considered analagous to the habit of nesting in 
hollows . 
1 1 0  
As the latter site is seen to provide protection from predators and 
allows the bird to produce and raise larger broods than those nesting in 
open nests , this reasoning may be inappropriately applied to swiftlets . 
As this thesis shows , swiftlets cannot raise an extra chick , thus any 
advantage deriving from safe nesting sites will only operate to improve 
the fledging rate and reduce adult mortality during breeding rather than 
to increase clutch size . However , as these two advantages contribute to 
higher survival and thus further increase the competition for food , they 
themselves will restrict clutch size . 
As the general theories on the regulation of clutch size predict 
that clutches will be larger in environments that have more extreme or 
unpredictable climate , it would be expected that chillagoensis has a 
larger clutch than assimilis . It seemed that this exception to the 
prediction of the theories on clutch size would be worth examining . By 
artificially increasing normal broods of chillagoensis to broods of two 
and comparing their growth with that of single-chick broods , it  was 
learnt that in the first season not one double brood was raised to 
fledging . Because growth was significantly slower in the double broods 
it appeared that chillagoensis could not collect enough food to raise 
two chicks at a time. That the birds were away from the cave feeding 
for a shorter time each day when compared with assimilis , is  taken as a 
consequence of smaller colony size at Chillagoe (reducing the required 
time to travel to feeding areas ) , rather than that chillagoensis is  
having an easier time collecting as much food as assimilis . This does 
not deny that chillagoensis may work less to raise its brood of one than 
assimilis does to raise a brood of two , for that situation could exist 
and chillagoensis still not be able to adequately feed two chicks , even 
in a good year. 
l l l 
In the hope o? learning more about the ability of chillagoensis to 
raise two chicks under dif?ering climatic conditions , the experiments 
were repeated in 1986/7 . Fortunately this season proved to have a 
delayed wet season. 
The ef?ects o? the late rains in the 1986/7 season at Chillagoe 
provided several pieces o? evidence that support the theory that the 
birds were unable to gather enough food to raise two chicks under such 
conditions . Aerial insect netting indicated very low densities o? 
available insects - with noticeable increaSe after the rains . Not only 
did the density o? the available insects increase when the rains 
eventually arrived , but so also did the average size of the available 
insects . More noteworthy was the death o? several chicks ?rom the 
manipulated broods . This did not happen the previous year or in Fi ji . 
When the surviving chicks experienced a weight increase of up to 36% in 
the 24 hours ?ollowing the arrival o? the first good rains it became 
clear that the parents had been hard pressed to adequately feed one 
chick and so were unable to gather sufficient food to raise two chicks . 
Because one o? the chicks in the artificially enlarged broods o? 
both assimilis and chillagoensis generally died by ?alling from the 
nest , it could be reasoned that insufficient nest space was the ?actor 
restricting clutch size . Support for this view comes from the 
observation that the larger chick was just as likely as the smaller 
chick to be the one that ?ell ?rom the nest . Evidence from the 
ecologically similar hirundines infers such a relationship . However , by 
watching chicks compete ?or ?ood and noting how o?ten they fell ?rom a 
nest while settling down after being handled , it is likely that 
disturbance created by increased competition for less food (per chick) 
is the real cause ?or chick loss where additional chicks are given to 
either assimilis or chillagoensis . 
1 l 2  
Enlarging nine chillagoensis nests to the size of those of assimilis 
only to find that one of the twin chicks still ?died or fell out , clearly 
shows that nest size might at best make only a small contribution to 
preventing extra chicks from fledging . That five of the two-chick 
broods which were provided with enlarged nests were also given 
supplemented food without showing improved fledging success , does not 
refute the finding that food supply limits clutch size in assimilis and 
chillagoensis . It might be that the dog food used to supplement the 
chicks ' intake was inappropriate to digestive systems adapted to an 
insect diet , that two feeds an hour or two apart per day were 
insufficient , or that the parents ' saliva used to bind the bolus 
together may be an essential part of the chicks ' diet . The latter is 
suggested because saliva normally helps digest food but also because 
swiftlet saliva has been shown to have nutrient value in its own right 
(Wang 1921 ) ? 
A trend found in both the aerially feeding families of Apodidae and 
Hirundinidae is that larger species lay smaller eggs ( in relation to 
adult size) than do smaller species . This means swiftlets lay large 
eggs relative to the parents size . However , variation in egg size 
itself does not show a significant correlation with the period of 
incuba.tion. This lack of correlation seems unreasonable if one reasons 
that large eggs should hold warmth better than small eggs . However , 
that there is a significant inverse correlation between adult weight and 
incuba.tion period suggests that it is larger adult size rather that; 
larger egg size that accompanies a reduction in the incubation period. 
Obviously other factors may affect this general trend because assimilis , 
which is lighter than chillagoensis , has a shorter incubation period. 
This variation is best explained by the warmer temperatures in Fijian 
caves . 
1 1 3  
Although assimilis and chillagoensis nest at similar sites and 
similar distances from cave entrances , the type of limestone foundation 
and nest materials used are different . That either of these factors 
affects the incubation period is unlikely , as the egg is mostly in 
contact with the air or brooding bird . In addition , assimilis uses a 
wide range of nesting materials in caves away from Nastnu , and in at 
least one nesting site these include grasses . 
It is  at the nesting site where predators concentrate their efforts 
to catch swiftlets . All known predators (with the exception of the cat 
in Fiji ) rely upon the concentration of breeding swiftlets to be able to 
defeat this fast and manoeuvrable bird . Whereas snakes and cats 
generally cannot reach swiftlet nests , they can reach flying individuals 
if there are small constrictions in the passages between the nest site 
and the cave entrance .  Snakes have the advantage of being able  to use 
sections of passage that are narrow or low,  whether dark or light . Owls 
and Goshawks take swiftlets from concentrations of birds at cave 
entrances . Despite these losses , adult White-rumped swiftlets have a 
further life expectancy greater than that of many similar sized 
passerines . This enhanced survival probably results from predator 
avoidance strategies such as nesting in the dark , feeding young at 
infrequent intervals , leaving and entering (where possible) through 
different cave entrances and doing so at high speed . 
While gathering food without being taken as food occupies almost the 
whole of a swiftlets active day , some aspects of its diet tend to 
indicate that it  takes whatever it  can find within the range it  can 
handle . While the birds vary their feeding areas and time of feeding in 
? ... ;- of 
a given area, this behaviour is  best interpreted as groups proceeding to 
. A 
an area known to be productive and working it if the present density is 
rewarding , taking all available prey until the catch rate is low and 
r then moving to other less prefe?ed areas . 
1 1 4 
Because flocks feed more frequently over well watered areas than over 
open savannah shows they do select feeding areas . Finding that food 
boluses of some individuals consisted almost entirely of one species 
indicates that their feeding group had located an area with 
mono-specific high density prey. That food boluses of other individuals 
contained up to six or seven hundred individuals of up to 83 species 
indicates that the group these birds fed with took whatever was 
available as they did not encounter any swarms or other high density 
concentrations of one or two species . 
While this thesis shows that neither assimilis nor chillagoensis are 
raising less than their capacity for gathering food will allow within a 
given season, it also shows that increasing clutch size is not the only 
way to improve production within a breeding season . While chillagoensis 
is shown to be unable to feed two chicks this thesis has uncovered a 
novel strategy that allows it to match the annual production of 
assimilis . By laying a second egg when the first chick is 32 days old , 
at which stage it is probably no longer poikilothermic , both parents are 
able to concentrate on gathering food for the chick doing the 
incubating. As the second egg does not hatch before the first chick 
fledges the parents are able to raise two chicks per season without 
having to attempt that which they cannot do : that is , find enough food 
to raise two chicks at the one time . 
There are several potential advantages to nestling incubation:  1 )  
More chicks can be fed at the time of greatest food abundance .  2 )  Adults 
can forage continuously rather than spending half their time incubating. 
3 )  Predation risk to adults is reduced by making fewer trips into the 
cave . 
These potential advantages suggest broader questions about the 
ecological and developmental constraints that may account for incubation 
by nestlings . 
Why do the Chillagoe birds rear two young in succession? Among 
birds that feed their young in the nest , clutch size is largely limited 
by the number of young that can be fed (Lack 1968 ) . The feeding rate of 
adult swiftlets depends on the abundance of flying insects . In the rain 
forest near Suva (annual rainfall 3200 mm} the swiftlets have a fairly 
stable food supply except when it is raining. However , on the Chillagoe 
savannah (annual rainfall 855 rrm) insects are only abundant for several 
days following infrequent and unpredictable rains . This climatic 
difference seems to account for the swiftlets ' ability to rear two young 
simultaneously in Fiji but only one at a time at Chillagoe . 
For many species of birds , clutch size has been related to annual 
climatic variance with larger clutches found in areas of greater 
variance (Ricklefs 1980) . There are two possible reasons for this . 
Firstly , a bird population in which numbers are severely limited in the 
non-breeding season will experience a competitive release in the 
breeding season . Secondly , adults with a low probability of surviving 
to breed in the next year may increase their breeding effort in the 
current year (Williams 1966 ) . The rainforest near Suva has a low 
. 
climatic variance . The ratio of mean rainfall in the driest month to 
mean rainfall in the wettest month is 1 : 2 . 6 .  In the savannah at 
Chillagoe it is 1 : 54.  However ,  it is in the rainforest rather than the 
savannah, that the swiftlets have the larger clutch. If chillagoensis 
experiences any competitive release it is insufficient to rear more than 
one nestling at a time.  However, they do prolong their breeding effort 
by rearing a second offspring in succession to the first . The success 
of this effort rests in part on incubation by the first nestling .  It is 
worth noting that the time taken for eggs to hatch when incubated or 
largely incubated by chicks , was not significantly different ( T38 = 0 . 2 ,  
ns ) to that taken when hatched by adults . 
1 1 6  
When parents incubated , incubation took 26 . 9  ? 0 . 2  days ( x  ? SE ,  n = 32 ) 
and when chicks incubated , incubation took 26 . 8  i 0 . 5  days (n = 8) . 
Why don ' t  the young swiftlets develop faster? Lack ( 1968) argued 
that evolutionary selection for the rate of nestling development is 
driven by the average food supply during the nestling period . There 
should also be an effect from the temporal variation of the food supply 
for nestlings . Slowly developing nestlings do not require food at a 
fast rate and can better withstand several days without food as compared 
with quickly developing young. The generality of Lack ' s  hypothesis of 
average food supply has been challenged recently ( Shea & Ricklefs 1985 ) , 
but it  seems to apply to chillagoensis especially if modified to include 
the effect of a variable food supply . As a group , swifts have long 
nestling periods for their size and during the nestling phase must cope 
with adverse weather which severely restricts feeding by adults . A long 
incubation period is linked to the overall slow rate of development for 
a species (Lack 1968 ) . Thus slow growth as a nestling is a 
developmental constraint on the rate of growth during the incubation 
phase . Amongst swifts the incubation period is dependent on both the 
nestling period and the size of the species : incubation period ( days) = 
0 . 28 ? 0 .02 nestling period (days ) - 0 . 073 ? 0 .012 female weight (grams ) 
+ 11 .7  (multiple linear regression of data from 16 species of swifts ?
Section 1 ,  p . 28) . So altering the rate of egg development independent 
of the rate of nestling growth, is not a feasible evolutionary strategy 
for swifts . Incubation by nestlings is one way around this constraint . 
How do nestlings perform what is typically a breeder ' s  behaviour and 
actively incubate the second egg? Work on the endocrinological basis of 
incubation behaviour has often stressed the importance of increased 
levels of progesterone and prolactin at the end of the pre-incubation 
stage for both male and female adult breeders ( Follett 1984) . 
1 1 7  
However , those hormones seem to be as much a result of incubation 
behaviour as they are a cause of it (Brown 1985 )". Apparently little is 
known about the levels of these hormones in nestling birds but it can be 
assumed that the nestling swiftlets will not have been through the same 
hormonal cycle as the adults . Nevertheless , young birds are known to 
have high levels of growth hormone during the first half of the nestling 
period (Bates et al 1962 ) and growth hormone is functionally similar to 
prolactin ( Scanes et al 1983) . It should be interesting to look at the 
hormone levels in incubating and non-incubating nestlings to see if 
there is feedback from behaviour to hormones such as might , for example ,  
lead to increased vascularization of the skin in the ventral region 
( swiftlets do not develop brood patches ) . 
Alternatively it might be that the chicks are passive in their 
settling over the eggs as in most nests there is little room to sit 
beside an egg , without moving the egg to one side . (Although I have 
found one adult sitting beside its egg) . 
Incubation by nestlings is presumably linked to the cost of 
incubating relative to the gain in inclusive fitness . The cost of 
incubation . in energy terms would seem to be fairly low. Using estimates 
of energetic cost (Drent 1975 ) , it should be less than 5% of the basal 
metabolic rate of a bird incubating a single egg at an ambient 
temperature of 23?C . Unlike an adult ,  a nestling has no conflicting 
behavioural activities . The gain is a higher probability that a sibling 
will be reared. However ,  whatever the proximate or ultimate explanation 
for the behaviour , the nestling is exhibiting behaviour that in a 
similar adult context would be totally appropriate . In thi.s sense there 
is a parallel to other ' helping ' behaviour such as the provisioning 
behaviour of communal breeders where the response to a feeding stimulus 
is  identical for both juveniles and adults (Jamieson & Craig 1987 ) . 
1 1 8 
Why don ' t  more species use this incubation method? I am confident 
that similar behaviour will  be found in other species in similar 
circumstances . These circumstances would probably include : 
1 )  no post-fledging parental care , 
2 )  a savannah habitat where limitation of food supply is indicated by 
one-egg clutches and where survival during the non-breeding season 
is low, and 
3) conditions suitable to breeding which last long enough to feed two 
nestlings in succession. 
1 1 9  
REFERENCES 
ADAMS , L .E .G .  1957 . Nest Records of the Swallow. Bird Study 4: 28-33 . 
AFFRE,  G .  & L .  Af'fre. 1967 . Observations automnales sur lli1e colonie de 
martinets pales , (Apus pallidus) , a Toulouse . Alauda 35 : 108-117 . 
ALEXANDER, B .  1898. Further notes on the Cape Verde islands . Ibis 7 :  
277-285 . 
ALI , S .  & S .D .  R ipl ey . 1970 . Handbook of the birds of India & 
Pakist3.11 :4 .  frogmouths to pitta' s .  Bombay . 
ARMSTRONG, J . S .  1932 . Handlist to the Birds of Samoa. London. Bale sons 
& Danielson .  
ARN-WILLI , H .  1960. Biologische studien urn alpensegler . Verlag 
Vogt-Schild , Solothum. 
ASHMOLE, N .P .  1961 . The biology of certain terns. D . Phil .  thesis 
(Oxford ) : cited by Lack & Moreau 1965 . 
-----. 1963 . The regulation of numbers of tropical oceanic birds . Ibis 
103b : 458-473 . 
BALDWIN, P .H.  & N.K.  ZACZKOWSKI . 1963 . Breeding Biology of the Vaux 
Swift . Condor 65 : 400-406 . 
BANFIELD, E . J. 1912 . My tropical isle . London. Fisher . 
BARTELS, M.  1931 . Beobachtilllgen lli1 brutplazen des alpenseglers . 
(Micropus melba melba) (L . ) J .  fur Ornithol . 79 : 1-28 . 
BATES, R.W. , R.A.  MILLER, & M.M.  GARRISON. 1962 . Evidence in the 
hypophysectomized pigeon of a synergism among prolactin, growth 
hormone , thyroxine , and prednisone upon weight of the body , 
digestive tract , kidney and fat stores . Endocrinology 71 : 345-360 . 
BECKING, J.H.  1971 . The breeding of (Collocalia gigas ) .  Ibis 113 : 
330-334. 
BEHLE, W.H. 1973. Clinal variation in White-throated Swifts from Utah & 
the Rocky Molli1tain region . Auk 90 : 299-306 . 
BELCHER, W.J.  1972 . Birds of Fi ji in Colour . Auckland . Collins . 
BENT, A.C .  1940. Life histories of American Cuckoos , Goatsuckers , 
Hummingbirds & their alies . U . S .  Nat . Mus . Bull . 176 : 254-271 . 
----- . 1942 . Life histories of North American flycatchers , swallows , 
larks and their allies . u. s .  Nat . Mus . Bull . 179 . 
BOLOTIN, E . I .  & G.V. KOLONIN. 1979 . New data on the fauna of ixodid 
ticks in the Primorsky Krai ( Soviet far east ) .  Zool . ZH. 58 : 
435-436 . 
BRANDT, J.H.  1966 . Nesting notes on the Collocalia of Micronesia & 
Peninsular Thailand . Oologists Rec . 40 : 61-68 . 
BROOKE, R.K.  1957 . On the breeding of' the Whi te-rumped Swift (? 
caff'er caffer) in Gatooma. Ostrich 28 : 164-169 . 
1 2 0 
-- . 1963. Little & Palm Swif'ts breeding on man-made structures in 
Rhodesia. Ostrich 34: 27-35 . 
--. 1966 . The bat-like spinetail Chaetura boehmi Schalow (Aves) . 
Arnoldia 2 :  1-18 . 
- . 1969 . Age characters in swif'ts . Bull . Brit .  Ornith. Club 89 : 
78-81 . 
--. 1970. Geographical variation & distribution in Apus barbatus , 
A .bradf'ieldi and A.niansae (Aves : Apodidae) .  Durban Mus . Novit . 8 :  
363-374. 
----- .  1971a. Breeding of' Swif'ts in Ethiopian Africa & adjacent islands . 
Ostrich 42 : 5-36 . 
---. 1971b. Taxonomic notes on some lesser known Apus swif'ts . Bull . 
Brit . Ornith. Club 91 : 33-36 . 
-- . 1971c . Geographic variation in the Alpine Swift ? 
(Tachymarptis)  melba (Aves : Apodidae ) .  Durban Mus . Novit . 9 :  
131-143. 
---. 1971d. Geographic variation in the swif'ts Apus horus & Apus 
caff'er (Aves : Apodidae) .  Durban Mus . Novit . 9 :  29-38 . 
----- . 1971e . Taxonomic & distributional notes on the African 
Chaeturini . Bull . Brit . Ornith. Club 91 : 76-79 . 
BROWN, R .E .  1985 . Hormones and paternal behaviour in vertebrates . Arner. 
Zool . 25 : 895-910 . 
CAM, C .R .  & A.E .  CAM. 1975 . A report on Nankeen Kestrel banding in New 
South Wales . Aust . Bird Bander 13 : 27-30 . 
CHARNOV, E .L .  & J .R .  KREBS. 1974 . On clutch-size and fitness . Ibis 116 : 
217-219 . 
CHIH-TUNG, K. 1973 . Preliminary studies on the breeding behaviour of' the 
House Swift . Acta Zoologica Sinica 19 : 293-304. Eng sum. p303-304. 
CLUNIE , F .  1984.  Birds of' the Fiji bush. Suva. Fiji Museum. 
CLYMO, R .S . 1983 . A trio of' troubles . Brit . Ecol : Soc . Bull . 14 : 
150-155 . 
CODY, M.L .  1966 . A general theory of' clutch size . Evolution 20 : 174-184. 
----- . 1971 . Ecological aspects of' reproduction. Avian Biology (Ed. by 
D . S .  Farner & J.R .  King) pp 461-512 . New York . Academic Press . 
COLE, L . J .  1917 . Determinate & indeterminate laying cycles in birds . 
Anat. Record 11 : 504-505 . 
COLLINS , C . T .  1963. The "downy" nestling plumage of' swif'ts of' the genus 
Qypseloides . Condor 65 : 324-328 . 
--- -. 1968a. The comparative biology of two species of swifts in 
Trinidad, West Indies . Bull . Florida State. Mus .  11 : 257-320 . 
1 2 1  
---. 1968b . Notes on the biology of Chapman ' s Swift Chaetura chapmani 
(Aves :Apodidae) .  Am. Mus . Novit . 2320 : 1-15 . 
-----.  1972 . Weights of some birds of north-central Venezuela. Bull . 
Brit . Ornith. Club . 92 : 151-153 . 
-- . 1973a. Notes on survival and band wear in White-throated Swifts . 
W .  Bird Bander 48 : 20-21 . 
---. 1973b. Development of temperature regulation in the House Swift . 
Pavo 11 : 1-11 . 
-- . & M.J .  LANDY. 1968 . Breeding of the Black Swift in Veracruz 
Mexico . Bull . Sthn.  Calif . Acad . Se . 67 : 266-268 . 
--. & R.K .  BROOKE. 1976 . A review of the swifts of the genus 
Hirundapus (Aves :Apodidae ) .  Contributions in science ; Los Angeles 
County Museum 282 : 1-22 . 
CORT , M. P.  1975 . A New species of Pellonyssus , Clark & Yunker, from the 
Ethiopian zoogeographical region (Acarina:Dermanyssidae ) Acarologia 
(Paris )  16 : 208-213 . 
Crouther, M.M. 1983. Observations on White-rumped Swiftlets breeding at 
Finch Hatton Gorge 1981-82 . Aust . Bird Watcher . 10 : 1-11 . 
CULLEN, J .M. & ASHMOLE , N.  P .  1963 . Black Noddy , Anous tenuirostris on 
Ascension Island. Part 2 .  Behaviour . Ibis 103 : 423-446 . 
DAVIS ,  W.B .  1937 . A Vaux Swift and its young. Condor 39 : 222-223 . 
DEMENTIEV, G .P . , N.A. GLADCOV, E . S .  PTUSHENKO, E .P .  SPANGENBURG , & A.M. 
SUDILOVSKAYA. 1951 . Birds of the Soviet Union , Vo1 . ? 1 . (English 
translation 1966 ) . Israeli program for scientific translations , 
Jerusalem. 
DEXTER, R. 1952 . Extraparental cooperation in the nesting of Chimney 
Swifts . Wilson Bull . 64 : 133-139 . 
--- . 1981 . Nesting success of Chimney Swifts related to age & the 
number of adults at the nest , & subsequent fate of the visitors . J .  
Field Ornith. 52 : 228-232 . 
DONNELLY, B .G. 1974. Vertical zonation of feeding swallows ano swifts at 
Kariba, Rhodesia. Ostrich 45 : 256-258 . 
DRENT, R .  1975 . Incubation . Pages 333-420 In Avian Biology Vol . 5 ( eds . 
Farner, D . S .  & J .R .  King) . London . Academic Press . 
DYRCZ , A. 1984. Breeding biology of the Mangrove Swallow (Tachycineta 
albilinea) & the Grey-breasted Martin (Progne chalybea) at Barro 
Colorado Island, Panama . Ibis 126 : 59-66 . 
EMLEN, S .T .  1982 . The evolution of helping 1 .  An ecological constraints 
model . Am. Nat . 119 : 29-39 . 
FEARE , C .  1984. The starling . Oxford . Oxford Univ. Press . 
1 2 2  
FENTON, M.B .  1975 . Acuity of echolocation in Collocalia hirundinaceae 
(Aves : Apodidae) ,  with comments on the distributions of 
echolocating swiftlets & molossid bats . Biotropica 7 :  1-7 . 
FILIPPOVA, N.A. & I .V. PPlNOVA. 1975 . Ixodes caledonicus Nuttall 1910 
( Ixodoidea: Ixodidae ) ,  a little known parasite of wild birds , from 
the fauna of the U . S . S .R .  Parazitologiya ( Leningrad) 9 :  339-347 . 
FINLEY, W.L.  & I .  FINLEY. 1924. Changing habits of Vaux Swift & Western 
Martin. Condor 26 : 6-9 . 
FISCHER, R .B .  1958 . The breeding biology of the Chimney Swift Chaetura 
pelagica (Linnaeus) .  N .Y .  State Mus .  & Se . Serv . Bull . 368 : 1-141 . 
FOLLETT , B .K .  1984 .  ?rds . Pages 283-350 In Marshall ' s  physiology of 
reproduction , 4 Edition ( ed .  Lamming , G . E . ) .  Edinburgh. Churchill 
Livingston. 
FRITH, H .J . (Editor) . 1976 . Readers Digest complete book of Australian 
Birds . Sydney. 
GADGIL, M.  & W.H. BOSSERT , 1970 . Life historical consequences of natural 
selection.  Am. Nat . 104: 1-24 . 
GARNETT, M.C .  1981 . Body size , it ' s  heritability and influence on 
juvenile survival among Great Tits , (Parus Major) . Ibis 123 : 31-41 . 
GAUD, J .  & W. T .  ATYEO. 1976 . Ascouracarinae , a new subfamily of 
Syringobiidae , feather mites . Ascarologia ( Paris)  18 : 143-162 . 
GIBB , J .A.  1960. Populations of tits and goldcrests and their food 
supply in pine plantations . Ibis 102 : 163-208 . 
GIBSON, D .D .  1981 . Migrant birds at Shemya Island Aleutian Islands 
Alaska. Condor 83 : 65-77 . 
GIBSON-HILL, C .A.  1948. The Malayan Swiftlets . Malayan Nat . J .  3 :  
190-200. 
GOODERS, J .  1975 . The great book of birds . New York . Dial Press .  
GOODMAN, D .  1974. Natural selection and a cost ceiling on reproductive 
effort . Am. Nat . 108 : 23-31 . 
GREENSLADE, P . J .M.  1983 . Adversity selection and the habitat templet . 
Am. Nat . 122 : 352-365 . 
GRIFFIN, D .R .  1958 . Listening in the dark . New York . Yale Uni . Press . 
HADOON, D .  1981 . Birds of the North Solomons . Wau Ecol . Inst . Wau. 
HAILS, C . J .  1984. The breeding ecology of the Pacific Swallow (Hirundo 
tahitica) in Malaysia.  Ibis 126 : 198-211 . 
HANNEC.ART, F .  & Y.  LETOC.ART. 1980 .  Oiseaux de Nouvelle Caledonie et des 
Loyautes ,  vol . l .  Auckland. Clark & Matheson. 
HARRIS , M.P .  1982 . Seasonal variation in fledging weight of the Puffin 
Fratercula arctica. Ibis 124 : 100-103 . 
1 2 3 
HARRISSON, T .  1976 . The ?ood o? Collocalia swi?tlets (Aves :Apodidae ) at 
Niah Great Cave in Borneo . J .  Bombay Nat . Hist . Soc . 71 : 376-393 . 
HAUKIOJA, E .  1971 . Summer schedule o? some subarctic passerine birds 
with re?erence to postnuptial moult . Rep . Kevo Subarctic Res .  Stat . 
7 :  60-69 . 
--- . & J .  REPONEN. 1969 . On the molt o? the House Sparrow, Passer 
domesticus . Ann. Rep . Ornithol . Soc . Pori . 1968 : 49-51 . 
HEINROTH, 0 .  1922 . Die beziehungen zwischen vogelgewicht , eigewicht , 
gelegegewicht und brutdauer . J .  ?ur Ornithologie 70 : 172-285 . 
HESPENHEIDE, H.A. 1975 . Selective predation by two swi?ts & a swallow in 
Central America. Ibis 117 : 82-99 . 
HESSE, R .  1922 . Die bedertung der tagesdauer ?ur die vogel . Sitzber . 
Nathist . Ver. Bonn ?or 1922-1923 , pp . A 13-17 . 
HOGSTEDT, G .  1981 . Should there be a positive or negative correlation 
between survival o? adults in a bird population & their clutch 
size? Am. Nat . 118 : 568-571 . 
HOOGSTRAAL ,  H . , S .  KADARSAN, M.N.  KAISER & P .F .D .  VAN DEENEN. 1974. 
Ornithodoros (Alectorobius) Collocaliae , new species ( Ixodoidea: 
Argasidae) ,  parasitizing cave swi?tlets (Aves : Apodidae ) in Java. 
Ann. Entom. Soc . Am. 67 : 224-230 . 
11fJ7. 
JAMIESON, I .  G .  & J. L.  CRAIG.I\Cri tique o? helping behaviour in birds : 
departure ?rom ?ctional explanations . In Perspectives in Ethology 
7 :  ( ed .  Bateson, P .  & P .  Klop?er) New York. Plenum. 
JARVINEN, A .  1986 . Clutch size o? passerines in harsh environments .  
Oikos 46 : 365-371 . 
JOHNSTON, D .W.  1958 . Sex & age characters & salivary glands o? the 
Chirmey Swi?t . Condor 60 : 73-83 . 
KAINADY, P.J!;G .  1976 . Presence o? down-like nestling plurr:age in Pallid 
Swi?t "'(Apus pallidus ) ? Bull . Basrah Nat . Hist . Mus.  3 :  139-142 . 
KENDEIGH, s . c .  1952.  Parental care and its evolution in birds . Illinois 
Bio . Monog. 22 : 1-356 . 
--- . 1976 . Latitudinal trends in the metabolic adjustments o? the 
house sparrow. Ecology 57 : 509-519 . 
KLOMP, H. 1970. The determination o? clutch-size in birds , A review. 
Ardea 58 : 1-124 . 
KOLONIN, G .V. & E . I .  BOLOTIN. 1977 . Rare species o? ixodid ticks in the 
Primorski krai .  Zool . zh 56 : 475-476 . 
KOSKIMIES, J .  1950. The li?e o? the Swi?t , Micropus apus (L ) . in 
relation to the weather . Ann. Acad . Sci . Fenn. Ser . A. 15 : 1-151 . 
LACK, D.  1947 . The signi?icance o? clutch size . Ibis 89 : 302-352 , 90 : 
25-45 . 
1 2 4  
-- . 1954 .  The natural regulation of aniriBl numbers . Oxford Uni . 
Press . London. 
--- . 1956a. Swifts in a tower . London. Methuen. 
-- . 1956b . A review of the genera & the nesting habits of swifts . Auk 
73 : 1-32 . 
-- . 1956c . Further notes on the breeding biology of the swift (? 
?) Ibis 98 : 606-619 . 
-----.  1966 . Population studies of birds . Oxford . Clarendon Press . 
---. 1968 . Ecological adaptations for breeding in birds . London . 
Methuen. 
---. & H.  ARN 1947 . Die bedeutung der gelegrobe beim alpensegler. Orn. 
Beob . 44: 188-210 . 
---. J .  GIBB & D .F .  OWEN. 1957 . Survival in relation to brood-size in 
tits . 
Proc . Zool . Soc . Lond . 128 : 313-326 . 
-- . & E .  Lack. 1951 . The breeding biology of the Swift Apus apus . 
Ibis 93 : 501-546 . 
- & ---. 1952 . The breeding behaviour of the Swift . Br. Birds 45 : 
186-215 . 
----. & R . E. MOREAU, 1965 . Clutch-size in tropical passerine birds of 
forest & savanna. L ' Oiseau 35 : 76-89 . 
---- . & D . F .  Owen. 1955 . The food of the swift . J .  Anim. Ecol . 24 : 
120-136 . 
LANGHAM, N.  1980 . Breeding biology of the Edible-nest Swiftlet 
(Aerodramus fuciphagu8) . Ibis 122 : 447-460 .  
LAYARD, E .L .  1876 . Description of a new species of a flycatcher 
(Mlfiagra) from the Fijis , & some remarks on the distribution of the 
birds found in those islands . 3 , Ibis 6 :  389-394 . 
----- .  & E . L . C .  La.yard , 1878 . Notes on the avifauna of New Caledonia. 
4 , Ibis ,  2 :  250-267 . 
, 1882 . Avifauna of New Caledonia.  4 , Ibis 6 :  493-546 . 
LEACH, J.A.  1928 . Notes made during a holiday trip to New Caledonia.  Emu 
28:  20-42 . 
LILL , A. 1975 . The evolution of clutch size & male "chauvinism" in the 
White-bearded Manakin . Living Bird 13:  211-231 . 
LOBKO-LOBANOVSKI , M. I .  1956 . On the biology of the Needle-tailed Swift 
in the Sakhalin. Zool . Zhurnal 35 : 1586-1587 . 
LOHRL, H.  1973 . Einflusser brutraumflache auf die gelegegrosse der 
kohlmeise (Parus riBjor ) . J .  fur Ornithologie 114: 339-347 . 
1 2 5  
LUDESCHER, F . B .  1973 . Sumpfmeise (Parus palustris palustris L . ) und 
weidenmeise (Parus montanus salicarius Br . )  als symatrische 
zwillingsarten. J .  fur Ornithologie 114:  3?56 . 
MAA, T . C .  1980 . A preliminary revision of the genus (Myophthiria )  
Rondani , (Diptera:Hippoboscidae ) .  Revue Suisse Zool . 87 : 835-861 . 
McARTHUR, R.H.  & E . O .  WILSON , 1967 . The theory of island biogeography . 
Princeton. Princeton. Uni .Press . 
MACKAY, R .D .  1968 . Unusual nesting site of Glossy Swiftlets . Emu 68 : 
219 . 
McLACHLAN, G.R .  & R .  LIVERSIDGE. 1978 . Roberts birds of South Africa. 
Gapetown. John Voelcker . 
MARCHANT, S .  1984. Nest-records of the Eastern Yellow Robin Eopsaltria 
australis . Emu 84: 167-174. 
----- . & P . J .  Fullagar, 1983 . Nest records of the Welcome Swallow .  Emu 
83 : 66-74. 
MARSHALL, A.J .  1961 . Breeding seasons and migration. In Biology and 
comparative physiology of birds Vol . 2 .  Ed. Marshall , A.J .  New 
York. Academic Press . 
MAUNDER, S .  1854. The treasury of natural history. London. Longman, 
Brown & Green. 
MAYR, E .  1937 . Notes on New Guinea birds . I ( Collocalia) . Am. Mus.  
Novit . 915 : 1-19 . 
----- . E .  1945 . Birds of the south-west Pacific .  Hertsfordshire . Wheldon 
& Wesley . 
MEDWAY, L .  1962a. The swiftlets ( Collocalia) of Niah Cave , Sarawak part 
1 .  Breeding Biology. Ibis 104: 45-66 . 
--- . 1962b. The Swiftlets (Collocalia) of Niah Cave , Sarawak part 2 .  
Ecology & the regulation of breeding. Ibis 104: 228-245 . 
----- . 1962c. The relation between the reproductive cycle , moult & 
changes in the sublingual salivary glands of the swiftlet 
(Collocalia maxima) Hume. Proc . Zool . Soc . Lon. 138 : 305-315 . 
---. 1962d . The swiftlets (Collocalia) of Java & their relationships . 
J .  Bombay Nat . Hist . Soc . 59 : 146-153 . 
-----. 1966 . Field characters as a guide to the specific relations of 
swiftlets . Proc. Linn. Soc . Lon. 177 : 151-177 . 
-----. 1967 . The function of echonavigation among swiftlets . Anim. 
Behav. 15 : 416-420 . 
----- . & J .D .  Pye . 1977 . Echolocation & the systematics of swiftlets . in 
Stonehouse ,  B .  & c .  Perrins , Evolutionary Ecology. London. 
Macmillan. 
MERCER, R .  1966 . A field guide to Fiji birds . Suva. Fi ji Museum. 
1 2 6  
MICHAEL, C .M.  1927 . Black Swift nesting in Yosemite National Park. 
Condor 29 : 89-97 . 
MILLER, A.H.  1961 . Molt cycles in equatorial Andean sparrows . Condor 63 : 
143-161 .  
MOLLER, A.P .  1982. Clutch size in relation to nest size in the swallow 
(Hirundo rustica) . Ibis 124 : 339-343 . 
MOREAU, R .E .  1941 . A contribution to the breeding biology of the 
Palm-swift eypselus parvus. J .  E .  A:fr . & Uganda Nat . Hist . Soc . 15 : 
154-170. 
----- .  1942a. (Colletoptera affinis)  at the nest . Ostrich 13 : 137-147 . 
-- . 1942b . The breeding biology of (Micropus caffer streubeli i ) ,  the 
White-rumped Swift. Ibis 84: 27-49 . 
-- . 1944. Clutch size : a comparative study with special reference to 
African birds . Ibis 86 : 286-347 . 
MOREL, M.  1967 . Les Oiseaux Tropicaux Elevent-ils Autant de Jeunes 
Qu' ils Peuvent en Nourrir? Le cas de 1agoostica senegala. La Terre 
et la Vie 21 : 77-82 . 
MORGAN, R.W. 1897 . On the nidification of certain South-Indian birds . 
Ibis 5 :  313-323 . 
MOSS , D.  1972 . A statistical analysis of clutch size in the Great Tit 
Parus major L .  Unpublished M .Sc .  thesis ,  Univ. of Oxford . 
MURPHY, G . I .  1968 . Pattern in life history & the environment . Am. Nat . 
102 : 390-404. 
NAIK, R.M. & S .  NAIK. 1965 . Studies on the House Swift , Apus affinis 
(G.E .  Gray) 5 .  Moult cycle in the adults . Pavo 3 :  96?120. 
-- , & A. RAZACK. 1967 . Studies on the House Swift , Apus affinis 
(G .E .Gray) 8 .  Breeding seasons and their regulation. Pavo 5 :  75-96 . 
- - ,  & N. SHIV.ANARAYAN. 1969 . Moult of the primary feathers in the 
House Swift , Apus affinis . Pavo 7 :  19-42 . 
NEUFELDT, I .  & A. I .  IVANOV. 1960 . Some notes on the biology of the 
Needle-tailed Swift in Siberia. Brit . Birds 53 : 431-433 . 
NOVICK, A. 1959 . Acoustic orientation in the cave swiftlet . Bio .  Bull . 
117 : 497-503. 
NUR, N. 1984a. The consequences of brood size for breeding Blue Tits I .  
Adult survival , weight change & the cost of reproduction. J .  Anim. 
Ecol . 53 : 479-496 . 
---. 1984b. The consequences of brood size for breeding Blue Tits II . 
Nestling weight , offspring survival & optimal brood size . J .  Anim. 
Ecol . 53 : 497-517 . 
OBERHOLSER, H .C .  1906 . A monograph of the genus Collocalia. Proc . Acad. 
Nat . Se . Phil . 53 : 177-212 . 
1 2 7  
O ' CONNOR, R. J. 1975 . The influence of brood size upon metabolic rate and 
body temperature in nestling Blue tits Parus caeruleus and House 
Sparrows Passer domesticus . J .  Zool . Lond; 175 : 391-403 . 
PARKER, S .  1967 . Some eggs from the New Hebrides , south-west Pacific . 
Bull . Brit .  Ornith. Club 87 : 90-92 . 
PELLANTOVA, J .  1975 . The course of breeding of the swift (Apus apus ) .  
Zool . Listy. 24: 249-262 .  
PALOMEQUE ,  J . , J .D .  RODRIGUEZ, J . D .  PALACIOS & J .LANAS. 1980 . Blood 
respiratory properties of swifts . Comp . Biochem. & Physiol . 67a: 
91-95 . 
PAYNE, R .B .  1969 . OVerlap of breeding & moulting schedules in a 
collection of African birds . Condor 71 : 140-145 . 
PECOTICH, L .  1974. Grey Swiftlets in the Tully River Gorge & Chillagoe 
Caves . Sunbird 5 :  16-21 . 
----- . 1982 . Speciation of the Grey Swiftlet Aerodramus spoctiopygius . 
Tower Karst 4 :  53-57 . 
PERRINS , C .M. 1964. Survival of young swifts in relation to brood size . 
Nature 201 : 1147-1148 . 
----- .  1965 . Population fluctuations & clutch-size in the Great Tit 
(Parus major) L .  J .  Anim. Ecol . 34: 601-647 . 
----- . 1977 . The role of predation in the evolution of clutch size . 
Chapter 16 In Stonehouse , B .  & C .M. Perrins , Evolutionary Ecology. 
London. Ma.cmillan.  
PIANKA, E.A. & W.S .  PARKER, 1975 . Age specific reproductive tactics . Am. 
Nat . 109 : 453-464. 
PITELKA, F .A. 1958 . Timing of molt in Steller Jays of the Queen 
Charlotte Islands , British Colombia. Condor 60 : 38-49 . 
PRIGOGINE, A. 1966 . Note on Chapin ' s  Swift . Bull . Brit . Ornith. Club. 
86 : 1-5 . 
RAND, A.L.  1936 . The distribution and habits of Madagascar birds . Bull . 
Amer. Mus. Nat . Hist . 72 : 143-499 .  
RATHBUN, S . F. 1925 . The Black Swift & it ' s  habits . Auk 42 : 497-516 . 
RICKLEFS, R .F .  1970 . Clutch size in birds : outcome of opposing predator 
& prey adaptations . Science 168 : 599-600. 
-- . 1980 . Geographic variation in clutch size among passerine birds : 
Ashmoles hypothesis . Auk 97 : 38-49 . 
ROBERTS , J .R. , D .M. SMYTH & A. V. SPAIN. 1976 . Preliminary studies on 
acoustic orientation in Collocalia spodiopygia. Search 7 :  97-98 . 
ROWLEY, J .S .  & R.T .  ORR , 1962 . The nesting of the White-naped Swift . 
Condor 64: 361-367 . 
SALOMONSEN, F .  1972 . Zoogeographical and ecological problems in Arctic 
birds . Proc. XV Int . Om. Congr . : 25-77 ? 
1 2 8  
SCANES, C .G . , T . J .  LAUTERIO , & F . C .  BUONOMO . 1983 . Annual , developmental 
and diurnal cycles of pituitary hormone secretion. Pages 307-326 In 
Avian endocrinology ( ed .  Mikami , S . , K. Horrrna. & M. Wada) . Tokyo. -
Japan Scientific Societies Press I Berlin . Springer-verlag . 
SCHAFFER, W.M.  1974. Optimal reproductive effort in fluctuating 
environments . Am. Nat . 108 : 783-790 . 
SCHMIDT, R .K. 1965 . Brutbiologie des weissburzelseglers (Apus caffer 
caffer) (Lichtenstein) auf der kaphalbinsel . J .  fur Ornith. 106 : 
295-306 . 
SCHONWETTER,  M.  1970. Handbuch der Oologie . Akademie-Verlag. Berlin. 
SERVENTY, D .L .  & H.M.  Whi ttell . 1962 . Birds of Western Australia.  Perth. 
Paterson Brokensha. 
SHAW, T .H .  1936 . Birds of Hopei Province ( 2  vols . )  Peking. Fan Memorial 
Institute of Biology. 
SHEA, R .E .  & R .E .  RICKLEFS , 1985 . An experimental test of the idea that 
food supply limits growth rate in a tropical pelagic seabird . 
American Naturalist 126 : 116-122 . 
SICK, H .  1948 . The nesting of Reinarda squamata (Cassin) . Auk 65:  
168-174. 
--- . 1959 . Notes on the biology of two Brazilian Swifts , Chaetura 
andrei & Chaetura cinereiventris .  Auk 76 : 471-477 . 
SKUTCH, A .F .  1949 . Do tropical birds rear as many young as they can 
nourish. Ibis 91 : 430-455 . 
- . 1967 . Adaptive limitation of the reproductive rate of birds . Ibis 
109 : 579-599 . 
--- . 1976 . Parent birds and their young. Texas . Univ. of Texas . 
SMEDLEY, D . I .  1977 . Sexing of Lewin ' s  Honeyeaters . Corella 1 :  34.  
SMYTH, D .M. , L .  PECOTICH, & J .R. ROBERTS . 1980 . Notes on the 
distribution & breeding of the Grey Swiftlet , (Aerodramus 
spodiopygius) . Sunbird 11 : 1-19 . 
SNOW, D.W.  1962 . Notes on the biology of some Trinidad Swifts . Zoologica 
47 : 129-139 . 
----- . 1978 . The nest as a factor determining clutch-size in tropical 
birds . J .  fur Ornith. 119 : 227-230 . 
SOMADIKARTA, S .  1968 . The Giant Swiftlet , ( Collocalia gigas) ,  Hartert & 
Burler . Auk 85 : 549-559 . 
STRESEMANN, E .  1928 . Sauropsida:Aves . in W.  KUKENTHAL (Ed. ) ,  Handbuch 
der Zoologie ,  7 .  Halfte , Lieferung 4 .  Berlin: Walter de Gruyter . 
--- , & V. STRESEMANN. 1966 . Die mauser der vogel . J .  fur Ornithologie .  
107 : 1-445 . 
1 2 9 
SUTHERS , R.A.  & D . H .  HECTOR . 1982 . Mechanisms for the production of 
echolocating clicks by the Grey Swiftlet , (Collocalia spoctiopygia) .  
J .  Comp . Physiol . ;  a sens Neural Behav . Physiol . 148 : 457-470 . 
--- JA. -- L<fSG": Procl .. r_c.:nof'l ?<;- sona.r- c.lic.k.s by Grey S....if+le'\s. lOwer l<o.r.s-t S": 4-"1-53. 
TARBURTON, M.K.  1986a. The food of the White-rumped Swiftlet ( Aerodramus 
spodiopygius ) in Fi ji . Notornis 33 : 1-16 . 
----- , 1986b. A comparison of the flight behaviour of the White-rumped 
Swiftlet and the Welcome Swallow. Bird Behaviour 6 :  72-84. 
----- , 1986c . Breeding of the White-rumped Swiftlet in Fi ji . Emu 86 : 
214-227 . 
--- - , 1987a. An experimental manipulation of clutch and brood size of 
White-rumped Swiftlets Aerodramus spodiopygius of Fi ji . Ibis 129 : 
107-114. 
----- , 1987b. The population status , longevity & mortality of the 
White-rumped Swiftlet in Fi ji . Corella 11 : 
----- , & E . O .  Minot 1987 . A novel strategy of reproduction in birds . 
Animal Behaviour in press . 
TAYLOR, J . S .  1942 . Notes on the martins , swallows & swifts :  
Graaff-Reinet . Ostrich 13 : 148-156 . 
(I 
TOVORNIK, D .  & V .  CERNY, 1974. F?ding of Haemaphysalis erinacei erinacei 
Pavesi 1884 on birds in Yugoslavia. Folia Parasitol (Prague) 21 : 
282 . 
VAN TYNE , J .  & A. J .  BERGER . 1976 . Fundamentals of Ornithology. New york . 
John Wiley .  
VUILLEUMIER, F .  & M.  GOCHFELD . 1976 . Notes sur L '  avifauna de 
Novelle-Caledonie .  Alauda 44:  237-273 . 
VINCENT , E .  1963 . Acoustic signals for auto-information or echolocation. 
In BUSNEL , 1963 . Acoustic behaviour of animals . 
WAGNER, H . O .  1957 . Variation in clutch size at different latitudes .  Auk 
74 243-250 . 
WALTERS , M. 1980 . The complete birds of the world . Sydney. Reed . 
WANG , C . C .  1921 . The composition of Chinese edible birds nests and the 
nature of their proteins . J .  Bio . Chem. 49 : 429-439 . 
WATLING , D .  1982 Birds of Fi ji , Tonga & Samoa. Wellington. Millwood . 
WEITNAUER , E .  1947 . Am neste des mauerseglers , (Apus apus apus ) ( L . ) .  
Ornith. Beobachter 44:  133-182 . 
----- , & D .  Lack. 1955 . Daten zur fortpflanzungsbiolog des mauerseglers 
(Apus apus) in Oltingen und Oxford . Ornith. Beobachter' 52 : 137-141 . 
WELLS, D . R .  1975 . The Moss Swiftlet ( Collocalia vanikorensis) Quoy & 
Gaimard in Sumatra. Ardea 63 : 148-151 . 
WELTY, J . C .  1975 . The life of birds . Philadelphia. Saunders . 
l 2 9 a 
WHITMEE, S . J .  1875 . List of Samoan birds , with notes on their habits & 
c .  Ibis 5 :  436-447 . 
WILLIAMS , G . C .  1966 . Natural selection , the costs of reproduction , and a 
refinement of Lack ' s  principle .  Am. Nat . 100 : 687-692 . 
WINKLER, D .W.  & J .R .  WALTERS . 1983 . The determination of clutch size in 
precocial birds . Current Ornithology 1 :  33-68 . 
WYNNE-EDWARDS ,  V . C .  1962 . Animal dispersion in relation to social 
behaviour . London. Oliver & Boyd . 
----- . 1963 . Intergroup selection in the evolution of social systems . 
Nature 200: 623-626 . 
YUREN, G .  & BENXIANG , z .  1985 . The breeding behaviour and population 
dynamics of the large White-rumped Swift ApUS pacificus pacificus 
(Latham) at Cheniushan Island in the Yellow Sea. Acta Zool . Sinica 
31 : 84-92 . 
ZAMMUTO , R.M.  & E . C .  FRANKS . 1979 . Trapping flocks of Chimney Swifts in 
Illinois . Bird Banding 50 : 201-209 . 
----- , & ----- , & R . P .  PRESTON. 1981 . Factors associated with the 
interval between feeding visits in brood-rearing Chimney Swifts 
( Chaetura pelagica) . J .  Field Ornith. 52 : 134-139 . 
1 3 0  
Appendix lA. 
CLIMATOLOGICAL SUMMARY TABLE FOR KORONIVIA,  FIJI 
AND OTHER BREEDING LOCATIONS 
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC YEAR 
RAINFALL (rrrn) 
1950-1979 367 300 399 359 239 183 171 154 204 221 305 296 3198 
1981 508 381 163 464 228 78 25 212 125 169 555 135 3043 
1983 267 321 305 135 121 85 101 289 86 187 212 330 2440 
HOURS / SUN 
1971-1979 171 161 149 143 135 127 124 145 128 157 149 167 1756 
1981 197 
1983 142 
1965-1979 26 . 5  26 . 5  26 . 3  25 . 4  24 . 3  23 . 8  22 . 6  22 . 8  23 . 1  24 . 0  24 . 8  25 . 7  24. 7  
?: . 
RAINFALL at TULLY ( rrrn ) 
Average 671 790 827 515 345 217 157 133 125 106 159 276 4321 
RAINFALL at CHILLAGOE ( rrrn) 
Average 217 216 147 30 14 10 4 4 6 14 55 138 855 
RAINFALL at NOUMEA (rrrn) 
Average 114 108 119 115 94 114 83 70 47 47 47 64 1020 
RAINFALL at POINDIMIE (rrrn ) 
Average 378 307 398 245 157 238 171 96 102 99 180 172 2542 
'? 
1 3 1  
Appendix 1B 
DAILY RAINFALL DATA FOR THE PERIOD OF THE 
MANIPULATION EXPERIMENTS 
KORONIVIA, FIJI . 
1981 Rainfall 1983 Rainfall 
(rrm) (rrm) 
November 25 74 . 2  
26 24 . 4  
27 1 . 8  
28 86 . 5  
29 2 . 8  
30 
December 1 0 . 4  90 . 8  
2 
3 1 . 8  12 . 7  
4 3 . 7  16 . 0  
5 4 . 0  
6 6 . 8  0 . 5  
7 13 . 2  14 . 5  
8 23 . 8  
9 
10 
11 1 . 2  
12 0 . 7  
13 12 . 5  0 . 3  
14 
15  1 . 3  
16 4 . 1  15 . 6  
17 4 . 9 3 . 8  
18 24. 4  
19  5 . 4  68 . 2  
20 7 . 5  15 . 7  
21  7 . 9  3 . 8  
22 9 . 9  9 . 5  
23 0 . 3  
24 3 . 5  0 . 5  
25 5 . 4  2 . 3  
26 2 . 0  0 . 5  
27 0 . 5  
28 11 . 3  
29 0 . 4  
30 20 . 4  
31 3 . 4  17 . 0  
Dec . Total 135 . 0  330 . 0  
1 3 2  
Appendix 1C . 
CLIMATOLCGICAL SUMMARY TABLE FOR CHILLAGOE , QLD. 
RAINFALL (mm) 
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC YEAR 
1902-1986 217 216 147 30 14 10 4 4 6 14 55 138 855 
S .D .  116 117 106 43 27 19 10 11 15 18 48 100 276 
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
1902 179 133 6 4 0 0 0 0 0 0 6 76 404 
1903 319 117 352 118 10 31 0 3 0 14 80 410 1453 
1904 206 276 180 84 0 0 0 0 14 62 134 160 1116 
1905 463 82 15 13 50 0 0 0 0 0 41 0 663 
1906 168 222 80 13 21 0 0 0 6 10 174 269 961 
1907 113 196 71 0 24 45 0 0 0 34 35 154 672 
1908 174 169 294 41 11 1 0 0 3 48 2 10 753 
1909 82 79 284 0 0 43 3 0 14 23 67 122 715 
1910 302 370 122 0 0 6 1 0 37 28 20 366 1252 
1911 117 253 129 279 0 0 0 0 0 20 35 112 945 
1912  124 51 154 1 3 24 0 0 1 25 9 90 482 
1913 402 253 252 105 31 7 0 0 0 3 1 229 1281 
1914  136 201 268 40 21 30 0 0 0 0 24 78 799 
1915 72 154 41 0 3 0 0 0 5 0 35 122 432 
1916 189 260 207 54 7 3 8 0 0 14 75 173 989 
1917 278 329 154 17 48 0 0 1 4 25 86 150 1091 
1 918 244 181 90 26 0 0 0 0 0 0 50 122 713 
1919 127 124 188 12 9 0 0 0 0 0 4 55 519 
1920 427 103 69 3 71 10 5 3 0 0 7 107 806 
1921 115 170 233 18 7 5 17 0 5 25 8 326 930 
1 922  181 414 84 4 1 2 8 0 0 10? 37 162 904 
1923 129 3 72 15 0 7 0 5 0 0 0 86 316 
1924 138 258 151 52 0 3 0 18 5 25 93 123 868 
1925 273 214 257 5 2 0 2 0 0 1 19 85 857 
1 926 171 19 41 7 14 2 0 0 12 0 3 201 470 
1927 184 345 145 16 0 64 35 0 0 0 8 110 906 
1 928 97 106 54 6 0 0 0 0 0 12 38 196 508 
1929 287 352 153 63 0 0 0 0 0 0 84 31 969 
1 930 465 149 48 29 136 0 0 0 0 41 70 46 983 
1931 242 110 9 15 0 0 0 0 1 1 38 223 639 
1932 312 94 52 0 9 0 0 0 0 0 0 394 861 
1933 163 188 61 0 0 14 19 42 14 0 139 149 789 
1 934 239 241 45 32 18 52 14 14 4 6 157 18 839 
1935 195 59 55 29 63 5 3 0 0 12 5 23 448 
1 936 207 245 270 55 0 28 1 0 3 49 27 241 1125 
1937 140 121 293 0 12 0 6 0 0 1 54 52 681 
1 938 248 215 20 0 0 1 61 0 0 17 29 27 618 
1939 240 239 161 39 2 60 0 1 0 19 115 110 985 
1940 176 392 114 17 4 3 0 0 0 0 98 20 824 
1941 263 161 279 14 19 6 1 0 0 0 51 51 844 
1 942 25 353 22  14 2 9 27 3 8 0 47 360 868 
1943 136 439 38 0 0 12 0 0 19 79 11 90 822 
1944 163 394 173 0 0 10 32 8 0 16 33 417 1245 
1945 100 241 299 73 0 0 0 0 0 9 56 78 857 
1946 407 196 81 14 0 0 2 0 0 0 48 145 894 
1947 138 239 124 0 0 5 0 16 50 0 55 163 789 
1948 122 101 101 5 2 3 4 0 0 1 44 58 440 
1949 182 334 278 24 0 0 0 0 2 48 80 76 1023 
1950 104 2 10 155 117 0 95 29 0 0 10 155 358 1234 
1 3 3  
CLIMATOLOGICAL SUMMARY TABLE FOR CHILLAGOE , QLD . cont . 
RAINFALL (mm) 
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC YEAR 
1951 415 251 188 0 0 0 1 0 0 15 50 41 960 
1952 85 82 17 56 0 23 0 0 0 51 66 112 492 
1953 153 261 27 0 0 0 0 26 1 0 12 105 587 
1954 291 456 74 164 3 4 0 0 0 33 25 26 1078 
1955 240 322 311 15 17 2 4 0 4 2 16 45 976 
1956 262 568 257 36 112 50 3 12 43 3 74 376 1796 
1957 290 176 162 44 0 1 0 0 0 32 63 86 853 
1958 334 306 116 45 2 0 0 0 0 3 12 68 886 
1959 299 115 154 20 18 15 1 0 0 0 41 291 953 
1960 144 169 33 1 60 1 0 0 0 9 75 90 581 
1961 110 109 117 6 3 0 0 1 1 3 52 92 492 
1962 251 264 27 27 25 1 1 0 9 3 141 127 875 
1963 214 374 79 17 0 1 0 0 0 0 16 87 787 
1964 200 127 156 3 11 0 3 0 0 57 178 197 932 
1965 137 49 160 24 0 25 0 0 2 0 4 90 491 
1966 233 123 18 5 0 0 0 0 0 4 71 1 455 
1967 81 251 371 14 2 93 0 0 0 3 28 158 1000 
1968 276 261 130 3 14 0 10 0 9 9 37 146 894 
1969 148 78 105 4 3 14 3 0 0 9 10 171 544 
1970 134 90 66 8 0 0 0 82 0 14 102 219 715 
1971 206 416 165 41 0 24 0 22 4 15 53 30 974 
1972 232 322 389 15 11 8 0 0 1 0 28 44 1050 
1973 91 409 158 34 7 2 0 20 37 12 238 177 1185 
1974 667 380 425 1 5 0 0 18 1 0 29 153 1678 
1975 262 114 214 17 0 6 0 5 102 1 148 180 1110 
1976 207 238 426 26 2 1 0 0 0 31 106 188 1126 
1977 66 353 270 12 lOO 0 0 0 0 0 74 163 1037 
1978 77 126 19 80 1 0 30 1 0 0 47 153 534 
1979 472 261 327 36 0 8 5 0 0 15 7 45 1176 
1980 294 185 124 8 6 4 0 0 0 30 54 101 808 
1981 459 355 46 27 79 3 0 0 3 7 52 179 1212 
1982 240 40 130 34 3 3 0 1 2 0 39 131 617 
1983 77 65 143 94 87 15 0 8 27 37 89 73 716 
1984 385 217 72 8 3 0 0 0 5 19 20 100 828 
1985 134 114 173 117 12 0 0 0 2 29 124 219 923 
1986 320 179 58 40 2 0 0 12 35 48 11 48 753 
1 3 4  
Appendix 1D 
DAILY RAINFALL 1985 
CHILLAGOE ?mm? 
DATE JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC 
1 8 . 0  
2 1 . 8  36 . 8  
3 7 . 4  
4 3 . 8  3 . 0  
5 4 . 0  14 . 2  28 . 0  
6 2 . 0  19 . 0  * 2 . 0  
7 16 . 6  
8 6 . 0  2 . 6  
9 1 . 0  2 . 0  
10 129 . 0  
11 1 . 6  11 . 2  
12  3 . 8  
13 10 . 0  
14 18 . 4  2 . 4  
15 23 . 6  
16 5 . 6  
17 64. 0  1 . 6  
18 10 . 0  
19 2 . 4  8 . 8  4 . 6  
20 
2 1  4 . 0  3 . 0  4 . 0  2 . 6  
22  108 . 4  5 . 4  30 . 8  
23 7 . 6 26 . 0  3 . 8  
24 1 . 6  33 . 0  2 . 8  
2 5  4 . 6  30 . 0  0 . 8  9 . 0  
26 74 . 8  0 . 1  21 . 0  4 . 0  
27 5 . 6  43 . 0  1 . 6  
28 6 . 0 
29 1 . 6  24 . 0  
30 2 . 0  
31 0 . 6  
TOTAL 134 114 173 117 12 NIL NIL NIL 2 29 124 219 
NOTE: * represents estimated date for the laying of the first egg . 
1 3 5  
Appendix 1D cont . 
DAILY RAINFALL 1986/7 
CHILLAGOE ?mm) 
DATE JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC JAN 
1 
2 6 . 0  30. 0  
3 33 . 6  4 . 8  
4 6 . 0  4 . 0  * 1 . 6  
5 1 . 0  6 . 2  2 . 8  
6 
7 99 . 0  7 . 0  
8 
9 2 . 0  
10 4 . 4  1 . 2  3 . 0  
11  25 . 0  
12 2 .0 
13 
14 17 . 0  2 . 4  
15 1 . 6  58 . 0  
16 1 . 0  8 . 0  36 . 0  
17 1 . 8 
18 9 . 4  2 . 2  14 . 0  
19  1 . 8  35 . 0  
20 26 . 6  1 . 0  4 . 0  
21 19 . 0  
22 51 . 0  
23 27 . 0  10. 6  2 . 0  
24 25 . 0  2 . 8  43 . 6  0 . 8  
25 33 . 4  19 . 0  
26 3 . 2  8 . 2  10 . 0  
27 5 . 6  19 . 8  1 . 2  8 . 6  
28 1 . 8  1 . 2 6 . 4  32 . 8  
29 7 . 2  18 . 8  1 . 6  
30 6 . 0  "31,.?6 35 . 0  
31 7 . 6  24 . 2  
TOTAL 320 179 58 40 2 NIL NIL 12 35 48 11 48 250 
NOTE : * Represents the estimated date for the laying of the first egg .  
1 3 6  
Appendix lE . 
MAXIMUM TEMPERATURES AT CHILLAGOE 1985/86 ( OC ) 
1985 1986 
DATE JUN JUL AUG SEP OCT NOV DEC JUN JUL AUG SEP OCT NOV DEC 
1 30 27 32 36 32 35 29 31 30 34 37 37 
2 31 28 29 35 30 35 31 32 32 30 37 37 40 
3 27 27 33 32 36 29 29 31 33 30 36 36 37 
4 30 27 28 33 32 35 31 32 33 30 33 35 36 36 
5 30 28 28 29 32 35 34 30 34 29 34 38 36 36 
6 29 29 30 30 ( 32 )  33 33 30 32 29 34 39 35 38 
7 28 29 30 30 32 34 33 30 34 29 35 (39 )  37 38 
8 26 29 30 27 32 36 33 31 33 29 39 35 39 
9 27 25 31 30 37 38 38 32 34 30 34 38 38 37 
10 30 26 31 24 35 37 38 30 32 33 33 37 37 39 
11 30 28 32 33 35 3 29 32 32 33 34 37 39 
12 29 27 33 33 31 40 37 27 30 33 34 37 36 39 
13 30 27 34 32 37 38 30 31 32 36 36 37 40 
14 28 26 32 35 37 34 38 34 33 31 36 38 41 41 
15 30 22 30 35 38 35 39 34 33 32 40 40 40 45 
16 29 29 30 35 37 37 39 35 36 33 39 40 36 41 
17 30 28 30 34 38 36 37 34 34 30 37 40 40 39 
18 29 28 33 35 38 38 39 33 35 29 36 39 36 39 
19 29 27 33 36 39 40 38 33 31 34 37 37 37 39 
20 29 26 32 30 39 40 39 32 32 34 38 33 36 38 
21 29 26 32 35 39 39 39 31 32 33 39 37 36 40 
22 28 27 29 34 38 38 39 30 32 33 39 39 37 37 
23 26 32 33 35 37 39 37 29 30 34 40 36 38 40 
24 27 30 28 35 36 36 39 30 33 34 37 31 38 
25 28 29 35 34 36 36 30 32 33 39 37 37 39 
26 28 27 34 35 36 34 39 30 33 34 39 39 38 39 
27 28 27 35 38 34 33 37 33 34 34 37 37 39 40 
28 26 28 33 38 35 37 32 29 35 36 37 40 40 
29 26 28 34 35 35 35 38 36 26 36 35 37 33 42 
30 26 28 35 34 35 36 37 33 31 34 37 37 40 41 
31 29 35 36 36 29 30 37 
NOTE: Dates in parentheses are the estimated days on which the first eggs 
were laid in the two caves that were visited daily .  
APPEND IX  2 
NOTORNIS 
is the journal of the Ornithological Society o f  New Zealand ( Inc . )  
Editor: B .  D .  Heather, 
10 Jocelyn Crescent 
SIL VERSTREAM 
Volume 33 Part 1 March, 1986 
THE FOOD OF THE WHITE-RUMPED 
SWIFTLET (Aerodramus spodiopygius) IN FIJI 
By M. K. TARllURTON 
ABSTRACT 
Diptera (flies), Homoptera (planthoppers), Hymenoptera (social 
insects), Isoptera (termites), and Coleoptera (beetles) were the most 
numerous prey in 32 food boluses being delivered by parent White?
rumped Swiftlets (Aerodramus spodiopygius) to their chicks inside two 
Fijian caves. Numerically the main food items were flies (37%) and 
planthoppers (33%). Both the season and the habitat over which the 
birds had been feeding seemed to determine whether flies or 
planthoppers predominated in a particular bolus sample. Flies 
predominated in the prey of swiftlets foraging over open country, 
whereas planthoppers predominated in the prey of swiftlets foraging 
over both forest and open country. 
The number of insects in each food bolus ranged from 47 to 
750 (x = 236). The average weight of a bolus was 0.225 g (range 0. 1-
0.43 g) .  The average length of  all prey was 2.48 mm, which is larger 
than the average length of available prey ( 1 .63 mm). The number of 
prey species ranged from 2 to 83 (X = 30 per bolus). Altogether, 167 
species were recorded in food boluses. The White-rumped Swiftlet bred 
during the wet season, when insects were more abundant. 
This study, along with others (largely unpublished), shows for 
the first time that flies are often the most common insect in the prey 
of swifts, swiftlets and swallows. 
INTRODUCTION 
Swifts have been shown to collect more food on fine days than on 
wet days, although the reasons differ with latitude. Lack ( 1956) found that, 
in temperate latitudes, nestling Common Swifts grew more in wing length 
and weight on sunny warm days than on dull, cold, wet days. He also found 
that the food boluses fed to chicks contained larger insects on warm days 
than on wet days. Aerial tow netting showed that flying insects were in greater 
densities on warmer days and so the swifts could select larger prey (Lack 
& Owen 1955). 
NOTORNIS 33: 1? (1986) 
1 3 7  
2 TARBURTON NOTORNIS 33 
In the tropics, however, Hespenheide ( 1975) found from tow net 
sampling that flying insects were at higher densities in wet weather. Despite 
this, he found that swifts and swallows 
I. Took the same average size of insects on both wet and dry days; 
2. Caught a greater size range on wet days, probably because the rain reduced 
their foraging time, forcing them to be less selective; 
3. Showed a preference for swarms, when present; and 
4. Did not favour flies, presumably because flies manoeuvre better than other 
insects. 
The preference for swarms applied particulary to the larger swifts. 
From these findings, Hespenheide suggested that flies are scarce in 
the diet of all aerially feeding insectivores because they are harder to catch, 
being more manoeuvrable than other insects. He also proposed that certain 
behaviour, characteristic of each insect order, caused the average size of prey 
taken from each insect order to be significantly different. 
This paper has two purposes. The first is to show the number, size 
and identity of the White-rumped Swiftlet's prey in Fiji. The second is to 
determine whether Hespenheide's findings apply to this swiftlet, which is widely 
distributed in the tropical south-west Pacific, or to other aerial feeders such 
as the swifts and swallows, as reported in other studies. 
METHODS 
In December 1981 and 1983, I studied the food of swiftlets nesting 
at Nasinu Nine-mile, 9 miles north of Suva. Of the two nesting colonies in 
separate caves at Nasinu Nine-mile, I chose that in the larger Waterfall Cave, 
where my longevity studies that had run since 1974 had shown that the birds 
are disturbed less by the public than those breeding in the smaller colony 
in Dry Cave. 
Birds were captured as they carried their prey to their chicks, mostly 
in nests built in totally dark sections of the cave. I caught the birds in a butterfly 
net before they reached their nests because Lack ( 1956) and Fischer ( 1958) 
had found with the Common Swift (A pus apus) and the Chimney Swift (Ciuietura 
pelagica) that disturbing birds at their nests made some desert. 
Whenever a bird had its throat distended with a food bolus, I gently 
prised open its mandibles using my thumbnail and pencil and, holding the 
bird upside down, rolled the food bolus out with the pencil. 
I collected the food boluses in the wetter of Fiji's two seasons, the 
season shown by other studies to have more abundant insects. I weighed each 
food bolus and then preserved it in formaldehyde. In the laboratory, I sorted 
the prey into orders and into unnamed but distinctive groups, presumably 
species, and then ?ounted and measured them. 
I sampled potential prey by the methods of Hespenheide ( 1975). The 
two areas sampled were the 4.3 km along Wainibuku Road from the Suva?
N ausori road to near the entrances of Dry and Waterfall Caves, and in Tamavua, 
10 km from the cave. The first area consisted of small horticultural farms, 
together with some young scrub regrowth and occasional trees. Farm crops 
were mainly pineapples, taro and cassava among scattered coconut trees. The 
1 3 8  
1986 WHITE-RUMPED SWIFTLET 3 
Tamavua area was a well-vegetated well-spaced residential area with food crops, 
flowering shrubs, trees and lawns. Swiftlets were feeding down to 0.5 m in 
both areas and at times were feeding while I was collecting samples in the 
tow net. 
RESULTS 
Identity of prey 
Flies were found in all food boluses but one and were the most numerous 
prey in 16 of the 32 boluses taken in December (Table 1) .  Flies made up 
43% of the total sample of 7433 invertebrates. Planthoppers were in all 32 
food boluses and were the most numerous in seven of them. Planthoppers 
made up 24% of the total sample. 
TABLE 1 - Composition of White-rumped Swiftlet prey in 32 food boluses. 
1981 & 1983 combined 
No. % % Range X %  of No. of No . in % Indi.vi<l un l s  
Where Where In All Samples Where Boluses Total In  Total 
Order Dominant Dominant Boluses Presen t ? SS Present in Snmple Sample 
Diptera 1 4  3 7  - 88 0 - 88 37 + 4. 8 3 1  3 1 76 43 
Homoptera 7 37 - 1 00 I - 1 00 33 +" 6. 2  32 1 748 24 
Hymenoptera 2 62 - 83 0 - 83 1 8 :; 5. 5 30 16 15  22  
Coleoptera I 53 0 - 53 9 :; 2 . 1  28 484 7 
Isoptera I 45 0 - 45  1 5  :; 4 . 1  1 2  168 2 
Heteroptera 0 0 0 - 2 < 1 .0 7 9 
Trichoptera 0 0 0 - 3 2 + 1 . 0  3 7 
Thysanoptera 0 0 0 - 3 1 :; 1 . 0 8 23 
Megaloptera 0 0 0 - 1 1 .0 1 3 
Lepidoptera 0 0 0 - <I < l . O  2 4 -
Psocoptera 0 0 0 - 2 < 1 . 0  4 6 
Ephemeroptera 0 0 0 - 1 1 . 0 1 1 
Neuroptera 0 0 0 - <l  < l .O 1 1 
Unidentified 0 0 0 - I < 1 .0 7 1 2  
Aranae 0 0 0 - 8 2 .?. 0 . 5  1 7  S2 
Social insects were in 30 of the boluses but were the most numerous 
iri only two boluses. They made up 22% of the total sample. Termites and 
beetles were the most numerous in one bolus each, but beetles were present 
more often than termites. Although termites occurred in only 12 of the 32 
boluses, they sometimes did so in reasonable numbers ( 17-43 or 9%-45% of 
total insects in the bolus). They are available to swiftlets only while swarming, 
when they are the preferred food. Spiders, although very small, were found 
in 17 of the 32 boluses. 
The 1983 samples, which were colle"cted on two days, had a very different 
composition. The averages for the six boluses taken on 1 1  December were 
B4% planthoppers and only 3% flies (one bolus containing 100% planthoppers). 
However, in only two of the six boluses collected on 5 December were 
planthoppers predominant (an average of 59%). Thus the diet of swiftlets cannot 
be apequately assessed by means of brief and intermittent sampling. 
Size of prey 
? 
The largest prey found in this study were two adult moths 1 1  mm 
long. Two moth larvae 4.5 and 9 mm long were also well above average prey 
size. Termites were the largest of the common prey, averaging 4.5 mm, then 
planthoppers (2.5 mm), social insects (2.3 mm), flies (2.2 mm) and beetles 
1 3 9  
' / '  
4 TARBURTON NOTORNIS 33 
( 1 .9 mm). The average size of the prey was 2.48 ? 0 . 1 1  mm (x ? SE), which 
is significantly greater (t=6.4, p<0.0 1 ,  df=39) than that of the prey available 
( 1 .63 ? 0. 12). The data for total prey was based on the means of all 32 boluses 
rather than that of each type so that the extreme means of the uncommon 
types did not swamp those of the majority. The average size of the flies, social 
insects and beetles was each significantly larger than that available (t=3.2-
3.5, p<0.01 ,  df=27-38). 
The average size of the smallest group of insects (beetles) commonly 
found in the prey was not significantly smaller than that of the flies (t= l .63, 
p>O . l ,  df=54). The flies were not significantly smaller than the Hymenoptera 
(t=0 . 12, p>0. 1 ,  df=57), which however were significantly smaller than the 
termites (t=9.5, p<0.001,  df=40). 
The average size of each major insect order found in the boluses, 
whenever it was predominant in a bolus, was compared with the size of the 
same order from boluses when it was in the minority. The size of insects 
from a swarm (arbitrarily decided by Hespenheide to be when more than 
20 of a species occur in a bolus) was compared with the size of the same 
insect order when found in fewer numbers. None of the comparisons were 
shown to be significant, except that of beetles. In the one bolus where beetles 
were dominant (54%), their average size of 5.7 mm ? 0.2 was significantly 
greater than the average of all others ( 1 .7 mm ? 0.09). 
A significant difference in size (p<O.OO l )  was found between three of 
the four major insect orders when the two samples, each of six boluses and 
each taken in December 1983, were compared. These are shown in Table 2. 
These two groups of samples had three important differences. Those taken 
on the 5th were collected earlier ( 1300-1555 hours) than those taken on the 
1 1 th 1900-1918 hours). The 5th was largely an overcast day, but the 1 1th was 
the fourth consecutive sunny day. Both these differences may be expected 
to cause those collected on the 1 1th to be larger (Lack 1956, Hespenheide 
1975). In addition, the boluses on the 1 1 th were taken one hour after sunset, 
when the swiftlets were probably catching dusk-flying insects, which have 
been shown to be larger than those flying during the day (Lewis & Taylor 
1967, Hespenheide 1975). So then, both prey size and prey type show daily 
changes. 
The range of2 1  White-rumped Swiftlet boluses was 0 . 1 -0.43 g, averaging 
0.23 g ? 0.02. A significant correlation was found between the number of 
insects in a bolus and the weight of a bolus (Spearman rank correlation rs 
= 0.66, p<0.002, n=21) .  This, together with a negative correlation (rs = -0.84, 
p<0.001 ,  n=21 )  between the number and size of the insects in a bolus, indicates 
that a bird returns to feed its chicks when it has all it can hold. 
TABLE 2 - Average size of common prey 
( 1 9 8 3  sample in mm) 
5 December 
Coleoptera 1 .  45 
Hymenoptera 1. 70 
Homoptera 2 . 15 
Diptera 2 . 16 
0.084 
0. 1 10 
0 . 180 
0. 301 
1 1  December 
2 . 35 
3 .06 
2 . 86 
2 .8 1  
0. 236 
0.425 
0.081 
0. 482 
p<0.001 
p<O.OOJ 
p<0.001 
N .  S. 
1 4 0  
1986 WHITE-RUMPED SWIFTLET 5 
The almost spherical food boluses were about 6-7 mm in diameter. Some 
boluses were firm but others fell apart easily, making them hard to measure. 
The number of insects in a bolus varied from 47 to 750. The average 
number for all 32 boluses was 236 ? 32. The 198 1 sample averaged 269 ? 
44 (n=20) and the 1983 sample averaged 178 j_ 36 (n? 1 2). 
Combining the dai:a for December 198 1 and December 1983, as shown 
above, hides certain information. Whereas flies were dominant in most of the 
combined sample of food boluses, they were exceeded by planthoppers in seven 
of the 1 1  boluses from the 1983 sample. Further analysis of the numbers of 
individuals and species in the major orders is shown in Table 3. 
TABLE 3 - Frequency of major prey in food boluses (x ? SE) 
Individuals 1 98 1 1 Species 4 1981 3 [ndi  vidua 1 s 19832 Species 19832 
Diptera 1 23 . 9  + 3 1 . 0  1 2 . 4 + 1 . 7  58. 4  + 25.1? 
Hymenoptera 7 1 . 2  + 1 5 . 9  7 . 3  + 1 .8 1 6 . 3  + 5 .6  
Homoptera 40. 2 + 1 2 . '? 1 .9  + 0.4 R4 . <J  + 1 0 . 1, 
Coleoptera 2 1 . 8  + 4 . 2  5 . 6  + 1 . 2  :1 . H  + J . O  
Isoptera 7 . 9  ? 1 . 8  0 . 3  ? 0. 1 ! . 2 ? 0. 6  
Total 269. 2 ? 44 32 .6  ? s. 2 1 70.0 ? 33 . 7  
NarE: Numbers in 20 boluses taken 2-24 December 
Numbers in 1 2  boluses taken 5 , 1 1  December 
Numbers in 1 2  boluses taken 2-24 December 
1 1 . 4  + 3 . 1  
7 . 2  + 1 . 0  !, . C) + 1 . 0  
2 .8  + 1 . 0 
0 .3  ? 0 . 1  
29 . 25 ? 7 . 4  
4 'Species' is  not a named species but i s  ascribed t o  individuals 
that are morphologically similar 
The decrease in total insects per bolus between the years was not 
significant (t=l .79, p>O . l ,  df=30, two-tailed). Neither was there a significant 
change in the number of species within each major order or the total number 
of species per bolus between the years. This uniformity suggests that further 
comparative analysis would be valid. Such analysis shows that the decrease 
in the number of individuals per order in a bolus between 1981 and 1983 
was significant (t=3.09-3.53, p<0.01 ,  df=30) in the social insects, beetles and 
termites. This decrease was offset by a significant increase in planthoppers 
(t=2.23, p<0.05, df=30). The number of flies did not decrease significantly 
(t=l .63, p>0. 1 ,  df=30). 
The number of species found in a bolus varied from 2 to 83 and averaged 
29 in 1983 and 33 in 198 1 .  
DISCUSSION 
Prey size compared with that of other swiftlets 
Prey size has been positively related to the body size of insectivorous 
birds (Hespenheide 197 1 ,  1975; Dyrcz 1979). The Whii:e-rumped Swiftlet, with 
its light weight and small prey, fits into the general trend. It takes the smallest 
prey of any apodid so far studied (Table 4). 
Table 5 shows that the White-rumped Swiftlet is typical of all aerial 
feeding birds studied to date (Hespenheide 1975, Waugh 1979) in that it takes 
larger prey than the average of that available. 
1 4 1  
6 TARBURTON NOTORNIS 33 
TABLE 4 - Prey size of various Apodidae and Hirundinidae 
Predator 
White-rumped Swiftlet 
Aerodramus seodiOEI&ius 
Glossy Swiftlet 
Collocalia esculents 
Mossy-nest Swiftlet 
Aerodramus vanikorensis 
Black-nest Swiftlet 
Aerodramus maximus 
Barn Swallow 
Hirundo rus tica 
Horus Swift 
Apus horus 
Short-tailed Swift 
Chaetura brachyura 
Chimney Swift 
Chaetura eelagica 
Common Swift 
Apus apus Fine 
Wet 
Pacific Swallow 
Hirundo tahitica 
House Swift 
Apus affinis 
Chestnut-collared Swift  
CU8t!loide6 rutilUM 
Black Swift 
Cu!:seloides niger 
Grey-breasted Martin 
Progne chalybea 
Mangrove Swallow 
Tachycineta albilinea 
TABLE 5 
Diptera 
Homoptera 
Hymenopera 
Isoptera 
Coleoptera 
Trichoptera 
-
Thysanoptera 
Megaloptera 
Lepidoptera 
Psocoptera 
Neuroptera 
Ephemeroptera 
Heteroptera 
Unidentified 
Aranidae 
Total 
X Size S E Range Mode Source 
(mm) 
--- - -?-- - - - ---- - - - + - - - -- - ---
2 . 48 0. 1 1  0 . 3- 1 1  This paper 
2 . 6  Waugh & Hails 1983 
3 .05 c . l . 5- 1 2 . 5  Harrisson 1976 
3 .05 c . l . 5- 1 2 . 5  Harrisson 1976 
3.  3 Waugh & Hails 1983 
3 .  7 1  0 .08 0.8-9.0 2 . 6-3.0 Collins 1980 
4 . 0  0 .07 1-9 Collins 1968? 
c . 4 . 0  <5 Fisc her 1958 
3 .  5 <2- Lack & Owen 1955 
6 .  ') - > 1 0  l.nck & Owi'n J C)')) 
4 . 8  Waugh & llai ls  1983 
5.09 Waugh & llni t:-; 1 983 
(,.IJ n . :' r, _ I O  7 - B  Co l i I nN I 'JMHl .. 
8.68 2- I 2 <)- 1 0  d- !..Cl""'-Y Coi l  ins 1 968 
.... 
10 .5  Dyrcz 1984 
1 5 . 7  Dyrcz 1984 
Size of prey of White-rumped Swiftlet 
(total sample) 
Actual prey Potential prey 
SE SE 
2. 2 1  0. 1 1  3 1  1 . 64 0 . 14 
2 .47  0 . 10  32 
2.  38 0. 1 5  29 1 . 80 0. 10  
4 . 50 0 . 1 9  1 1  
1 .88 0. 1 7  26 1 . 26 0 .05 
3 .06 0.65 5 3 . 3  
1 . 49 0 . 1 4  1 1  1 . 38 0.03 
2 .67  1 
9 .00 2 3 . 4  
3 . 4 5  0 . 6 1  4 3 . 3  
! . 50 1 
3 .00 1 
2 . 47 0. 29 6 2 .0  0 .26 
2 .  28 0 . 34 5 
1 . 67 0. 1 2  1 7  4 .  5 
2 . 67 0. 1 1  32 1 . 63 0. 1 2  
1 4 2  
1986 WHITE-RUMPED SWIFTLET 7 
Hespenheide ( 1975) expected that the average size of each insect order 
in a swift's prey would be significantly different from that of the other orders. 
He derived this by assuming that the different orders of insects have different 
average flight abilities and that the birds spend about the same amount of 
energy in capturing any given prey item. Hespenheide ( 1975) found some 
evidence for these expectations in the prey of other swifts. However, this study 
shows evidence to the contrary in that swarming insects can negate both of 
Hespenheide's assumptions. An insect is seldom using or likely to use its full 
flight capabilities (in terms of high speed and manoeuvrability) while swarming, 
and an aerial predator will expend less energy in procuring a bolus of any 
high-density collection of insects. 
The food bolus 
Since Bartels ( 193 1 )  demonstrated that the Alpine Swift fed its chicks 
infrequently with large boluses of food, such feeding behaviour has been shown 
for other Apodidae. The wet weight of the White-rumped Swiftlets' food boluses 
varied about as much (0. 1 -0.43 g) as those of the Common Swift ( <O. 7-2.5 g, 
Lack & Owen 1955), although less than those of the Edible-nest Swiftlet 
(Aerodramus fuciphaga) (0. 13-1 .08 g, Langham 1980) and the Chimney Swift 
(Chaetura pelagica) (0.2-0.9 g, Fischer 1958). 
The average number of insects in a bolus (236) is much larger than 
the 94 average of 10 boluses from the same species in Queensland (Smyth 
1980). From this one could predict (assuming that the above correlations between 
size and number of insects in a bolus hold) that the Queensland subspecies 
takes larger prey than the Fijian subspecies does. This is expected (Bergmann's 
rule) as the Queensland subspecies A. s. terraereginae is much larger ( 12.2g) 
than Fijian birds (8. 1 g). In the Edible-nest Swiftlet, which is similar in size, 
the prey numbered 100 to over 1200, with an average of more than 500 per 
bolus (Langham 1980). The much larger Common Swift usually has 300-1000 
prey in a bolus, but the recorded range is 58- 1500 (Lack & Owen 1955). 
The number of species in a bolus varied from 2 to 83 and averaged 
29 in 1983 and 33 in 198 1 .  This is about half the number of species found 
in similarly sized samples from the stomachs of Short-tailed Swifts (Hespenheide 
1975), perhaps because fewer species are available in Fiji than in Panama and 
Costa Rica, as one would expect by Fiji's small area and isolation. However?, 
the average number of species taken by the White-rumped Swiftlets is lower 
than might be because 2 1  of the 24 birds apparently fed at swarms (as defined 
by Hespenheide 1975). The highest number of species in a bolus is only nine 
less than the highest in the Short-tailed Swift. One swiftlet had fed at six 
swarms and another at only two swarms, neither taking any other species. 
Five of the birds fed on fewer than 10 species to produce a bolus - a 
characteristic proposed for the larger swifts (Hespenheide 1975). The 24 boluses 
contained 167 species, of which 67 were flies, 44 social insects, 23 beetles, 
18 planthoppers, 1 1  spiders, 5 each of sap-suckers (Heteroptera) and thrips 
? (Thysanoptera), 2 book lice (Psocoptera) and 1 termite. An additional 29 species 
were taken in the tow net. 
The above results show that only in one bolus, dominated by beetles, 
was the average size significantly different from the average for insects of 
the same order in all other boluses. In this case the beetles in? the beetle-
1 4 3 
8 TARBURTON NOTORNIS 33 
dominated bolus were larger than in all other boluses. This is the reverse 
of that expected if a bird feeding on a swarm is less selective, as Hespenheide 
( 1975) proposed. As only two of the 50 beetles in the bolus were below the 
mean size of beetles in all other boluses, this bolus seems to have resulted 
from nothing more than the chance location of a swarm of larger than average 
beetles. 
Taxonomic comparison between available prey and captured prey 
For the most valid comparison between potential prey as sampled by 
the tow net and actual prey from the food boluses, both samples should be 
collected in the same season. Although this means ignoring the mass of data 
from 1981 ,  I have chosen to do so because several of the 1983 net samples 
were taken at the same time as the swiftlets were capturing the insects in 
the food boluses. On several occasions swiftlets were foraging in the same 
air space and at the same time as the net samples were being taken. The 
resulting data are shown in Table 6. 
TABLE 6 - Taxonomic proportions of prey compared with aerial 
invertebrates 
X %  in X %  in X %  in X %  in liomoptera Diptera 
Tow Net Food Boluses Dominated Dominated 
Order (Dec 83) ( Dec 83) Boluses Boluses 
Diptera 66.9 + 4 , 5  25 . 7 + 8 . 3* 8 .0  + ) . (J* (> 1 . 0  + 8. 5 
Homoptera 3 .8  + 0 . 6  58.0 + 9 .8* 77 .8  + 7 . 0* 18. 5 + 6.4+ 
Hymenoptera 1 1 . 1  + 3 . 5  9 .9  + 2 . 9  8 .9  + 4 . 3  1 1 . 7 + 2 . 1 
Coleoptera 1 5 . 9  ? 4 . 7  2 . 5  ? 1 . 0+ 0 . 7  ? 0 . 3* 6 .2  :?: l .  7+ 
Isoptera 0 1 . 6  2 .  5 0 
Trichoptera 0. 1 0 . 2  
Thysanoptera 1 . 2  0. 2 
Lepidoptera o. 7 0 
Psocoptera 0. 1 0 . 3  
Heteroptera 1 . 1  0. 2 
Ephemeroptera 0 . 2  0 
Unidentified 1 . 3  0. 2 
Araneae 0 . 1  1 . 3  
NOTE: * Shows significant difference to tow net samples ( p  <0 .001 ) .  
+ Shows significant difference to tow net samples ( p? <0.05) . 
Because two planthopper species (both Delphacidae) formed a clear 
majority in 8 ot the 12 boluses and only one of these species was rarely taken 
in the net, the birds with an abundance of planthoppers had apparently spent 
much of their foraging time in some other habitat than that sampled. Further 
confirmation of this is given by the significant difference between the percentage 
of flies in the bo1uses having mostly planthoppers and the percentage of flies 
in the tow net samples (t=4.4, p=<O.O l ,  df=lO) and no significant difference 
between the percentage of flies, social insects or beetles in boluses dominated 
by flies and the percentage of them in the tow net samples. Taken together, 
these data suggest that the birds with predominantly flies in their food boluses 
had been feeding in the open habitats that I had sampled with the net, whereas 
those with predominantly planthoppers had been feeding over the forests (which 
I did not sa:mple with the net) to the west of the caves. 
1 4 4  
1986 . WHITE-RUMPED SWIFTLET 9 
Of the fly species in the net samples, a similar proportion was found 
in the fly-dominant boluses (44%) and the planthopper-dominant boluses (47%). 
This similarity may mean that the swiftlets feeding on planthoppers foraged 
over the fly-rich open habitats as well as over the planthopper-rich forests. 
This is confirmed in that the planthopper-dominant boluses contained a larger 
percentage (43%) of fly species not found in the fly-dominant boluses than 
the small percentage (24%) of f1y species found only in the f1y-dominant boluses. 
This conclusion is consistent with my observation that the swiftlcts periodically 
feed in the open habitat on their way to the forest. It is also consistent with 
the finding that a greater number of insect species fly over forest, which has 
a greater diversity of plants than open habitat (Hespenheide 1975, Waugh 
& Hails 1983). 
It is interesting that the average percentages of the three most common 
insect orders taken in the net are each very close to those taken in Costa 
Rica and Panama with a similar net by Hespenheide ( 1975). The largest 
deviations from any of his results (which varied by season and location) are 
flies 8.2%, social insects 8. ')<y.', and beetles 9.4'!\, . The main interest in this 
comparison arises from two phenomena. The first is that it would seem unusual 
for oceanic islands such as Fiji to have a similar proportion of flying insect 
groups to a region that is attached to two large land masses. The second is 
that, whereas the two swifts and the swallow studied in Central America did 
not make proportionate use of f1ies, the most common insect order, the White?
rumped Swiftlet, did in Fiji. 
The most common group of flying insects available to Fijian swiftlets 
was the flies. Hespenheide suggested that flies arc more manoeuvrable than 
most insects and that this helps explain their infrequent occurrence in the 
prey of large swifts in particular and in aerial predators in general. 
He cited studies of six species of large swifts that took a small range 
of prey species with flies not a major component. He reasoned that, because 
the larger swifts have greater foraging ranges than smaller swifts, they may 
specialise on insects in mating or dispersal swarms. However, there are two 
problems with this argument. The first is that some studies (seven of?which 
have not been previously published) have shown that t1ies can be the 
? predominant prey of large swifts. Thblc 7 shows that f1ies have dominated 
in the studied diets of eight species, three of which were large swifts. By 
comparison, the social insects were found to be dominant in the prey of 1 1  
species, planthoppers dominant in the prey of three species and beetles dominant 
in the prey of two species. 
The second problem is that, if flies were more difficult to capture and 
the difficulty increased with the size of the swift, as proposed by Hespenheide, 
there should be a good negative correlation between the weight of the swift 
and the percentage of flies in its diet. There is however only a low negative 
correlation between the predator's weight and the proportion of flies in the 
prey for the 37 studies in Tables 7 and 8 that provide numerical data as 
percentages (rs= 0.28, 0. 10>p<0.05). It would appear that, regardless of .the 
size of the predator, swifts, treeswifts or swallows do not show any preference 
for or against flies. The birds presumably take what is available, giving preference 
to swarms or other high-density concentrations, which are just as likely to 
1 4 5  
TABLE 7 - Percentage composition (numerically) of major prey of swifts 
Philippine Spinetail 
Chaetura celebensis 
Alpine Swift 
Apus melba 
White-collared Swift 
Cy_pseloides zonaris 
San Geronimo Swift 
Hymenoptera 
99 
4th 
14 
94 
Panyptila sanctiheronymi_ 
Black Swift 
Cr.pseloides niger c . 14' 
100 
Common Swift 
A? 6 
3rd 
Zimmers Swift 
CyQseloides cryptus 
Whi te-throated Swift 
Aeronautes saxatalis 1st  
3rd 
Fork-tailed Swift 
Apus pacificus 1st 
97 
lst 
Homoptera 
< !  
c . 37 
66 
1 st  
3 1  
5th 
Coleoptera Isopter a Diptera 
3rd 1st  
54  23  
2 0 1 
l st? 
7 1 c .  31 
0 
8 1 7  
4th 2nd 
57 
0 
2nd 
2nd 1st  
2nd 3rd 
<3 
Bird ' s Sample 
Weight Size 
180 14 
100 
100 6 
98 2 
57  2 
46 6 
46 
43 24 
43 
43 13 
39 l 
33 2 
33 2 1  
30 40 
30 
30 
Source 
Morse & Laigo ! 968 
Arn 1945 
Arn 1945 
Rowley & Orr 1965 
Carr & Dickinsoo 1951  
Rathbun 1925 
Collins & Lane;- 1968 
Lack & Owen 1955  
Koskimies 195C1: 25  
Collins pers.  ::>r.tm. 
Collins pers. c<Jnm. 
flespenheide 19?5 
Bent 1940 
Li tvinenko 1972 
Chiba 1968 
Lea & Gray 1935 
0 
);! 
::0 
OJ 
c 
::0 
d z 
z 
? 0 ::0 z 
(j) 
c.:> c.:> 
I-' 
.p-.. 
0'1 
T AB!-E 7 - continued 
Horus Swift 
Apus horus 
House Swift 
Apus affinis 
Cayene Swift 
PanJ:I?tila cayennensis 
Chimney Swift 
Chaetura pelagica 
White-rumped Swift 
Apus caffer 
Chestnut-collared Swift 
CyQseloides rutilus 
White-tipped Swift 
Areonautes monti vagus 
Vaux Swift 
Chaetura vauxi 
Short-tailed Swift 
Chaetura brachl:ura 
Grey-rumped Swift 
Chaetura cineriventris 
Band-rumped Swift 
Chaetura SJ;!inicauda 
Neotropical Palm Swift 
Reinarda squamata 
Hymenoptera 
1 7  
78 
3rd 
1st  
5 1  
29  
82 
55 
Homoptera Coleoptera Isoptera 
4 1  6 19  
10  
1st  
2nd 
4th 2nd 5th 
0 0 49 
1st  
2 8 5 
1 1  
25  
Note:  1st  = most common, 2nd = second most common, etc . 
Diptera Bird ' s Sample 
Weight Size 
8 27  
4 27 I 
64 24 I 
2nd 24 1 2  
1st  24 
22 
0 20 4 
35 20 1 1  
1 8  
31  18  10  
42 18 17 
4 1 8  6 
27  1 6  7 
4 14  
53 10 6 
Source 
Collins 1 980 
Waugh & Hails 1 983 
Collins pers .  comm. 
Warren 1890: 183 
Fischer 1 958 
f!oreau 1942 
Collins 1968a 
Collins pers . comm. 
Davis 1 937 
Collins pcrs .  co::1m . 
Coll ins 1 968a 
Hespenheide 197  5 
Collins pers. comm. 
Hespenheide 1 975  
Collins pers .  comm. 
<D 
CO 
(J) 
::E 
I 
=i 
m :D 
c 
$: 
"U 
m 
0 
(/) 
? 
01 
-i 
r 
m 
-i 
? 
? 
,__. 
+:?
-....] 
TABLE 8 - Percentage composition (numerically) of major prey of swiftlets and swallows 
Purple Martin 
Progne subis 
Grey-breasted Martin 
Progne chalybeil 
House Martin 
Delichon urbica 
Barn Swallow 
Hirundo rustica 
Pacific Swallow 
Hirundo tahitica 
Black-nest Swiftlet 
Aerodramus maximus 
Rough-winged Swallow 
StelgidoEter):X 
ruficollis 
Bank Swallow 
Riparia riparia 
White-rumped Swiftlet 
Aerodramus spodiopyRius 
Mossy-nest Swiftlet 
Aerodramus vanikorensis 
Hymenoptera 
2 
75 
32 
82 
59 
60 
69 
31 
33 
6 
5 
7 
26 
10 
60 
Indian Edible-nest Swiftlet 
Aerodramus unicolor 
Edible-nest Swiftlet 
Aerodramus fuciehaga 4 1  
Grey-rumped Swiftlet 
Aerodramus francica 
Glossy Swiftlet 
Collocalia esculenta 48 
55 
Mangrove Swallow 
Tachycineta albilinea 94 
20 
Homoptera Coleoptera 
0 68 
0.3  -
I 4 1  
50 
6 
7 
23 
8 14 
24 3 
13  1 1  
2 7  12  
I S  8 
48 3. 5 
c .80 c . IO 
1 5  
1 5  
1 2  34 
3 3 
Isoptera Diptera Weight Sample Source 
Size 
0 20 51 c . IO Johns ton 196 7 
5 33. 5  65 Dyrcz 1985 
0 3 51 c.  9 Johnston 1967 
19.8 Bryant 1975 
2 8 19 .5  \laugh & Hails 1983 
2 31 17 .8  \laugh & Hails 1983 
27 <I 15 .9  Harrisson 1976 
4 13 .7  4 Hespenheide 1975 
33 15 .75 Beal 1918 
27 14 .6  Beal 1918  
32 1 4 .6  Stoner 1936 
69 14 .6  \laugh 1979 
5 1  1 2 . 2  10 Smyth 1980 
3 46 8 .2  20 This study 
0. 5 36 8 .2  1 2  This study 
27 <I 1 1 . 3 Harrison 1976 
c . l l . O  A l i  & Dillon 1 980 
8 10.8 Langham 1980 
49 c . IO 2 Collins pers. comm. 
26 8.3 Hails & Amirrudin 1981 
2 2 8.3 \laugh & Hails 1983 
0 <I 1 4 . 5  2 Ricklefs 1971 
26 Dyrcz 1984 
i\) 
):! ::0 
CD 
c ::0 
d z 
z 
? 0 ::0 z 
en 
(,) (,) 
I-' 
? 
CXl 
1986 WHITE-RUMPED SWIFTLET 13 
be flies as any other group. This is not surprising because many flies congregate 
at feeding or mating sites and so attract feeding swifts, swiftlets and swallows. 
To explain the greater dominance of social insects over beetles in prey 
taken than in prey available, Hespenheide pointed out that the social insects 
tended to congregate more and so the birds could presumably find such 
concentrations. There is a similar disproportion in the prey of the White?
rumped Swiftlet and the same reasoning could apply. My observations offeeding 
swiftlets flying in 10-30 m diameter circuits confirms that they do feed on 
insects that are swarming or in other high-density concentrations. 
Hespenheide ( 1975) found that swifts and swallows preferred the larger 
catchable prey of the range they could manage. If the same holds for swiftlets, 
flies, the most abundant but second smallest prey taken of White-rumped 
Swiftlets in Fiji, could not be taken because of their size alone. Flies must 
be chosen because they are easier to catch and/or more abundant. 
Tow net samples taken in Costa Rica and Panama consistently 
demonstrated that, although flies were 70-75% of airborne insects, they were 
only 4% of swift prey in the comparable wet season (Hespenheide 1975). 
Hespenheide presumed that the flies were harder to catch than other prey. 
If this is true of flies in Fiji, either the White-rumped Swiftlet is better able 
to catch flies than the swifts, swallows and other swiftlets whose prey contains 
few flies or the other kinds of flying insects are far less abundant in Fiji than 
in Central America and Malaysia. The latter cannot be so because the taxonomic 
proportions of the Fijian tow net samples (Table 9) are very like those of 
Central America. So perhaps the White-rumped Swiftlet has greater ability 
in securing more manoeuvrable prey, although, as Tables 3 & 4 show, it is 
not alone in this ability. 
A likely alternative for flies being chosen, other than their being easier 
to catch or more abundant, is that in Fiji they occur in high density in small 
areas. In Central America, flies may not have been in swarms or swarms of 
larger prey may have been more attractive to the swifts and swallows. 
Published comments suggest that mosquitoes are fewer in Fiji than 
elsewhere because the swiftlets hunt them tirelessly (Wood & Wetmore 1926, 
Sibson in Belcher 1972, Allison 1978/79). I doubt these statements because 
mosquitoes were 2.5% (2 11852) of free-flying insects but only 0.58% (4317433) 
of the swiftlet's prey. In addition, four of the six places I have lived at or 
visited within the range of the swiftlet had large numbers of mosquitoes. 
Food abundance and the timing of breeding 
Some evidence suggests that the dry season is a better breeding time 
than the wet season for birds that feed on the wing. Hespenheide ( 1975) noted 
that the sw;1llows and most other insectivorous birds nest in the dry season. 
He also suggested that, although in the wet season the density of flying insects 
is higher in cloudy but dry periods and ants and termites seem to swarm 
most, the more frequent rains must reduce the bird's foraging time. In Asia, 
the Edible-nest Swiftlet (Langham 1980), the Black-nest Swiftlet and the Mossy?
nest Swiftlet (Medway 1962) hatch most eggs during the dry period November 
to March. 
1 4 9  
14 TARBURTON NOTORNIS 33 
However, such is not always the case. The Indian Edible-nest Swiftlet 
Aerodramus unicolor (Abdulali 1942), the Pacific Swallow and the Glossy Swiftlet 
Col!ocalia esculenta (Waugh & Hails 1983) produce most of their first broods 
with the onset of the monsoon rains in May. 
In Fiji, the Whitc-rumpcd Swiftlct also breeds during the season of 
heavy rainfall.  Nests are built in September and October, corresponding with 
an increase in rainfall (Table 9). I suspect that increase to be the trigger because 
the increase in both rain and nest building occur so soon after August, the 
driest month of the year. Laying in November and early December corresponds 
with a further increase in rainfall. The high level of rainfall continues to April 
and so covers the period that young are being fed in the nest and the critical 
period during which the young are learning to feed themselves on the wing. 
TABLE 9 - Monthly rainfall averages in mil l imetres - Koronivia 
Research Station ( 1 950- 1 979 )  
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Year 
367 300 399 359 239 183 ! 7 1  ! 54 204 22! 305 296 3198 
Further evidence that there is an increase in the number of flying insects 
during the wet season in Fiji is the high correlation (rs = 0.8) between date 
and the number of insects caught in the aerial tow net during December. 
The raw data were 5 December 10 insects, 6 December 97 insects, 9 December 
68 insects, 11 December 265 insects, 15 December 162 insects. Confirmation 
of this trend is needed from net samples taken in every month. 
Although flies were dominant in most of the combined 198 1 and 1983 
boluses, that does not prove that this swiftlet specialises in flies. If I had 
taken more boluses in 1983, the overall result would probably show planthoppers 
as predominant because, as Table 10 shows, planthoppers made up 48% and 
flies only 36% of the total 1983 sample. 
TABLE 1 0  - Composition of White-rumped 
Swiftlet prey 
% boluses % boluses % of total 
Order dominant in  present in  sample 
'81  '83 '81  '83 ' 81 '83 
Diptera 60 36 100 9 1  46  36 
Homoptera 20 64 !00 lOO 1 5  48 
Hymenoptera !0 0 95 91 26 10 
Coleoptera 5 0 95 73 8 3 . 5  
Isoptera 5 0 40 27 3 0 . 5  
98% 98% 
Inadequate sampling or a real change in prey composition over time 
has led several workers to make generalisations which later study has shown 
to be incorrect. The large range of foods in boluses collected at the one time 
from this and other studies demonstrates how sampling could give biased results. 
The abundance of various insects can fluctuate greatly for various reasons 
such as current and past insect density, disease, predation, climate, and responses 
in prey or plant food species (Bos & Rabbinge 1976, Dixon & Barlow 1979, 
Anderson & May 1980, Barlow & Dixon 1980, Randall 1982). Such fluctuations 
are likely in many insects and will restrict the choice of prey for aerial feeders. 
1 5 0 
1986 WHITE-RUMPED SWIFTLET 
ACKNOWLEDGEMENTS 
15 
For published and unpublished data on the food of several swifts, I 
thank Charles T. Collins of California State University. I also thank, at Massey 
University, Edward Minot for comments on two drafts of this paper and Ian 
Stringer for identifying the more difficult prey. For editing and criticism of 
the final drafts I thank Nigel Langham and Barrie Heather. 
LITERATURE CITED 
ABDULALI, H. 1942. The terns and edible-nest swiftlct? at Vcngurla, wc?t coast India. j.  Bombay Nat. Hist. 
Soc. 43: 446-45 1 .  
ALl, S.; DILLON, R.S. 1970. Handbook o f  the: birds o f  India & Pakistan:4. frogmouths to pittas. Bombay. 
ALLISON, N.A. 1978/79. Notes on birds of Viti Lcvu & Vanua Lcvu islands, Fiji. Sea Swallow 29: 10?1 1 .  
ANDERSON, R.M.; MAY, R.M. 1980. Infectious diseases & population cycles o f  forest insects. Science 210: 658-
661. 
ARN, H. 1945. Zur biologic des Alpcnscglcrs Micropus melba melba (L.). Schwciz. Arch. Ornith. 2: 137?181 .  
BARLOW, N.D.; DIXON, A.F.G. 1980. Simulation o f  lime aphid population dynamics. PUDOC. Wagcningcn 
Netherlands. 
BEAL, F.E.L. 1918. Food habits of the swallows, a family of valuable native birds. Bull. U.S. Dept. AG. 619. 
BELCHER, W.J. 1972. Birds of Fiji in Colour. Auckland: Collins. 
BENT, A. C. 1940. Life histories of American Cuckoos, Goatsuckcrs, Hummingbirds & their allies. U.S. Nat. Mus. 
Bull. 176: 254-27 1 .  
BOS, VAN DEN, ] . ;  RABBINGE, R. 1976. The Grey Larch Bud Moth & its population dynamics. PUDOC 
Wageningen, Netherlands. 
BRYANT, D.M. 1975. Breeding biology of House Martins Deliclwn urbica in relation to aerial insect abundance. 
Ibis ll7: 180-216. 
CARR, M.H.; DICKINSON, J.C. 195 1 .  The San Gcronimo Swift in Honduras. Wilson Bull. 63: 271-273. 
CHI BA, S. 1968. Analysis of stomach contents of Nucifraga caryocaractes and A pus pacificus of Japan. Mise. Rep. 
Yamashina Inst. Ornith. 5: 287-297. 
COLLINS, C.T. 1968a. The comparative biology of two species of swifts in Trinidad, \Vest lndics. Bull. Florida 
State Mus. l l : 257-320. 
COLLINS, C.T. 1968b. Notes on the biology of Chapman's Swift Chaetura clzapmani (Aves:Apodidae). Am. Mus. 
Novit. 2320. 
COLLINS, C.T. 1980. Notes on the food of the Horus Swift (A pus horus) in Kenya. Scapus 4: l0-13. 
COLLINS, C.T.; LANDY, M.J. 1968. Breeding of the Black Swift in Vcracrus, Mexico. Bull. Sthn. Calif. Acad. 
Se. 67: 266-268. 
DAVIS, W.B. 1937. A Vaux Swift and its young. Condor 39: 222-223. 
DIXON, A.F.G.; BARLOW, N.D. 1979. Population in the lime aphid. Zoo!. J. Linn. Soc. 67: 225-237. 
DYRCZ, A. 1979. Die Nestlingsnahrung bei Drosselrohrsanger (Acroccphalus arundinaceus) und Reichrohrsanger 
(Acrocephalus scirpaceus) an den teichen bei milicz in Polcn und zwei seen in der Westschweiz. Om. Beobachter 
76: 305-316. 
DYRCZ, A. 1984. Breeding biology of the Mangrove Swallow (Tachycineta albilinea) and the Grey-breasted Martin 
(Progne chalybea) at Barro Colorado Island, Panama. Ibis 126: 59-66. 
?:ISCHER, R.B. 1958. The br.::eding biology of the Chimney Swift Clwcrura pclagica (Linnaeus). N.Y. State M us. 
& Se. Scv. Bull. 368: l-141.  
HAILS, C.J.; AMIRRUDIN, A.  198 1 .  Food samples and food selectivity of White-bellied Swiftlcts (Col/ocalia 
tsculenta). Ibis 123: 328-333. 
HARRISSON, T. 1976. The food of Collocalia swiftlets (Avcs:Apodidac) at Niah Great Cave in Borneo. J. Bombay 
Nat. Hist. Soc. 7 1 :  376-393. 
HESPENHEIDE, H.A. 197 1 .  Food preferences and the extent of overlap in some insectivorous birds, with special 
reference to the Tyrannidae. Ibis 113: 59-72. 
HESPEN?EIDE, H.A. 1975. Selective predation by two swifts and a swallow in Central America. Ibis 117: 82-
99. 
JOHNSTON, R.F. 1967. Seasonal variation in the food of the Purple Martin (Progne subis) in Kansas. Ibis 109: 
8-13. 
KOSKOMIES, J. 1950. The life of the Swift, Micropus apus (L), in relation to the weather. Ann. Acad. Sci. Fenn. 
Scr. A. 15: 1-151 .  
LACK, D. 1956. Swifts in a tower. London. Methuen. 
LACK, D.; OWEN, D. F. 1955. The food of the swift. J. Anim. Ecol. 24: 120-136. 
LANGHAM, N. 1980. Breeding biology of the Ediblc-ncst Swiftlct (Aerodramusfuciphagus). Ibis 122: 447-460. 
LEA, A.M.; GRAY, J.T. 1935. The food of Australian birds. Emu 35: 63-98. 
LEWIS, T.; TAYLOR, L.R. 1964. Diurnal periodicity of flight by insects. Trans. Royal Ent. Soc. Lon. 116: 393-
476. 
LITVINENKO, N.M. 1972. On the food of the Whitc-rumped Swift in the southern Primorye. Ornithologiya 
10: 361-362. Trans. from Russian by A. N. HILKEVITCH, Calif.Statc Uni. 
MED WAY, L. 1962. The swiftlcts (Colloca/ia) of Niah Cave, Sarawak. Part I:  Breeding Biology. Ibis 104: 45-66. 
MOREAU, R.E. 1942. The breeding biology of Micropus caffer streubelii, the White-rumpcd Swift. Ibis 84: 27-
49. 
MORSE, R.A.; LAIGO, F.M. 1968. The Philippine Spine?tailed Swift, Chaewra dubia McGrcgor, as a honey bee 
predator. Phi!. Entom. 1: 138-143. 
RAND ALL, M.G.M. 1982. The dynamics of an insect population throughout its altitudinal distribution: Coleophora 
a/teco/ella (Lepidoptera) in northern England. J. Ani m. Ecol. 51 : 993-1016. 
RATHBUN, S.F. 1925. The Black Swift and its habits. Auk. 42: 497-516. 
ROWLEY, J.S.; ORR, R.T. 1962. The nesting of the White-napcd Swift. Condor 64: 361-367. 
SMYTH, D.M. 1980. Notes on the diet of the Grey Swiftlct,Aerodramus spodiopygius. Sunbird l l :  20-2 1 .  
1 5 1  
16 TARBURTON NOTORNIS 33 
STONER, D. 1936. Studies on the Bank Swallow, Riparia np11ria ripariu (Linnacus}, in the OncJda Lake region. 
Roosevelt Wldl. Ann. 4: 127-233. 
WARREN, B.H. 1890. Report on the birds of Pennsylvania. Cwhh. Pcnn. Harrisburg. 
WAUGH, D.R. 1979. The diet of Sand Martins Riparia ripariu during the breeding ?cason?. Bird Study 26: 123-. 128. 
WAUGH, D.R.; HAILS, C.J. 1983. Foraging ecology of a tropical aerial-feeding bird guild. Ibb 125: 200-2 17. 
WOOD, C.A.; WET MORE, A. 1926. A collection of birds from the Fiji i::.land:.. ( 12) Ibis 2: 9 1 - 136. 
MICHAEL K. TARBURTON, Depamnem of Botany and Zoology, Massey 
University, Palmerston North 
---??- * 
1 5 2  
THE FOOD OF THE WHITE-RUMPED SWIFTLET (Aerodrarnus spodiopygius ) IN 
QUEENSLAND 
ABSTRACT 
1 5 3  
Homoptera (planthoppers ) ,  Diptera (flies ) , Hymenoptera ( social 
insects ) and Isoptera ( termites ) ,  were the most numerous prey in 44 food 
boluses being delivered by parent White-rumped Swiftlets (Aerodramus 
spoctiopygius chillagoensis ) to their chicks inside six Chillagoe caves . 
NUmerically the main food items were planthoppers ( 47%) and flies ( 24%) . 
The number of insects in each food bolus ranged from 7 to 587 (x = 
149 ) . The average weight of a bolus was 0 . 33 g ( range 0 . 11 - 0 . 62 g ) . 
The average length of all prey was 3 . 6  mm , which is larger than the 
average length of available prey ( 2 .  2 mm ) .  The number of prey "species" 
ranged from 2 to 83 (x = 40) per bolus . Altogether , 303 invertebrate 
" species" were recorded in food boluses . 
The White-rumped Swiftlet breeds during the wet season , when insects 
are generally accepted as being more abundant . However the density of 
potential prey is shown to be significantly lower than that taken during 
the breeding season in Fiji . 
INTRODUCTION 
The White-rumped Swiftlet (A. s .  assimilis ) has been shown ( Tarburton 
1986 , Appendix 2 ,  pp 137 - 152 ) to take flies as its most common prey in 
some seasons and over certain habitats in Fiji . Because published studies 
(Hespenheide 1975 , Lack 1956 ) had not shown flies to be the dominant taxon 
in the prey of other swifts , it appeared worthwhile to make an analysis of 
the prey of the same species where it feeds in different habitats and 
climates . There are a number of factors which might alter the type and 
size of the prey taken over the savannah at Chillagoe . 
1 5 4 
Chillagoensis being larger than assimilis , should enable it to take 
larger prey (Hespenheide 1971 , 1975 , Dyrcz 1979 ) . Larger prey will only 
be available in similar taxonomic proportions as it was in ?iji if the 
different climate , food source for the prey and competition from what is 
likely to be a greater number of species of aerially feeding competitors , 
make them available . Having these variables as they apply to the same 
bird should help discern some of the causes for aerially feeding birds not 
taking prey in the taxonomic proportions that are found to be available . 
METHODS 
During December 1985 , January and December 1986 and January 1987 I 
studied the food of A. s .  chillagoensis nesting at Chillagoe . Food 
boluses were taken from adults caught in narrow sections of caves as they 
delivered the food to their nestlings . Such boluses were taken from Guano 
Pot , Gordale Scar Pot , New Southlander , Crack Pot , Keef ' s Cavern and 
Golgotha Cave . The approximate locations of these caves are shown in 
Section 1 ( p .  10) . 
Swiftlet food samples were collected and measured in the same way as 
described in the previous Appendix ( p  138 ) . 
In Fiji I sampled potential prey of assimilis by the methods of 
Hespenheide ( 1975 ) , but chillagoensis rarely fed below 8 m in the first 
season, so I could not sample the air they were feeding in with the net 
mounted on a vehicle . Instead , I sampled the flying invertebrates 
( potential prey) by placing the same net as used in Fi ji on a five m pole 
and then swung it through the air in circular and figure of eight motions . 
In order to reach the air space that the swiftlets were using I stood on 
the top of limestone outcrops (Suicide & Spring towers as well as tower 
5 126 in Chillagoe township)  or on the tank stand at the rear of two 
Queensland National Parks & Wildlife homes . 
1 5 5  
Swiftlets often fed at these locations and were doing so while some of the 
samples were being collected . Nine samples were collected in a similar 
manner in Fiji . 
Much of the statistical reporting is based on mean measurements and 
their standard errors and is shown in text and tables as x + SE . The data 
for determining the size of all prey were based on the means of all 44 
boluses rather than that of each type so that the extreme means of the 
uncommon types did not swamp those of the majority . Whereas Table 2 shows 
the actual minimum and maximum sizes of all potential prey netted , those 
few measurements that were above or below the size that the birds were 
found to have taken were not used in calculating the average of available 
prey. It is assumed that insects smaller than those commonly caught are 
' selected against ' for perceptual or energetic considerations , and that 
the one insect that was larger than the maximum caught by the birds 
sampled , would be too large for the bird to handle .  
RESULTS 
Identity of prey 
Planthoppers were found in all 44 food boluses and were the most 
numerous in 18 of them (Table 1 )  making 47% ( 3102 individuals )  of the 
total sample of 6549 invertebrates . Flies were found in 41 of the food 
boluses , and were the most numerous prey in 11 of them making 24% of the 
total sample.  Social insects were in 43 boluses and were the most 
numerous in three boluses . They made up 17% of the total sample . 
Termites were the most numerous in three boluses ,  but beetles , plant bugs 
(Hemiptera) and spiders were present in more boluses than were termites . 
Thirty-one percent ( 1 , 669 ) of the 5 , 334 insects taken from boluses in 
the 1985/86 season were just three "species" of jumping plant-lice of the 
family Psyllidae . 
Table 1 Composition of White-rumped Swiftlet prey in 32 food boluses : Chillagoe , 1 985/6 
No . % % Range x %  of No . of No . in 
Where Where In All Samples Where Boluses Total 
Order Dominant Dominant Boluses Present + SE Present in Sample 
Homoptera 14 36 - 98 0 - 98 38 + 3 . 9  30 2352 
Diptera 11 38 - 56 0 - 56 27 + 3 . 2  30 1440 
Hymenoptera 5 42 - 68 <1 - 68 23 + 3 . 0  32 880 
Isoptera 2 48 - 98 0 - 98 28 +u . o  8 81 
Aranae 0 0 0 - 12 5 + 0 . 7  24 265 
Heteroptera 0 0 0 - 33 4 + 1 . 6  20 182 
Coleoptera 0 0 0 - 9 3 + 0 . 6  21 109 
Thysanoptera 0 0 0 - 1 <1 7 ll Lepidoptera 0 0 0 - 7 2 + 1 . 1  6 11 Phasmatodea 0 0 0 - <1 <1 1 1 Unidentified 0 0 0 - <1 <1 1 1 
% Individuals 
In Total 
Sample 
44 
27 
16 
2 
5 
3 
2 
1 
I-' 
lJl 
C)\ 
1 5 7  
The total of 298 species for that season was made up of 90 Hymenoptera , 75 
Diptera, 54 Araneida , 36 Homoptera , 25 Coleoptera , 11 Heteroptera , 3 
Lepidoptera , 3 Isoptera, and 1 Thysanoptera. 
Size of prey 
The largest prey found in this study were three termites and a social 
insect , each 10 rrnn long . The next largest prey were 9 mm long and 
included four social insects , a moth , a fly and the only mantid in the 
sampled prey. Termites were the largest of the common prey , averaging 6 . 4  
rrnn , then social insects ( 4 . 2 mm) , plant bugs ( 3 . 0  mm) , flies ( 2 . 5  rrnn) ,  
spiders ( 2 . 4  rrnn) and planthoppers ( 2 . 4  mm) . The average size of all prey 
from the 44 boluses was 3 . 64 .?. 0 . 24 mm (x  .?. SE) , which is significantly 
greater ( t77 = 3 . 89 ,  p< 0 . 001 , )  than that of available prey ( 2 . 4  + 0 . 2 ) . 
While the average size of prey in each major taxon was not 
significantly greater than the average size of potential prey ( except for 
the Hymenopterans where t87 = 15 . 2 ,  f< 0 . 001 ) , the data as shown in Figure 
1 clearly shows the captured prey to be consistently larger than the 
available prey . A comparison of the maximum and minimum lengths of 
potential and actual prey (Table 2 )  shows that although prey smaller than 
1 rrnn is  available , the birds do not take i t .  Termites and moths smaller 
than 3 . 5  rrnn are not common in either available or captured prey . 
Table 2 Maximum and minimum sizes of potential and actual prey 
taxon 
Homoptera 
Diptera 
Hymenoptera 
Isoptera 
araneida 
Coleoptera 
Heteroptera 
Thysanoptera 
Lepidoptera 
Phasmotodea 
Blattodea 
bolus 
min . 
1 . 0  
1 . 0  
1 . 0  
3 . 8  
1 . 0  
1 . 0  
1 . 5  
1 . 0  
4 . 0  
prey 
max .  
10 . 0  
9 . 0  
10 . 0  
10 . 0  
5 . 5  
3 . 3  
8 . 0  
2 . 0  
9 . 0  
9 . 0  
potential prey 
min . max. 
1 . 0  2 . 6  
0 . 8  8 . 5  
0 . 4  7 . 0  
3 . 2  9 . 0  
2 . 0  
1 . 0  6 . 0  
1 . 7  2 . 2  
0 . 8  1 . 5  
9 . 0  
7 . 0  
1 5 8 
Abundance of potential prey 
The sweep-net samples of potential prey at Chillagoe caught an average 
of 6 ( SE = 1 ,  n =39 ) insects of the size range found to be taken by the 
birds ( 1-10 mm) .  This estimate of density of available prey is far below 
the average of 95 ( SE = 29 , n = 16 ) insects caught in the same net in 
Fi ji . 
The Food Bolus 
The weight of 32 White-rumped Swiftlet boluses ranged between 0 . 11 -
0 . 62 g ,  averaging 0 . 33 g ? 0 . 02 .  
The number of insects in a bolus varied from 7 to 587 . The average 
number for all 44 boluses was 149 + 21 . Further analysis of the numbers 
of individuals and species in the major orders is shown in Table 3 .  
Table 3 Frequency of major prey in food boluses (x ? SE ) 
Individuals Species 
Homoptera 71 . 0  +16 . 0  5 . 0  + 1 . 0  
Diptera 35 . 0  + 7 . 0  11 . 0  + 1 . 0  
Hymenoptera 26 . 0  + 3 . 0  11 . 0  + 1 . 0  
araneida 6 . 0  + 2 . 0  4 . 0  + 1 . 0 
Heteroptera 4 . 0  + 2 . 0  1 . 0  + 0 . 2  
Coleoptera 3 . 0  + 1 . 0 2 . 0  + 0 . 4  
Isoptera 4 . 0  + 1 . 0  0 . 4  + 0 . 1  
Total 149 . 0  + 21 32 . 0  + 4 . 0  
' Species ' i s  not a named species but i s  ascribed to individuals 
that are morphologically similar . 
DISCUSSION 
Identity of prey compared with that of Fi ji 
While flies made up 43% of the 32 boluses collected from Fi jian 
swiftlets they fell to being the second most common prey ( 21%) in the 44 
boluses collected from swiftlets at Chillagoe . 
1 5 9  
Planthoppers , which made up 24% and were the second most common prey in 
Fiji , were the most common prey ( 37%) in the samples from Chillagoe . Most 
other taxa were found in similar proportions except for spiders which 
composed only 1% of prey in Fi ji and made up ?? of prey at Chillagoe , and 
beetles which were 7% of prey in Fi ji and were only 2% of prey at 
Chillagoe . 
The most common flying insects available to swiftlets in Fi ji and the 
second most common available at Chillagoe were flies . Whereas in Fi ji the 
major portion of the prey was flies , they were not the most common prey in 
Chillagoe . This could be explained if the planthoppers which the 
Chillagoe birds concentrated on were to be found in swarms which I never 
sampled with my sweep-net . 
Alternatively , one can accept Hespenheide ' s  ( 1975 ) suggestion that flies 
are more manoeuvrable than most insects and that this helps explain their 
infrequent occurrence in the prey of aerial predators in general . 
If this is the case then the White-rumped Swiftlet in Fi ji has greater 
ability than most apodids , including chillagoensis ,  to capture more 
manoeuvrable prey . This is possible as assimilis weighs less and has 
longer wings than chillagoensis . That flies are not as large a majority 
in the available prey at Chillagoe as they are in Fi ji also helps explain 
their under-representation in the diet of Chillagoensis .  
It might be that non-fly prey i s  easier to obtain at Chillagoe . 
Evidence for this is that while Fijian birds forage for 15 . 5  hours a day 
swiftlets at Chillagoe forage for only 13 hours . Fijian swiftlets leave 
their caves just after 0400 hrs and return for the night mostly after dark 
between 1930 and 2000 hrs , while the majority of swiftlets at Chillagoe do 
not leave the cave until around 0530 hrs and most return before dark 
around 1830 hrs . However , there is also evidence against the view that 
the birds at Chillagoe have less difficulty finding enough food for their 
chic?. 
1 6 0  
Swiftlets at Chillagoe are only gathering food for one chick whereas 
Fijian swiftlets are collecting food for two chicks . Fi jian birds also 
breed in larger colonies than those at Chillagoe , which means that on 
average they have to fly further to their feeding areas . 
However , we need not appeal to either of these explanations for as was 
shown in the previous appendix (p  150) , such variation could result from 
the time of sampling . Examples given there show that the most abundant 
taxon in the prey of apodids varies with time even though flies are 
consistently the commonest prey available . In one season in Fiji this 
swiftlet did not capture a majority of flies even though they were the 
most abundant taxon. 
Table 4 Taxonomic Proportions of Prey Compared with Aerial Invertebrates 
x %  in x %  in 
Order Sweep Net Food Boluses 
Hymenoptera 41 27 + 3 
Diptera 39 21 + 3 
Thysanoptera 6 <1 
Homoptera 5 37 + 4 
Coleoptera 5 2 + 1 
Isoptera 2 8 "+ 3  
Heteroptera 2 2 + 1 
Aranea <1 3 "+ 1  
Lepidoptera <1 <1 
Blattaria <1 <1 
NOTE Taxa are arranged in descending order of the proportion available . 
Chillagoensis does not take its prey in similar taxonomic proportions 
to those available within the size range that it  handles ( Table 4) . If it 
did , flies would predominate , followed by social insects , thrips and 
planthoppers . The proportion of each taxon l.n available prey is very 
close to that available in Fiji and in Costa Rica and Panama (Hespenheide 
1975) , except that thrips replace beetles at Chillagoe . As Hespenheide 
( 1975 ) predicted , the proportion of flies caught was below that available .  
The reason he gave was that the manoeuvrability of flies is said to be 
better than that of social insects and beetles . 
1 6 1  
However , as pointed out in the previous Appendix ( p  150)  the proportion of 
flies in a sample varies largely with the time of sampling and these data 
may not contradict those from Fiji where flies predominated in one of the 
two large samples . The under-representation of social insects in the diet 
of swiftlets from both Chillagoe and Fi ji does not support Hespenheides ' 
suggestion that the poor manoeuvrability and tendency to swarm make this 
prey taxon preferred above flies and beetles . 
Size of prey 
Because the Chillagoe birds ( x  = 9 . 3  ? 0 . 04 g)  are significantly 
larger ( t400 = 15 .4 ,  f <0.001 ) than Fijian birds (x = 8 . 19 ? 0 . 06 g ) , and 
because prey size has been positively related to the body size of 
insectivorous birds (Hespenheide 1971 , 1975 ; Dyrcz 1979 ) , it would be 
expected that the Chillagoe swiftlets would take larger prey than the 
Fijian swiftlets . The average size of prey taken by the swiftlets at 
Chillagoe ( 3 . 64 ? 0 . 24 mm) was significantly larger ( t74 = 4 .32 ,  f <0 .001)  
than the average size of  prey taken by Fijian swiftlets ( 2 . 48 + 0 . 11 mm) . 
The available prey sampled by the sweep-net was significantly larger ( t32 
= 4 .37 ,  P <0.001 )  at Chillagoe ( 2 . 58 ? 0 . 17 mm) than at Fi.ji ( 1 . 74 ? 0.09 
mm) . 
It has been suggested (Hespenheide 1975 ) that prey smaller than the 
minimum size taken is not taken due to either perceptual reasons or 
because the relative ease of capture for the different taxa converges at 
small sizes to very similar values . 
Food abundance and the timing of breeding 
The discussion in the previous Appendix ( p  149 ) about breeding in the 
wet season versus breeding in the dry season hardly applies to birds in a 
savannah regimen as it is normal practice for birds to breed in the wet 
season. 
1 6 2  
However , in both Fiji (p  21 ) and Chillagoe (Table 5 )  it is the onset of 
the wet season that coincides with breeding in the White-rumped Swiftlets . 
Laying in November and early December corresponds with an increase in 
rainfall at both Chillagoe and Fiji . The high level of rainfall continues 
to April in Fi ji but only until March at Chillagoe . 
Table 5 Monthly rainfall averages in millimetres- Chillagoe P . O .  
( 1902-1986) 
Jan Feb Mar Apr May Jun Jul 
217 216 147 30 14 10 4 
Aug Sep Oct 
4 6 14 
CONCLUSIONS 
Nov Dec 
55 138 
Year 
855 
Although flies were the commonest of the insects available they were 
second most common to planthoppers in the prey of chillagoensis .  However , 
as only one of the 260 insects caught in the sweep-net was a jumping plant 
louse , which composed 31% of the prey in the good season , it is clear that 
the sampling technique was not adequate in all respects .  Despite this 
deficiency it  was shown that the swiftlets were taking larger prey than 
that which was available . 
Chillagoensis took larger prey in larger boluses which contained fewer 
individuals than was the case for assimilis . Each bolus taken by 
chillagoensis had an average of 10 species more than those taken by 
assimilis . Thirty-two boluses of invertebrates taken by chillagoensis 
contained 303 species , whereas the same number of boluses taken by 
assimilis contained 167 species . The fact that there are no other swifts 
or swallows resident in the area accounts , as much as the difference in 
size between Australia and Fiji , for chillagoensis having larger prey of 
more species available to it than does assimilis . 
1 6 3  
ACKNOWLEI:X}EMENTS 
I wish to acknowledge the helpful criticism of the drafts of this 
paper by Prof Brian Springett and the identification of the plant lice by 
Dr Ian Stringer . The Chillagoe National Parks staff were very helpful in 
allowing entry to their parks and homes for this study . 
REFERENCES 
DYRCZ , A.  1979 . Die nestlingsnahrung bei drosselrohrsanger (Acrocephalus 
arundinaceus) und teichrohrsanger (Acrocephalus scirpaceus) an den 
teichen bei milicz in Polen und zwei seen in der Westschweiz .  
Ornithologie Beobachter 76 : 305-316 . 
HESPENHEIDE , H .A. 1971 . Food preferences & the extent of overlap in some 
insectivorous birds , with special reference to the Tyrannidae . Ibis 
113 : 59-72 . 
HESPENHEIDE, H .A. 1975 . Selective predation by two swifts & a swallow in 
Central America. Ibis 117 : 82-99 . 
LACK, D .  1956 . Swifts in a tower. London. Methuen . 
TARBURTON, M.K.  1986 . The food of the White-rumped Swiftlet (Aerodrarnus 
spodiopygius) in Fiji . Notornis 33 : 1-16 . 
APPEND I X  4 
An experimental manipulation of clutch and brood size of 
White-rumped Swiftlets Aerodrarnus spodiopygius of Fiji 
1\11 . K .  TARDtJ RTO N 
1 29 :  1 07-1 1 4  Accepted l ') :\ l arch l lJ8(J 
By monitoring hatching success, chick growth and  fledging success in normal sized and experimentally 
enlarged clutches and broods of the \Vhite-rumped Swift let Aerodramus spodiopygius in  Fij i ,  this paper 
demonstrates the inability of this species to raise more young t han i s  normal. 
1 6 4  
1 6 5  
ro8 SHORT C0:\1 :\1 U K J CAT J O;-.; s  I B J S  I 2 
Winkler & V\' alters ( 1 98 3 )  han? stated that 'b irds lay that number of eggs that resul 
in the parents operating at the optimal working capacity ' .  In stating this ,  they accep 
Ashmole's update ( 1  W> 1 ,  1 %3 )  of Lack 's hypothesis ( 1 94-7) that  birds are maximiz 
ing their clutch size in  respect to the quantity o f  food that is  immediately availabJ, 
during the breeding season.  Howe\'er, they also recognize that parents may bo 
maximizing their reproductive output oYer their entire l i fetime rather than within : 
single breeding attempt.  A number of authors use the latter \?iew to predict that somt 
brood sizes wil l  be below the maximum that parents can feed in a particular seasor 
(e .g . ,  Will iams 1 966,  :VIurphy 1 968 ,  Gadgil & Bossert 1 970 ,  Charnm? & Krebs 1 974 
Goodman 1 974, Pianka & Parker 1 975 ) .  
It  would be  expected that l cmg-l iYed b irds  with small cl utches would be amongst 
those most l ikely to be working below ? their capacity to gather food in o rder to 
enhance their l ifelong capacity  to raise more young. It has been shown by mon i toring 
chick growth in artificial ly enlarged broods that some b irds hm?e been able to gather 
enough food to raise more chicks than normal at one t ime.  Howe\?er, all such results 
were obtained with temperate species and no such manipulations have been carried 
out on swiftlets which are long-l iYed and tropical . 
Methods 
The subjects of this manipulation experiment were \\'hite-rumped Swiftlets nesting 
in 'Dry'  and 'Waterfal l '  caves north of  Suva (Tarburton 1 986) .  Daily vis i ts were 
made to these colonies throughout December 1 98 1  and again in 1 983 .  On these visits 
eggs were checked for hatching and measurements were made of each chick's wing, 
tarsus and weight. Six natural single-egg clutches and 37 natural two-egg c lutches 
were monitored as controls along with the manipulated clutches. Single-egg clutches 
were manipulated by removing one of the two eggs from 24 normal clutches. The 
eggs that were removed were added, as were another three, to normal c lutches to 
make up 27 three-egg clutches. S ix two-egg clutches were also made up by swapping 
eggs so that none of these birds were incubating their own eggs. The rates of growth 
for 1 3  one-chick and 39 two-chick natural broods were compared with those for 37 
one-chick, seven two-chick and 23 three-chick, manipulated broods. 
Lack ( 1 956) reported some parental desertion when the nests of Common Swifts 
Apus apus were disturbed. To test the effect of d isturbance on the swiftlets, I visited 
one isolated group of nests less frequently than other nests in  the stuay. The chicks at 
frequently-visited nests grew at the same rate as those in  nests visited less frequently. 
Nests were approached slowly to allow incubating parents time to leave the nest 
without the haste that caused increased egg loss.  After being handled most chicks 
were restless and l ikely to fall from the nest. Thi s  made it  necessary to keep an eye on 
a brood for about a minute after replacing i t .  
Individual chicks within  a brootl were identified ini t ially by placing a daub of  
dark green quick-drying model paint on the  head , shoulders or ru:JfRp.  Once the  tibia 
was large enough to retain a band, individually numbered size one aluminium bands 
from CSIRO Canberra, Australia were used . I banded all  swiftlets on the tibia 
instead of the usual tarso-metatarsus. Bands applied to the tarsus often slip over the 
toes (Tarburton unpubl ished) .  The swiftlets usually stay on the wing all day (except 
when breeding) and this prev?nts the band from damaging the thin skin and feathers 
of the tibia.  
The day of hatching was known for only 22  chicks.  Their  wing,  tarsus and weight 
measurements exhibited an e\?en exponential growth form through Day 8. In order 
to increase the sample of aged chicks, those up to eight days old when found were 
aged with a formula derived using regression anah?sis of measu remcnts of known age 
chicks. 
S HOHT CO;\I ;\ I ll !' I CA T I O!'S 1 09 
Four ful l -day watches were made at a sample of  manipu lated and untouched 
nests to determine the daily feeding rate of the d ifferent sized broods .  The low beam 
on a miner's lamp was used to make the obsen?at ions.  
Results 
H atclzing success 
The hatching success of natural cl utches and of manipulated clutches were not 
significantly different .  Those with natural clutch size one hatched an average of 
0?43 ? 0 ? 1 8  (x ? s . e . ,  n = 9),  while the manipulated ones hatched an average of 
0?6 1 ? 0? 1 2, (n = 20); (:\led ian test;  1.7 = 0? 28, n . s . ) .  Pairs with a natural c lutch size of 
two hatched, on average, 1 ? 1 8  ? 0? 1 5 , (11 = 34) and the manipulated twos, 1 ?0 ? 0 ?26 ,  
1 . (n = 6) (;(2 = 0?6 ,  n.s . ) .  
The combined hatching success of single-egg cl utches was 0? 5 2 ? 0 ?09  (n = 29)  
and that  of  the  combined two-egg clutches 1 ? 1 5  ? 0? 1 4  (n = 40), showing a clear 
advantage to the two-egg clutch (t65 = 3 ?8 1 ,  P < 0?00 1 ) .  The manipulated three-egg 
clutches had an average clutch hatching of 1 ? 2 2  ? 0 ? 1 7  (n = 27) ,  which was not 
significantly larger than that of the two-egg clutches (I ss = 0?3 3 ,  n . s . ) .  
Chick growth 
Figures 1 ,  2 and 3 show the mean daily increase in length of wing and tarsus  and 
weight for individuals  in broods of all three sizes. The overlapping standard errors 
on the wing growth curves indicate no significant d ifference between broods of one 
and two whereas that of three i s  significantly less than both one and two after Day 10.  
Average adult  wing l ength was not reached before birds fledged,  though the 
minimum adult wing length was reached by most fledglings . 
. The weight of chicks in broods of three was also significantly lower. After Day I 2 
a significant difference could be detected between those of single and two-chick 
broods, but  it was not as great as between those of t\YO and three-chick broods.  One?
chick b roods reached adult weight on Day 1 9, two on Day 22 and those of three-chick 
broods on about Days 30 to 50.  
Tarsal growth exhibited the least  variation, with no consistent differences 
amongst chicks of d ifferent clutch sizes . The tarsus was also the most rapid in 
achieving adult  size, which on average took 1 3 ? 5  days. 
Fledging success 
In order to determine brood success, an arbi trary point of success needed to be 
established,  due to the difficulty of knowing when the nestling period ended . The 
difficulty was caused by the wandering habits of older chicks which,  although 
possibly beneficial in exerc is ing their wings as they 'walked' around on the cave 
walls ,  made location of marked birds more diflicult .  Taking a wing length of 90 mm 
as indicating that a nestling is  likely to Hedge, the fol lowing determinations were 
made. Parents with a brood of one raised 0 ?4] .!. 0 ?  i I eh icks (x  ? s .e . ,  n = 24 ) .  Those 
with a brood of two raised 0 ?02  ? 0 ? 1 5  (11 = 25 )  and t hose \\? i th  a brood of three raised 
1 ?09 ? 0? 3  (n = 1 1 ) .  
These data show that a pair  with a brood of t \YO rear s ign i ficantly more young 
than those with a brood of one, ( l44 = 2? 57 .  P < 0?0 1 ), while those with a brood of three 
do not rear significantly more than those with broods of  two ( 1 1 ;  = 0? 59 ,  n . s . ) .  
1 6 6  
1 6 7  
I IO SHORT CO:V1 !\1 U K I CAT I ONS I B I S  I 29  
Adult x -
100 Adult min - -o ?o ... f o t + I o O  ? 
+ 
80 0 - t ?? ? . 
E 60 ?H? '? j ? + E. f? ? ? "' 0 ? + + c:: 
? ?cto.l>l-: + + + 
40 
q.if<?. + f .  0-o,."e ? 
20 
20 30 40 
Days 
Figure 1 .  Mean daily increase in the wing length of 'White-rumped Swiftlet chicks in different sized 
broods: the means for broods of  one are represented by hollow circles, broods of two by horizontal bars 
and broods of three by solid circles. Standard errors of the means (vertical lines) for chicks in  broods of 
three rarely overlap those for chicks from single or double broods after the tenth day. Although n = 1 1 8 ,  
not  all  chicks were measured each day,  hence the unevenness of the  growth curves. 
10 
8 
s 6 
:E .  "' 
? 4 
2 
10 20 30 
Days 
__,.._ 
? ?  
40 
? 
0 
Figure 2. :\-lean daily increase in \\'hi tc-rumpl'll Swiftk t  chick weight :  the means for broods of one are 
represented by hollow circles, broods of two by horizontal  bars and broods of three by solid circles. 
Standard errors of the means (vertical l ines) for chicks in broods of three rarelv o\?erlap those for chicks 
from single or double broods after the tenth day. Although 11 = 1 1 8 ,  not all  chicks were measured each day, 
hence the une\?enness of the growth curves. 
S H 0 R T C 0 !V1 !\1 U N  I CAT  I 0 N S I I I 
1 0  
.c 8 
"& c "' 
..J 
1 0  20 30 
Days 
Figure 3 . Mean daily increase in tarsal length of \\'hitc-rumpcd Swiftlcts: the means for broods of one are 
represented by hollow circles, broods of two, horizontal bars and broods of three by solid circles. The 
consistent overlap between the standard errors (vertical lines) of chicks from all  three clutch sizes indicates 
the small effect food supply has on tarsal growth. 
Because average growth curves conceal certain characteristics of the individual 
growth curve and in particular the variation within a brood, the progress for a 
selection of individuals has been p lotted in Figures 4 and 5 .  It wil l  be noticed that 
even in the brood of three (and this was true for all other broods of three) death was 
not preceded by a drop in weight as in  the Common Swift in wet summers (Lack & 
Lack 1 95 1 )  and sometimes in the smaller sibl ings of the Edible-nest Swiftlet 
Aerodramus fuciphagus (Langham 1 980). 
Feeding rate 
The four ful l-days observation on a sample of the manipulated nests demonstrated 
that the average of 2 ? 9  ? 0? 34 visits 2 (x ? SE,  n = 1 1 )  to broods of three was 
significantly greater (Median test, X 1 = 4?97,  P < 0?05) than the 2 ?2 ? 0? 1  visits (n = 1 8) to broods of two. The 2 ? 1  ? 0?01 visits (n = 1 8) to broods of one was not 
significantly d ifferent (Median test) , X? = 0?8, n .s . )  from the number of trips to 
broods of two. The feeding rate per chick ( 1 ?0 feed each) in broods of three was not 
significantly different (t2i = 1 ? 1 )  to the feeding rate ( 1 ? 1  feeds each) per chick in 
broods of  two. 
During the day's observation in 1 981  a simultaneous collection of data was made 
on 20 undisturbed nests. These were not part of the manipulation experiment, and 
were so high as to be out of reach. An average of 2 ?8  ? 0 ?26  feeding visits were made to 
these nests. As these nests undoubtedly contained both one- and two-chick broods 
they can be compared to the combined average of 2 ? 1 7  ? 0?06 visits to man ipulated 
broods of one and two chicks . Because the difference is significant (t34 = 2 ?36 ,  
P < 0 ?05 ) ,  the possibi l i ty arises that my  presence was reducing the  number of feeding 
visits made by parents whose ch icks were being regularly handled .  Such an effect 
would be equal in  all sized clutches and so not affect the comparison between 
different-sized clutches. However, birds nesting at greater heights might be older, 
have closer feeding ranges and/or have some other benefit that will  increase their 
capaci ty to provide for their brood.  
1 6 8 
100 
80 
60 
E E 
"' c 
? 
40  
20  
10  20 30 
Days 
40 
I 
I 
1 /  
I /  
1 6 9  
Figure 4. Wing growth of\Vhite-rumpcd Swiftlct indi,?iduals from broods of all  sizes. Chicks from single 
broods are represented by continuous lines, those from broods of two by dotted lines, and those from 
broods of three by discontinuous Iincs. 
Conclusions 
Despite the apparently favourable weather for gathering food, the three chicks in 
artificially enlarged broods experienced significant delays in wing growth and weight 
increase. Such results are inconclusive because sur\'i\'al studies fol lowing brood 
manipulation of the Puffin Fratercula arctica (Harris 1 982) and the Blue Tit Parus 
caeruleus (Nur 1 984a,b) show that such a delay does not necessarily reduce the future 
survival of either chicks or parents. This should be particularly true in the Apodidae 
where chicks are adapted to withstand long periods without food.  
Because broods of three were fed significantly more than broods of two,  it  appears 
t hat the \Vhite-rumped Swiftlet of Fiji is not maximizing its han?est of the a\'ailable 
food supply during the breeding season . E\'en though the feeding rates to broods of 
two and three "vere not s ignificantly different ,  parents could  not  hatch significantly 
more chicks when given a third egg, or fledge significantly more ch icks when given a 
third chick. It is clear that parents are maximizing the number of fledgl ings thay can 
raise. The inability to rear three chicks may mean that food boluses are smaller when 
the feeding rate increases, thus prevent ing sign ificant ly  more chicks fledging fr?rn 
the larger broods.  
Although this swift let is  not maximizing its harvest of  the a\?ailable food supply, 
this does not mean that they are reduc ing annual p roduct ion in order to optimize 
their l ife-long production. However, the ell'o rt sm?ed by not \\'orking at i ts maximum 
1 1  
9 
7 
"" 
.:;:: 
"" ?o:; :;: 5 
3 
.... 
?'/ 
SHORT  CO:VL \1 l' 0: I CAT !Ol\'S  
/ 
....; ..
.. 
"'???. /:?:::;::;:::;::::::.
?
?::?:. 
1 0  
? 
\ __.. - - -..... "\':: ::. - - - -
20 30 
Days 
1 1 3  
40 
Figure 5 .  Weight increase in  White-rumped Swiftlet individuals from all  three sized broods. Chicks from 
single broods are represented by continuous lines, those from broods of two, by dotteq lines and those 
from broods of three by discontinuous lines. 
food?gathering capacity may help to explain why the \\'hite-rumped Swiftlet is a 
long-lived species for its size (Tarburton in prep . ) .  
The helpful advice and criticism gi,?en by  Dr Edward :\linot in  preparing this paper a s  part requirement 
for a Ph.D.? degree from Massey University is gratefully acknowledged. 
AsHMOLE, N.P. 1 96 1 .  The biology of certain terns. D .Phil .  thesis (Oxford):  cited by Lack & :\1oreau 1 965. 
AsH MOLE, N.P. 1 963. The regulation of numbers of  tropical oceanic birds.  Ibis  1 03b: 458-473. 
CHARNOV E.L. & KREBS, J .R. 1 974. On clutch-size and fitness. Ibis 1 1 6: 2 17-2 19.  
GADGIL, M. & BossERT, \\' .H. 1970. Life historical consequences of natural selection. Am. !\at.  104: 1-24 . . 
GooD MAN, D. 1 974. Natural selection and a cost cei ling on rcp?oductive effort. Am. !\at. I 08: 23-3 1 .  
HARRIS, M.P. 1 982. Seasonal variation i n  Aedging weight o f  the Puffin Fratercula arctica. Ibis 1 24: 1 00-
1 03. 
LACK, D.  1 947. The significance of clutch size. I bis 89: 302-352,  90: 25-45. 
LACK, D.  1 956. Swifts in  a tower. London: :\?lethuen. 
LACK, D. & LACK, E. 1 95 1 .  The breeding biology of the Swift Apus apus. I bis 93:  501-546. 
LANGHAM, !\. 1 980. Breeding biology of  the Edible-nest Swift let At?rodrtii/IIIS fuciplw?us. I his 1 22 :  447-
460. 
MURPHY, G. I. 1 968. Pattern in life history and the er1\'ironment. Am. !\at.  1 02:  39 1-403 .  
1\'VR, N.  1 984a. The consequences of brood size for breeding Blue Tits I .  Adult survival, weight change 
and the cost of  reproduction . .J. Anim. Ecol. SJ :  479-4%. 
NUR, N. 1 984b. The consequences of brood size for breeding Blue Tits I I. !\estling weight, offspring 
survival and optimal brood size. ] . Anirn. Ecol .  5 3 :  497- SI 7. 
1 7 0 
I I 4  SHORT COMMUK I CAT I ONS  I B I S  I 2 9  
PIAN'KA, E.A. & PARKEH, W.S.  1975 .  Age specific reproducti\'e tactics. Am. :--.:a t .  J O'J: +53?-464. 
TARBURTON, :\1 .K. 1 986. The food of the \\'hite-rumped Swiftlct (Aerodramus spodiopygius) in  Fij i .  
Notornis 33 :  1- 16 .  
\\'JLLIAMS, G.C.  1 966. :--.:atural selection, the costs of reproduction, and  a refinement of Lack's principle. 
Am. Nat. 1 00: 687-692. 
WJN'Kl.ER, D.\\'. & WAI.TEHS, j .R.  1 983 . The determination of  clutch size in prccocial hirds. Current 
Ornith. I :  33-68. 
Department of Botany and? Zoology, A1asse_v University , Palmerston North ,  New 
Zealand 
1 7 1  
APPEN D I X  5 
Bird Behaviour 6: 72-84 
A Comparison of the Flight Behaviour of the Wbite?rumped Swiftlet 
and the Welcome Swallow 
M. K.' Tarburton 
Department of Botany and Zoology, Massey University 
Palmerston North, New Zealand 
TARBURTON ,  M.K .  1 986. A comparison of the flight behaviour of fhc White-rumped Swiftlet and Welcome 
Swal low . Bird Behaviour 6: 72-84. 
A number of physical and behavioural characters are used to compare the flight of the White-rumpcd Swiftlct 
Aerodramus spodiopygius with that of the Welcome Swallow Hirwulo neoxena. Although the Swiftlct conserves 
energy by gliding more and having a slower wing beat .  it also reaches h igher speeds and alt itudes and spends more 
t ime in  flight each day than the Swallow. The publ ished morphological data on other members of  the swift and swal?
low families are compared to test the predictive value of two manoeuvrabil i ty indices (36) . l believe that the indi?
ces are invalidated by problems of al lometry and so cannot be used to compare birds of different sizes. An improved 
index formula is suggcst,?d .  
White-rumped Swi/ilet Welcome Swallow ,l,fanoeuvmbility cl llometrv Flight moul! 
lntrodm:tion 
Whitc-rumped Swiftlcts Aerodramus spudiopygius 
can tly cont inuously for up to 1 6  hours (pcrs . l)bscv . ) ,  
and must therefore b e  well  designed for flight .  B u t  for 
what type of flight are t hey best equipped? Because up 
to 30,000 White-rumpcd Swift le ts may l ive i n  one cave 
(64) , an abi l i ty to forage widely is probably important .  
Lack ( 4 1 )  suggested that  the wings of  the Apodinac arc 
adapted for high speed:  and certa in ly high speed flight 
would a l low a l arge popula tion to disperse rapid ly  to 
distant foraging are:.?s. But can wings adapted for high 
speed ,  also be adapll.'d for manoeuvrabi l i ty?) Cir?
cumstantial  evidence suggests that high speed and 
manoeuvrabil i ty can be combined . Hespenheidc (JX) 
bel ieved that the relat ive searcity of fl ies in the diets of 
birds i s  a result of  the fl ies? speed and agi l ity . White?
rumpcd Swi ft lets have a large proportion of !l ies in 
their  diet (60, 64) .  Thi s  fact and the smal ler s ize of  the 
Swift let  than a numbe r of swal low? (Table l )  suggests 
that Lack ( ?+ I )  was incorrect in postulat ing that swal?
lows arc bet ter designed than the smal ler  swif!s for 
catching t iny i nsects. One way of assessing the fl ight 
speed and manoeuvrabi l i t y  of  swifts and swal lows is to 
make interspecific morphnlogical and behavioural 
compa risons ,  both wi th in  and between the Apodinae 
and the Hirundines. 
Hai ls  (::15) has shown that  two mart ins ,  l'rogne subis 
and Delichon urbica, a swallow Hirundo rustica and 
the Common Swift Apus apus al l  have sim i lar  fl ight 
capacit ies i n  terms of encrgetics. When metabolic esti?
mates were made ( i n  cal .g . ? '  hr 1 ) ,  each species was 
shown to be, not only similar to one another but a lso 
simi lar"to the Golden Plover Pluvialis dominica, Rock 
Dove Columba livia. three l arids and two anat ids,  
though each wa:; more efficient than a number of  o ther  
birds. However, because the  energet ic cost of  fligh t per 
gram of  the birds weight, decreases with i ncreasing 
body weight , comparisons w:l l  only be val id  when 
either the metabol ic rate of  fl ight is compared with the 
rest ing metabolic rate of the same species or flight costs 
arc compared between species of s imi lar weight .  Either 
wav.  the Apodidae and the H irundincs have t he lowest 
flight costs of any birds measured so t ar (Table 2 ) .  
I n  order to fly ,  birds must expend energy in  t hree 
ways (54) :  tn generate l i f t .  to overcome d rag caused by 
nun-l ift ing parts (parasit ic drag) . and to overcome the 
drag c reated by movement of  the wings. The H i run?
d ines and Apodincs have morphological characteristics 
that tend to reduce the energy required for each of 
these act ivit ies (::15) .  Firstly , these b irds have re lat ive ly  
long wings and weigh l i t tle compared with their  wing 
arc:1 ( i . e .  they have a low loading) ; thus they generate 
i ift more easily than do birds of similar size . Secondly ,  
the stream l ined head,  short neck and tapered abdo-
1 7 2  
1 7 3 
Fiight Behaviour of a Swiftlet and a Swallow 73 
TABLE l 
MANOEUVRABILITY INDICES OF SOME SPECIES OF 
APODIDAE HEMIPROCNIDAE AND 1-!IRUNDlNIDAE 
Species Wing Tail Weight Sampk Wing Tail 
mm Outermost g Size Index Index References 
Rectrix 
Whitc-rumped Swiftlet 
Arrodramus spodiopygius 1 1 2 .0  44.7 8 . 1 9  1 02 1 3.68 5.46 A 
White-bellied (Glossy) Swiftlet 
Collocalia esculenta 102.0 40.4 B.2S 1 2.32 4.87 36 
Mossy-nest Swiftlet 
Aerodamus vanikorensis 124.0 53.8' 10.99 37 1 1 .28 4.90 3,58,69,8 
Neotropical Palm Swift 
Reinarda squamara 1 12 .5  69.5 . !0.36 10.86 6 .7 1  c 
Edible-nest Swiftlet 
Aerodramus fuciphagus 1 1 6.0 50.6 10.80 10.74 4.69 44,63 
Old World Palm Swift 
Cypsiurus parvus !29.2 50.6 10.80 421 9.4 7.79 12 
Wire-tailed Swallow 
Hirundo smithi 1 1 2.0 M 74 1 2.7 8.82 5.83 
F 5 1 . 5  r2.1 6 8.82 4 . 1  26,48 
Eastern Rough-winged Swallow 
Psalidoprocne orientalis 104.5 83.0 1 1 .9 3 8.78 6.97 47 
Black-nest Swiftlet 
Aerodrantus maximus Ll\ .9  47.9 1 5.9 21 8.30 3 .0 1  37,63 
Boehm's Swift 
Neafrapus (A-ft--arnsia) boehmi 1 2 1 . 7  !5 .3  7.95 26,69 
Wekome Swallow 
Hirundo neoxena 1 10.0 67.5 1 3.92 6 7.90 4.85 A 
Sabines Spinetail 
l?haphidura sabini 120.9 16.6 79 7.28 1 0  
Bank Swallow 
Riparia riparia 105.5 4H.O 14 .6 249 7.23 3 . 29 46,62 
Band-rumpcd Swift 
Chnctura spinicauda 102.6 1 ?1 .2 1 9  7.23 61 
Pacific Swallow 
l!inmdo tahitica Malaysia 14 .75 35 n.9R }.OH 36 
Fiji 109 . 1  J.6.(} 17 .H  1 1  6. 1 3  2.62 A 
Mangrove Swallow 
Tachycineta albinia 99.0 4 1 . 3  !4.5  6.83 2.85 57,C 
Short-tailed Swift 
Chaetura brachyura 122 .0  3 1 .0' 1B .J 240 6.67 1 .69 18 
Grey-rumpcd Swift 
Chaetura cinereil-?entns 1 0 1 .5 15 .71  53 6.46 18,6i 
Lesser-striped Swallow 
llirundo (Cecropis) ab_vssinica 1 10 .7  97.0 17.2  6. 44 5 .6-1 26,47 
Caffer Swift 
Apus caffer 1 4 1 .8 74.0 22. 1  229 6.42 3.35 9,26,47 
Barn Swt.tl!ow 
1-/irundo rustica 125.0 87.0 19 .5 10  6 .4 1  4.56 47 
Vaux Swift 
Chaetura vmui 1 15 .0  18.  i l !74 6.35 2 , 1 9  
Chestnut-collared Swift 
Cypseloides rwilus 1 24.0 40.0' 20.2 45 6. 1 4  1 .98 18 
Fork-tailed Swift 
Apus pacificus 180.0 80.0 29.6' 6.08 1 .75 32,56 
House Martin 
Dcllchon urhica 1 1 2 .0  58.5 19 .8 5.66 2.95 47 
Z: L T  
74 Tarburton 
TABLE 1 (continued) 
------- ?-- ?--- ---?------? 
Spectcs Wing Tail 
mm Outcnno:-;t " ? 
Horus Swift 
A pus horus 152.2  58.0 27.37 
Chtmncy Swift 
Chactura pl'!agica 1 28.0 }7.6 23.6 
Pallid S\\ift 
Arms pallidus 170. 1 36.7 
House Sv.?ift 
A pus affinis Asia 
Africa 1 3 1  () 4-\5 27.2 
White-throated Swift 
At'rO!Iaures saxatalis F l 3S 5 55.7 32.3 
M ! -!04 56.H .14 .2 
Common S\'.ift 
Apu.s apus 17:1.0 72.0 42.7 
Afncan Bearded Swift 
A pus barhatu.\ 1 72.0 7-1 .5  42 .2R 
Bradiit.?[d's Sv,:ift 
A pus bradfieldi 1 7 1 .7 o3.3 43.0 
Grcy-brca<;tcd Martin 
Progne chalrhea 1 3 1 .0 (l LS 33.5 
Black Swift 
( ?_vpseloidcs mgcr 1 70.0 57.5 ?15 .6 
?iosque Swallnw 
Cccropu: .senegalensis 1?13.5 95"' 40.7 
San Cicrnnimo Swift 
l'anyprtla .wmtihero11yrm l K.\7 K6.4 5fl.65 
Chapin:. Spinctatl 
( "harwra melmwpygw 167 ?l -t9.4 52 0 
Mottk?d Swtlt  
Apln equaroriali.\ 20.1 1 93.0R 
Alpt!il' S\vdt 
A pus nu.:! ha S. W. Angola 1 99.K 76.0 
Spain 222.6 li7.0 
S.\V. Afnca 202 6 o5.5  9 1 .0 
RU\'\1:1 ?20.0 102.5 
Uganda 22(, ,2  127.6 
\Vhitc? .. :ollared Swift 
Strt'fHOprocnt' ::.onaris 197,0 118. 1 
\Vhitc?thro41tcd Necdlctail 
( '!wl'ftml t'llwiacwa F 200.0 88.0+ 1 1 .1.7  
?! ::I?UJ XX.O+ i22 .2 
Swift 
Str!'{Jhlf'Ti)C!lc \'t'ntico!/,Jn?; 22X.O S L 7  C l 75.0 
Phthppm?..? Nct'dktail 
( 'hacwra n.:lt'henst.\ 2 i q  0 6-1 0 1 79.6 
??--?-------?- ---????------ ----- ----?? -?--?----------?-
Ba-:cd nn mc.u\ or mode and therc!(Hl' probably an undc1 or t)\'l"fC"t!ill,ltC 
A rlll" paper 
B /\m ... ?ncan Mu:-.cum of ?:amral History 
c Ln? AngL'h:? ( 'oullt) Mu!-eum lH N<ltural I fistory 
1 7 4  
Weight Sample Tail 
Size Index index References 
140 5.56 2. 1 2  9 . 1 8,47 
1805 5 .42 1 .59 27,30,C 
4 .63 52 
4.29 1 . 54 35 
4.R2 1 .64 17 ,47 
??  4. 2'1 1 .72 4 
22 4. 1 1  1 .66 
35 4 . 1 2  1 .70 43.47 
S9 .1.()7 1 .76 
82 4.00 1 .47 :1.6, 
3 9 1  1 .84 18,7 1 
i 2  3.73 1 .28 c 
3.53 2.33 26,47 
9 :-; 24 153 l , l 6,5'l.C 
3.22 0.95 -15 
2K 2 . 1 X  0.94 7, 1 ?  
20 2.63 R 
1 0  2.56 52 
1 5  2.23 0.'l4 X 
2X 2. ! 5  
1'1 1 .77 x 
i 9  2 . 0 1  19,7 1 
l .7h 0 77 5 1  
1 1  1 .6-l 0.72 5 1  
1 .30 0.-!7 I R.C 
00 L""' 0.36 50 
Specie' 
Soot; Tern 
1vial lard 
American B lack Duck 
Commou Swift 
Great Black-hacked Gul l  
Ring Bi l led Gul l  
House Martin 
Laughing Gull 
Golden Plover 
Pigeon 
Barn Swallow 
Purple Mart i n  
Budgerigar 
B luc-throated Bee-eater 
Loggcrhcad Shrike 
Pacific Swallow 
Common Roscfinch 
Bul lfinch 
Chaffi nch 
Brambling 
Costa's Hummingbird 
Purple Carib Hummingbird 
Siskin 
Gli ttering-throatcd 
Emerald Humminghird 
Savannah Sparrows 
TABLE 2 
ESTIMATES OF M ET A B OLISM D U R IN G  FLIGHT FO R 
A V A R I ETY O F  AVIAN SPECI ES 
S1ema ji1scaw 
Anas plmyrhynchos 
Arws ruhripes 
A pus apus 
Larus marinus 
Larus delawarcnsis 
De/ichon urbica 
Deliclwn urbica 
Delichon urbica 
Larus arrrici/la 
Pluvialis dominica 
Columba lil'ia 
Hirundo rusuca 
Hinmdo msrica 
Progne subis 
lvlelopsillacus wzdulaws 
M crops viridis 
Lanius ludo>?icianus 
Himndo whitica 
Carpodacus erYfhrinus 
Pyrrizula pyrrhula 
Fringilla coelcbs 
Fringil!a coe!ebs 
Fringilla monrifringiila 
Calypte costae 
Eulampis jugularis 
Carduelis spin us 
Amazilia fimbriala 
Passcrculus sandwichensis 
Body Weight 
(g) 
1 X7 .0  
1 000.0 
1 026.0 
40 . 5  
800 0 
4 10 .0 
1 9 .0  
20 .5  
1 7 .8 1  
350.0 
l 40 . fl 
384.0  
18 .99 
1 7 . 8  
5LO  
35.0 
33. 8  
48.6  
!4 . 1 
60. 0  
29.5  
22. 5  
22.0 
23 .3 
3.0 
8.3  
1 2 .5 
5 .5  
! 9 . 1 
Metabolism 
(J g?' hr?') 
92 
!52  
! 52 
! 56 
156 
1 80 
1 68}  
1 80}  18 1  
194) 
1 88 
220 
224 
236 
259 
264 
388 
421 
488 
581  
640 
654 
608} 656 
704} 
630 
792 
820 
820 
824 
992 
------ ------ - --
Source 
Flint & Nagy l 9!H 
Dolnik & Gavrilov 1 973 
Bcrger el a/ 1970 
Lyulecva 1 970 
Dolnik & Gavrilov 1 973 
Bcrger er a/ 1 970 
Keskpaik 1 968 
Lyulceva 1970 
Hails 1 979 
Tucker 1 972 
Johnston & McFarlane 1 967 
Lcfebvre 1 970 
Hails 1 979 
Lyuleevva 1970 
Utter & Lefebvre 1970 
Tucker 1 968 
B ryant er a/ I 984 
Weathers e1 a/ 1984 
B ryant et a/ I 984 
Berger et a/ 1 970 
Dolnik & Gavri1ov 1 973 
Dolnik & Gavri1ov 1973 
Dolnik & Gavrilov 1 973 
Dolnik & Gavrilov 1973 
Lasiewski 1 963 
Hainsworth & Wolf 1 969 
Dolnik & Gavrilov 1 973 
Berger & Hart 1 974 
Williams & Nagy 1 984 
::!1 
ao? 
;:r 
ttJ 
<1> 
:::r 
P> 
< 
a? 
c 
'"1 
0 
,_, 
P> 
(/) 
:;: 
? 
[ 
5 
0. 
CO 
en 
? 
2:. 
0 ::;:: 
---1 
V\ 
t-' 
"-.) 
V1 
1 7 6  
76 Tarburton 
men and tail reduce the energy required to overcome 
parasitic drag. Last ly .  the sickle-shaped wings and high 
wing a<;pect ratio (wing span to wing area) of Hirun?
clines , Hemiprocnidae (Crested or  Tree Swifts) and 
Apodidae results in reduced profile drag. This last 
characteristic is unexpected as it would seem logical for 
l arger than average wings to create larger than average 
wing drag when in motion. 
I f  these characteristics can be used to demonstrate 
the efficiency of swal lows, swifts and trce-swifts over 
other birds,  they can also be used for comparisons 
within and between the three orders. 
In the second assumption we have the naive expecta?
tior. that a measure of length is l inearly related to 
weight or  volume. For example , when the data for 
wing length and weight from Table 1 arc graphcd as in 
Figure 1 it is clear that a straight line does not provide a 
good  fit .  Thus the wing index used by Hails and Amir?
iudin (36) is very dependent on the weight of the bird 
25 
-..... 
e: ?20 .. 
..c +-
en .. 
c 
OJ 
-
.?15 0 
3 
.. .. 
f 
, .. 0 
10 ?'I 
0 50 
Weigh t 
(Figure 2) .  However, as with most al lometric problems 
the relationship between weight and wing length is bet?? 
tcr described by a power function i . e . ,  an equation of 
the form y = a Xh . The curve in Figure 1 is of this form 
and is the least squares fit to the power function , y = 
5 5 . 0  X" 29 (H2 = 0 . 9 3 ;  df = 50; P<0. 00 1 ) .  An index of  
manoeuvrability based on wing length might employ 
this power function and could take the  form of wing 
l ength I wing length predicted from weight.  Predicted 
wing length equals 54.95 (weight)''2" based on  the data 
in Table 1 .  This could be improved to give symmetry 
about the mean by taking the log of this  ratio i . e . ,  wing 
index = lg (wing length I predicted wing length) .  This 
new index may be used to compare all members of the 
Apodidae and Hirundinidae regardless of their weight 
(Figure 3). The new index in e ffect a ll ows an examina?
tion of deviations from the general relationship bet?
ween wing length and weight for a number of species. 
I t  is worthwhil e  noting that almost all of  the Hirun?
dines are found in  the lower left corner of  the graph in?
dicating that after adjustment for al lometry . hirundines 
have a smaller wing than apodincs. 
,. 
" .. 
.. 0 
1 _.,. ... 
100 150 
(g) 
Figure 1 .  The relationship between wing. length ,md weight in those swifts and swallows given in Table I .  
1 7 7  
12 
? 8  
"0 c 
en c 
3 
4 
t : 
.. 
.... 
00 
.. 
.. 
Flight Behaviour of a Swiftlet and a Swallow 
" 
?? 0 
(I 
50 
G 
' & 0 
.. 0 " 
100 150 
\aJeight (g) 
.. .. 
Figm?e 2. The relationship between the wing index used by Hails and Amirrud in  (36) and the weight of the swifts and 
swallows giwn i n  Table 1 .  
" 
0 ?2  
" 
.. ? "' 
0 ? 1  0 " X 0 " QJ 1/J , ? .. " "0 " " 
c ,. 
.. .t 
en 1\ 
c 0 0 
3 .. 0 0 
r:P " 3 " "  ? " QJ .. .. z -0?1  00 0 .. 
0 0 0 
00 
0 50 100 150 
Weight (g) 
Figure 3. The new wing i ndex plotted against the weight of the swifts and swallows given i n  Table 1 .  
77 
1 7 8 
78 Tarburton 
TABLE 3 
FLIGHT BEHAVIOUR 
White-rumped Swiftlet 
Circuit Feeding 
High Movement 
Welcome Swallow 
Circuit Feeding 
High Movement 
Wing Beats 
per second 
X ?  S .E .  
4 .7  ? 0 .20 
4 .7  ? 0. 1 4  
5 . 7 1 ? 0. 1 4  
5 .79 ? 0 .29 
n 
33 
50 
3 1  
16  
Other Flight Behaviour. During feeding flights glid?
ing is undoubtably important, for as Pennycuick (55) 
points out, i t  is  achieved at  a much lower cost than 
Flapping flight .  Hai ls (35) has calculated that the swifts 
and house martins ,  which ghde the most, have a flight 
cost saving of 72% ; whereas swallows , which glide the 
least, have only  a 62% saving over other birds of simi?
lar size. I have tested these findings in the field by com?
paring the flight behaviour of  the White-rumped 
Swift! et  with that of the Welcome Swallow. The results 
are shown in Table 3. 
The average length of time in each glide for the 
Swallow ( 1 .5 s) and the Swiftlet ( 1 .2 s )  was not signific?
antly different (t = 1 .94;  df=63; P<0.05 ) .  However, 
s ince the time spent !lapping by Swallow?s ( 17 .6  s) is 
considerably longer than that spent flapping by 
Swiftlets ( 1 . 8 s), i t  is  clear that the Swiftlct glides more 
than the Swallow when feeding. Two other factor also 
show that the Swiftlct conserves energy better than 
does the Swallow. Firstly, feeding Swallows have a sig?
n ificantly faster wing-beat rate (0.57 s)  than the feeding 
Swiflets (0.47 s) ( t  = 3 .8 :  df  = 34; P<O.OOl ) .  Secondly 
Swallows rest from flying by perching for an  average o f  
3 . 0  min  after an average flight time o f  2.45 m in ,  com?
pared with the Swiftlet ,  which does not perch or rest at 
al l  through the day ( in the non-breeding season) .  Even 
in the breeding season parents take on ly one to three 
rests (lasting one to 20 m in) at their nest.  
Methods 
Comparisons between the vVhitc-rumped Swiftlet 
and Welcome Swallow were based largely upon two 
manoeuvrabil i ty i ndices used by Hails and Amirrudin 
(36) to compare the White-bell ied Swiftlet Collocalia 
esculema, House  Swift Apus affinis and Pacific Swallow 
Length of Glide Length of Wing-beat 
(seconds) Sequence (seconds) 
X ? S.E.  n X ? S.E. n 
1 . 1 7 ?0.09 35 1 .84 ? 0. 1 6  35 
10.36 ?2. 1 45 4:40 ? 0.70 45 
l . 5 l  ? 0. 1 5  30 l7 .55 ? l .S 44 
1 .0 6 39.84 + 6 
Hirwulo tahitica. These authors use the long wings and 
tail  as well as the l ight weight of the White-bellied 
Swiftlet to explain not only its reduced energetic costs 
but also its greater manoeuvrability, wh ich  ai lows the 
Swiftlet to forage much closer to vegetation than other 
swifts. The first manoeuvrability index (tai l  index) is 
the ratio of the outer ( longest) rcctrix l ength to body 
weight and is attributed by Hails and Amirrudin (36) to  
Waugh (68) and Bryant and Westerterp ( 14) .  The 
second index (wing i ndex) is  the rat io  of  the wing 
length to  weight. ror both indices the h igher the  index 
the greater the manoeuvrability.  
For this study indices were derived from measure?
ments of live White-rumped Swiftlcts caught in Fij i ,  of 
skins of the Pacific Swallow in the National Museum of 
New Zealand and of l ive Welcome Swallows caught at 
Longburn , New Zealand. All other measurements 
were taken from published data or from mu?eums as 
shown in Table 1 .  Those tail measurements marked 
with an asterisk were based on the mean tail length and 
so slightly underestimate the length of  the outer rcctrix 
in species with forked tails and overestimate the length 
of the outer rectrix in species with only s l ightly bifur?
cated tails. 
Other features of flight that were measured were ( i )  
average length of  t ime  of a sequence o f  wing  beating , 
( i i )  number of beats per flapping sequence,  and (i i i ) 
average length of a glide. Single beats were ignorect ht'?
causc they appeared to be used for manoeuvring 
rather than propulsion, and also because they were 
below my reaction time to record with a (Casio C-80) 
stopwatch . All three measures were determined for 
feeding flights and travel fight s .  During feeding flights ,  
the  birds flew in fcedmg circuits that  approximated cir?
cles with a diameter range of about four to 50m . Dur?
ing travel flight ,  the birds usually flew at  greater 
1 7 9 
Flight Behaviour o f  a Swi ft le l  and a Swallow 79 
alt i tudes (30- 1 30 m for the Swiftlcts, 10-40 m for the  
Swallows) and were moving to o r  from the i r  feeding 
areas. Alth ough the Swift let  travel flights were mea?
sured close to the Nasinu caves,  15  km north of Suva, 
i n  Fij i , feeding c ircuits en route to feeding areas were 
not uncommon and these were included in the data for 
travel fl ight .  
Results and Discussion 
Manoeuvrability in Flight. I n  applying the manoeuv?
rabi l i ty i ndices to 40 add i tional species of H i rundincs 
and Apodines , I have found that t here i s  considerable 
variation (Table 1 ) .  Hails and Ami rrudin 's (36) asser?
t ion of h igh manoeuvrab i l i ty in the Whi te-bel l ied 
Swiftlet is  substantiated to the extent that  as measured 
by the ta i l  i ndex, the ir  manoeuvrab i l i ty is  h igher than 
that  of  most  swifts. However, the same index shows 
that  the Wh ite-rumpcd Swiftlct is even more man?
oeuvra ble .  while the wing i ndex indicates greater 
manoeuvrabi l i t ics among four swiftle ts .  Thus it is  no  
longer possible to claim that the Whi te-bell ied 
Swift let "s manoeu\Tabi ! i ty is except ional in  the  
Apodidae . 
For the wing index the Whire-rumped Swiftlct 
had the h ighest manoeuvrabi l i ty rating, but  for the tai l  
i ndex two palm swifts and three  swallows were given 
the h ighest rat ing .  This inconsistency between the two 
ind ices leads one to q tH:st ion the i r  val id i ty .  !n using the  
1ndices we make two a?sumptions, both of  which arc 
possible sources for error. The first i s  that  each i ndex 
actua l ly measures manoeuvrabi l i tv ,  and t he second is 
th: 1 t  each index overcome-; the a l lometric problem of 
comparing  species of d i ffe rent weights .  
One way of  test ing the first assumpt ion would be to 
compare the manoeuvrabi l i ty i ndices wi th  the propor?
t ion of  prey in the diet  that  is  d i fficult to catch. The d i f?
ficulty o f  capturing each type of prey could be deter?
m ined by comparing the proportion o f  each type in the  
diet  w i th  i ts avai labi l i ty where and when the b i rd for?
ages . This procedure has been considered e lsewhere 
(64 ) ,  where it was shown that fl ies,  wh ich arc said to 
have high 111anoeuvrabi l i ty (31S) , arc equal ly l ike ly to 
dominate in sampled diets nf b i rds wi th  high or low 
manoeuvrabi l i ty  indices. However. to determine 
whether species wi th  low ind ices take only those fly 
species tha t  lack manoeuvrabi l ity wi l l  requ i re further  
study. 
During travel fligh t ,  d i fferences between species 
were even greater.  The Swift lct  beats 4.7 t imes a second 
and the Swal low 5.1S t imes a second.  The Swift let glides 
for much longer ( 10 .4s)  than the Swallow (l .Os ) ,  but 
beats for much shorter periods (4.4s) than the Swallow 
(39.8s) . A Swift let  that  started flapp ing  overhead 
rarely tlew out o f  sight i n  the same flapping sequence, 
whereas Swal lows in  travel fligh t usually d id so. 
Visual d i fferences i n  the fl ight of  the two species 
were obvious. Wctrnorc (70) noted that  swifts generally 
fly higher than the smaller swallows. The Whi tc?
rumped Swift lct ,  which is smal ler in size to both the 
Pacific flirundo tahitica and Welcome Swallow, feeds 
from 10 cm to 1 30 m above the ground.  This is consi?
derably h igher than t he  a l titudes a t  which t he  two swal?
lows feed (from nearly zero over water to 30m) .  
A second d ifference i s  that  when glid ing both swal?
lows hold the ir  wings sl ightly closed and close to the  
body, whereas Swiftlcts hold the i r  wings well away 
from t he  body result ing in  the character ist ic sickle?
shaped curve. Mayr ( 48) used this character to separate 
swift lets from swal lows in the south-west Pacific. The 
sickle-shaped curve of the Swift lct is  accen tu ated by 
the pointed wing t ips where the first primary is 8mm 
shorter  than the second primary as well a s  having an 
cmarginatcd inner web . Lack ( 42) suggested that the 
shori first primary (which he found to be 5mm shorter 
than the second in  other species of swiftlcts) and the 
wing s lot  (caused by the feather cmarginat ion)  he lp 
give greater contro l  i n  s low fl ight .  
A th ird d ifference is the ampl i tude of th<: wing beat . 
Both the Pacific and Welcome Swallow a !mo?t always 
us,; "large ampl i tude strokes.  whereas the Swift let only 
uses such stt okcs in tight manoeuvres and slow fl igh t .  
especially in  cav,?s .  
A fourth d ifference occurs when feeding fl ight is in a 
straight l ine :  both swallows exhib i t  a regular  undula?
t ion but the Swift le t  does not . Rather .  swiftlets change 
alt i tude i rrcgulariy,  sometimes pul l ing up to a halt to 
catch prey and then divi ng down again  to gain 
airspeed .  The tai l  is often ful ly spread in t h is activity 
and i ts considerably larger area than that  of the swal?
lows' forked ta i l  must aid in performing the manoeuvre . 
Fifthly, deviat ion to the side is infrequent and 
gradual in both swal lows . but not the Swift lc t . A glid?
ing Swift lct may quickly drop one wing to swerve side?
ways ; ur i f  the i nsect prey is further to the s ide the b ird 
may rapidly twist the body and tai l  to bring the 
momentari ly gaping mouth in  i ine with the prey.  
A further d ifference i s  that  only Swiftlets fly i n  total  
darkness. Audible cl icks arc used to guide their way in 
the dark . 
1 8 0  
80 Tarburton 
B o t h  species of swal lows and the S w i ftlet sometimes 
p u t  o n  a burst of rapid wing beat i n g  to alter alt i tude i n  
pursui t  o f  p r e y .  At t i m es the Swift let w i l l  turn such 
s eq u e n ce s  into im pressive power dives, re m i n iscen t  of 
the l a rger swifts. By contrast , descriptions o f  the feed ?
i n g  flight o f  some swift species include comments o n  
t h e i r  l a c k  o f  manoeuvra b i l i ty. For exam p l e ,  W hite?
t h ro ated Spinetai ls feed i n g  o n  a grasshopper swarm i n  
Queensland ( 1 5) were never seen t o  make a second 
attem p t  t o  catch the same grasshopper for the swift 
would rather select another target 'more or less on t he 
bird ' s  l i n e  of tl ight' a n d  P h i l i pp i n e  Spinetai ls feeding 
on bees, near their hives (50) had such 'straight' fl ight 
that t h e  local people regularly caught t h e  birds b y  
swingint; pole-mounted fish nets i n t o  their flight p a t h s .  
A n o t h e r  problem f o r  aeria l  feeding b i rds i s  heat pro?
duction. Ricklefs (57) found that on sunny afternoons 
the Three Swallow lridoprocne hicolor only siightly re?
duced its foraging i n  its tem perature habitat; whereas 
the closely re lated, tropical Mangrove Swal low 
Tachycineta albilinea, drastical ly reduced its m idday 
foraging regularl y . Ricklcfs ( 57) demonstrated t h a t  the 
rad i a t ional heat absorbed by a bird i n  the m i d dle of the 
day approached that generated b y  i ts metabolism 
d u ri n g  fl ight and thus caused problems o f  regu la t i ng its 
temperature . That there i s  reduction in i nsect abun?
dance and size around m idday (38) could also contribute 
t o  t h e  M angrove Swallow's relative inactiv i t y ;  but 
R i c k l c fs (57) noted that sympatric flycatchers feed 
most frequently d u ri n g  the m iddle of the cl a y .  S i n ce the 
Whitc-rumped Swiftlet does not stop fl ying i n  t h e  heal 
o f  t h e  day , i t 's  greater flight efficiency presumably not 
o n l y  a l lows i t  to catch h ighly manoucvrahle pre y ,  un?
obtain able by most other aerial  predators. but also 
a l l ow s  i t  t o  d o  so right through t h e  heat o f  t h e  tropical 
day.  
A nother aspect of fl ight efficiency i n  t h e  Apod idac 
can b e  seen in B ryant's ( 1 3 )  observations o n  the habit  
o f  leg tra i l i ng d u ri ng fl ight i n  17 M a laysian bird s .  H e  
showed t h a t  the Pacific Swallow a n d  Rcd-rumped 
Swal l ow Hirundo striolata both i n crease their a m o u n t  
o f  leg t r a i l i n g  around m i dday, evidently for t h e  pur?
pose of dissipa ting excess heat . Two Apod i d a c ,  the 
White-vented Ncedletail  I-lirundapus cochinchincnsis 
and H ouse Swift Apus ajjlnis a n d  one l lc m i procn i d ,  
t h e  Crested Swift flcmiprocnc longipcnnis, d i d  not 
show an i ncrease i n  leg trailing during t h e  hottest part 
of t h e  d a y .  
S e veral physiological processes have a l s o  b e e n  shown 
to help fac i l i tate e ffi c i e n t  fl ight i n  the A pndidae . For 
example , the Commo n ,  A l p i n e  and Pall id Swifts have 
l arger erythrocytes with greater oxygen affinity than 
passc rincs (52) .  The i r  haematological values and ra t i os 
of heart to body size arc equal to those o f  montanc 
birds l i ving a t  alt i tudes above 2500m.  B e cause the 
breast m uscles of the swifts are smaller ( 1 5 . 6  to 1 9 . 7% 
o f  body weight) than those o f  t h e  Pigeon Calumba 
livia ( 26.5%  ) , they make more effi c i e n t  use of t h e  
hacmatological adaptations. This effi c i e n c y  h a s  been 
dcmonstrakd by comparing the energy used in fl ight 
by the Common Swift , four H irundincs and four other 
birds of similar weight (Table 2). The S w i ft used 39 .0  
ca i . g? '  hr?'  which i s  5 .4  times i t s  average m etabolic rate 
while roosting (Sleeping Metabolic Rat e ) ,  whereas t h e  
other birds averaged 1 52 ell .  g ? '  hr _ ,  w h i ch is 12  t imes 
their S M R .  
Alt hough these physical a n d  physiological features 
work together t o  reduce the e n e rget i c  needs o f  t h e  
Whitc-rumped Swiftlet t h e i r  habit  o f  fly i n g  l o nger t h a n  
the d i u rn a l  perio d ,  without rests, m e a n s  t h a t  i n  a day's 
flying they probably d o  not save as m u c h  energy i n  a 
clay when compared with t h e  H i ru n d i n e s .  The gross 
daily energy consumption of the Common Swift i s  30.7 
kcal , whereas t h a t  o f  the H ouse M a r t i n  a n d  B a r n  Swal?
low is J 7.0 kcal and 1 7 . 4  kcal per day (25 ) .  When t h ese 
energy expenditures arc corrected for d i fferences i n  
body size , b y  dividing them by t h e  body weight of each 
bird (Table 1 ) ,  the comparisons s t i l l  show an advantage 
to the Swift .  The Common Swi ft consu m e s  0.72 kcal g '  
cl ' .  the H ouse Martin 0.86 kcal g? 'd? ' ,  a n d  the Barn 
Swal low O.R9 kcal g ' d ? ' . H owever, some o f  this advan?
tage will result from the reduced thermoregulatory 
costs of t h e  Swift due t o  its larger body size than t h e  
t w o  H i ru n d i n c s .  
? S'pecd of Flight. The flight speeds attainable by swifts 
are both breathtaking and legendary. The Needlctailcd 
Swift Chaetura caudacuta has been described as o n e  of 
the fastest-flying birds in t h e  world , fee d i n g  and cruis?
ing at 50 to 1 30 km/h ( 3 1  ). The same b i rd has been de:;;?
cribed as the fastest bird in the Soviet U n io n  with level 
fl ight speeds o f  up to 1 70 km/h (29 ) .  H oweve r ,  two 
other swifts have also been descriticd as t h e  world's fas?
test birds:  the Brown-throatcd N c e d l e t a i l  Clwl!lllra 
gigantea with reputed speeds of up t o  200 mph (33) imd 
the P h i l ippine Spinctai l  Swift Chaellim cl!lehensis(22) 
with re puted speeds o f  well  over 1 00 m p h .  The fl ight 
speed of t h e  Whi tc-rumpcd Swift let can also be i m pres?
sive , particularly when they are e n tering or departing 
through the open i ng o f  the caves used for roosting and 
nesting. In Fij i this Swift lct  usually i n c reases its speed 
as it fl ies t h rough the cave e n t rance . Why these bursts 
o f  speed were made at the e n t rance puzzled m e  u n t i l  I 
realized that the cave entrance was where Barn Owls 
1 8 1  
Fli!!ht Behaviour of a Swiftlet and a Swallow 8 1  
Tyto alba lulu sometimes preyed upon the Swiftlets. 
The Owls appan:ntly still find the concentration of 
swift lets rewarding enough to periodically hunt at cave 
entrances. Similar behaviour has been reported for a 
small colony (24 nests) of Mottled Swifts Apus 
aequatorialis in Rhodesia (2 1 )  that occupied a cave 
whose entrance was patrolled by a pair of nesting fal?
cons. 
The entrance to the Wate1fall Cave at Nasinu was 
suitable for marking off a 1 Om horizontal section 
through which White-rumped Swiftlets could be timed 
and 92 birds were timed through this section .  The 
slowest took 1 . 5 1  s and the fastest 0 .34 s .  The average 
speed was -15 km/h ; the slowest 24 kmih ; and the fastest 
I 06 km/h . Most birds appeared to fly even more 
quickly in their dives towards the cave entrance , but 
the difficulty of timing them there was not overcome. 
Six Welcome Swallows timed across 1 19 m of pad?
dock on their way to their evening roost ranged in 
speed from 41 to 66 km/h. Their average speed of 
53 .6 km/h was greater than the average flight speed of 
the Swiftlets. The maximum SwaHow speed, however, 
which I estimated to be  close to their maximum, was 
well below the maximum speed of the Swiftlets. This 
suggested many Swiftlets fly well below their maximum 
speed en?n when flying through predator patrolled 
areas. 
My data (including Table 4) for moult in  the White?
rumped Swiftlet is similar to that of the Black-nest 
Swiftlet except that it starts earlier, in September or 
October instead of November or December. Moult in 
the primaries is accompanied by a burst of nest build?
ing and probably by enlargement of gonads as four out 
of six moulting birds found dead or accidental ly kil led 
were found to have en larged gonads. The moult i s  
reasonably synchronized, between individuals, sym?
metrical between right and left wings, and centrifugal 
(starting with the innermost or first primary). How?
ever, as Table 4 shows, 62% had not  commenced 
moulting by December and 36% were sti l l  moulting 
their last (tenth) primary in  June .  No birds were found 
moulting primaries in  August. The designated primary 
in the table was either missing or was the ?mallest re?
generating feather in the tract . Sometimes two feathers 
were regenerating simultaneously (the outermost 
always being the least mature), but in  only one bird 
were three growing at the one time. Wing coverts were 
found to be moulting in December, April and June . 
Tail feathers were found to be moulting in June , and 
head feathers in  September. 
Few tropical birds moult and breed at the same time 
though some show considerable overlap.  Payne (53) 
looked :J.t 1 ,050 specimens of tropical African birds and 
found that only 13% of 1 90 species showed overlap or 
synchronous moult and gonadal or breeding activity. 
The only other moult data for swiftless that I can find 
in  the literature relates to the Mossy-nest Swiftlet from 
Sumatra, where Wells (69) found two of five specimens 
were moulting their tenth primary. Although these 
birds had regressed gonads, a few nests contained eggs, 
hence synchrony is ruled out, though overlap would 
still be possible . 
The martin genus Progne shows similar variation to 
Aerodramus. The tropical . northern and southern 
species all breed and moult at different  times of the 
TABLE 4 
MOULT IN THE PRIMARIES 
Aug Oct Dec Pcb Apr Jun  
Number Examined 1 30 1 3  65 55 44 50 
Number not in  moult 1 30 1 3  ..j() 33 22 32 
Number in moult 
P I  ..j 
P2 1 0  
P.\ 3 
p..j 5 
1'5 2 
P6 5 
P7 6 
PH 3 
P9 5 5 
PlO  1 6  l R  
--- -- -?" 
1 8 2  
82 Tarburton 
year. The exception , P. l'vf. Modest a, is endemic to  the 
Galapagos I slands wh..:re i t  breeds while actively 
moult ing (28) .  
Sleeping on the Wing. The Common Swift ha?  been 
recorded as sleeping on the wing ( 4 1  ) . The evidence in?
cludes an observation at n ight by a pilot of a group of 
Swifts and observat ions of groups heading out to sea at 
nightfa l l .  Snow (6 1 )  suggests that  the Whik??co! lared or 
Cloud Swift Cypseloides (Strcptoprocne) zonaris may 
also sieep on the wing a t  t imes. Wattl ing (68) similarly 
be l ieves that  Whitc-rumped Swift lets sleep on the 
wing.  but  he gives no evidence. My banding and recap?
ture data indicate tha t  the populat ion is larger than that  
found roosting in  Dry Cave , Nasinu on  any one nigh t ,  
and t herefore i t  i s  possib le  that  non-roost ing birds may 
have spent the night on the wing. 
The resul ts  of a Swedish experiment further de?
monstrate  that swift s  do not often fly a t  their  fastest 
speeds. A l t imeters, capable of measuring and record?
ing frequent air pressure readings. were p laced on 
Common Swifts that  were then released 405 km from 
their  nests. As wel l  as showing that  the  homing birds 
reached al t i tudes of up to 3600 m and in clear weather 
held an average altitude of 2300 m, the a lt imeters 
showed t hat the b irds spent an average of 12 h flying 
(34) . This  means the birds averaged less than 34 km/h if 
they flew in  a straight l ine .  
Moult of f1ighf Feathers. Moult  of the fl ight feathers 
could conceivably affect fl igh t performance in  the 
Apodidae . Because of their  tota l  dependence on aerial 
feed i ng and small size i t  i s  to  be expected that they 
would display some strategies for reducing the disaci?
vantages of moult ing.  Many temperate b irds solve the 
problem of moult  affecting fligh t  by having dist inct and 
separate t imes of the vcar for breeding and moul t ing 
t hereby reducing the a
-
mount of energy
' 
required at a1;; 
one t ime. This does not mean that  a l l  temperate zone 
swifts have a common moult and breeding cycle .  The 
Lacks (43) noted that ' I t  is remarkable that  of the two 
swifts breeding in  central Europe,  A. a pus migrates a 
few day? after breeding and moults in winter quarters, 
whereas A .  me/ha undergoes a nearly complete m(nt l t  
before leaving the breeding colonies' . In  West Africa 
the  Sabines Spi nctail  Raphidura sabini commences 
moult shortly after laying; hut t he Stumpy-tailed 
Spinetail Neafrapus cassini moults at  any time qf the 
year, and mou l ting i tse lf  is no indicat ion of an indi?
vidual's breeding cond it ion (6) .  
The swift lets , which arc all t ropical , show even grea?
ter d ivers i ty .  The Edible-nest Swiftlet showed no sea?
sonal pattern in the number of primary flight feathers 
replaced over a seven month period (44) . Moult  in  the 
Edible-nest Swift lc t  was arrested i n  some breed i ng 
birds,  and apparent l y  resumed at the  i nnermost pri?
mary as well as a t  the position where p rimary moult  
was halted. In  the B lack-nest Swift l e t ,  however, the 
progress of moul t ing the flight feathers was orderly, 
synchronized and annual ( 49) . Moul t  of  the primaries 
in  the B lack-nest Swift le t  started in N ovember or 
December and was not  interrupted in b reeding b irds 
? though complete body moult  took seven months.  Two 
of five specimens of the Mossy-nest Swift let  were 
found to be moult ing tenth primaries ,  ye t  a l l  five had 
regressed gonads. 
Admowledgements 
l wish to thank Edward Minot  for constructive 
advice during the preparation of th i s  paper: I a lso 
thank Mary Lecroy (American Museum of Natural 
H istory) , Charles T. Collins (Universi ty  of Cal i fornia) ,  
Kimball  Garre t t  ( Los Angeles County  Museum) and 
N.  Hyde ( Nat ional  Museum of New Zea land) for 
measuring swifts and swal lows . 
References 
I .  Arnold, K.A.  and T.C. Maxwcll .  1 970. The G reat  Swal?
low-tai led Swift (Panyptila sactihieronymi) from t he state 
of  Guerrcro, Mexico. Condor 72: 1 08 .  
2 .  Baldwin, P . H .  and N.K.  ZaczkowskL 1 963. Breeding  
biology of the Yaux Swift. Condor 65: 400-406. 
.'\ . Bccking, J . H. 1 97 1 .  The breeding of Col!ocalia gigas. 
/his 1 13: 332-334 
.J Bchle, W.l l .  i <J73. Clinal  variation in Whitc-throatcd 
Swifts from Utah and the Rocky Mountain region.  Auk 
90: 2'!9-306. 
5. Brooke, H . K .  ! 970. Ges>graphical variat ion and distribu?
t ion in  Apus bar/latus, A. bradji'eldi and A. niansac 
(Avcs; Apodidae ) .  Durban Museum Novitat.: 8: 363-374. 
il. Hrookc, R . K .  J <J7 1 .  Breeding of swifts i n  Eth iopian Afri?
ca and adjacent islands. Ostrich 42: 5-36. 
7. Brookc, R.r<.  I <J7 1 .  Taxonomic notes on some lesser 
known A pus swifts. Bulletin of the British Ornithological 
Club 91 :  33-36. 
R.  Brooke, R.K. 197 1 .  Geographic variat ion in the Alp ine 
Swift Ap11.1 (Tachymarptis) me/ha (Av-:s; Apodidac ) .  
Durban Museum Novitate 9 :  D l - 1 43 .  
9.  Brooke, R.K.  1 97 1 .  Geographic varint inn in the  swifts 
Apus !toms and Apus caffer (Aves: Apodidac) .  Durban 
Museum Novitare 9: 29-38. 
1 8 3  
Flight Behaviour of a Swiftlet and a Swallow 83 
1 0. Brooke, R . K. 1 97 1 .  Taxonomic and distributional notes 
on the African Chacturinc. Bulle1in of !he British 
Ornilhologi.I'IS Club 91 :  76-79. 
1 1 .  Brooke, R . K .  1 972. Geographical variation in Palm 
Swifts ()psiurus spp. Durban l>1useum Novitale 9: 2 ! 6-
:231. 
1 2 .  Brooke, R. K .  1973. Distributional and biological notes 
on the Mottfed Swift in Rhodesia. Os1rich 44: 1 06- 1 1 0. 
1 3 .  Bryant, D.M.  1 983. Heat stress in tropical birds? be?
havioural thermoregulation during !l ight .  Ibis 1 25: 3 1 }-
323. 
l ?l .  Bryant, D.M. and K.R. Westerterp. 1 979. Energetics of 
foraging anJ free existence in birJs. Proceedings uf !he 
1 7ch lfllcnuuional Ornithological Congress ( Berl in) .  
1 5 .  Cameron, A.C.  1 968. Feeding habits of the  Spine-tailed 
Swift. Emu 68: 2 1 7-2 1 9 .  
1 6 .  Carr, M . H .  and J . C .  Dickinson J r .  1 95 1 .  The San 
Geronimo Swift in  Honduras. Wilson Bullelin 63: 27 1 -
273. 
1 7 .  Chih-tung, K .  1 973. Preliminary studies in the breediJig 
behaviour of the House Swift. Aclil Zoologica Sinica 19 :  
293-304 . 
JR .  Collins, C.T. l 96R. The comparative biology of two 
species of swifts in Trinidad , West lndics. Bullelin of !he 
Florida Slate .\1u.seum 1 1 :  257-320. 
1 9 .  Collins, C.T. 1 972. Weights of some birds of north-cen?
tral Venezu.:la. Hulle1in of 1he Brilish Omirlwlogists 
Club 92: 1 5 1 - 1 53. 
20. Collins, C.T. and R.K. Brookc. 1 976 . A review of the 
swifts of  the g.:nus flirwulapus (Aves: Apodidae) . Con?
lribulions in  Science; Los Angeles Coumy Museum 282: 
1 -22. 
2 ! .  Cooke, P. 1 %5 .  Breeding of  Mottled Swift A pus 
aequarorialis (Von Mul ler) in the Matapos H il ls ,  S .  
RhoJcsia. Ostrich 36:  38-39. 
22. Delacour, .) . and E. Mayr. 1946. Birds oflhc Phiilj;pines. 
23. 
25. 
Ne\?: York : Macmi llan . 
Dementyev, G.P. ,  N.A.  Gladcov, E.S. P!ushenko, E.P. 
Spangcnbuq? and A. M .  Sudilovskaya. 1 95 1 .  Birds of the 
Soviel Union. Volume I .  (English tran'ilation: Israeli 
program for scientific translations. Jerusalem). 
Dolnik, V.I? .  and V.M.  Gavrilov. 1 97:l. Energy 
metabolism during flight of some passerines. In Bird 
Migmlions, Ecological and Physiological Faclors. Edited 
by B . E. Bykhovski i .  New York: Halstead Press. 
Dolnik, V.R.  and L.I. Kimzhcvskaya; 1 98 1 .  Time and 
e nergy budgets during the nest period in  the swift (Apus 
liJ>us) and swallows ( Delichon urbica. Hirwulo rustica). 
Zoo/oJ?,icheskii Zlwrna/ 59: 1 84 1 - 1 85 1 .  
26. Donnelly, B.G.  I \174. Vertical zonation o f  feeding swal?
iows and swifts at Kariba, Rhodesia. Oslrich 45: 256-258. 
27. Dunning, J . B .  1 984. Body weights of  686 species of 
North American bird,. Wcslem Bird Banding Associa?
lion Monograph No I .  
28. Eisenmann, E.  1 959. South American migrant swallows 
of the genus Progne in Panama and northern South 
America: with comments on their indentificat ion and 
molt. A uk' 76: 528-532. 
29. Fe1?guson-Lees, LJ. 1 960. Studies of less fami liar  birds. 
107. Needle-taileJ Swift. Brilish Birds 53: 43 1 -433. 
30. Fischer, R.B. 1 958.  The breeding biology of  the Chim?
ney Swift Chaewra pelagica (Linnaeus) .  New York Srate 
Museum and Science Service Bullelin 368: 43 1 -433. 
3 1 .  
33.  
Frith, H.J .  Editor. 1 976. Readers Digesl Complete Book 
of Australian Birds. Sydney: Readers D igest. 
Gibson, D.D. 1 98 1 .  Migrant birds at Shenya Island Aleu?
tian I slands A laska. Condor 83: 65-77. 
Gooders, J ,  1975. The Greal Book of Birds. New York: 
Dial Press. 
34. Gustafson, T., B .  Undkvist, L.  Gotborn and R.  Gyllin. 
1 977. Altitudes and flight t imes for swifts Apus ?apus L. 
Omis Scandinavia 8: 87-96. 
35. Hails, C.J. 1 979. A comparison of flight energetics i n  
hirundines and  other b i rds. Comparalive Biochemistry 
and Physiology 63a: 5 8 1 -585. 
36. Hails, C.J. and A.  Amil-rudin. 1 98 1 .  Food samples and 
food selectivity of  White-bellied Swiftlcts (Colloca/lia 
l'sctdenta). Ibis 1 23: 328-333. 
37. Harl'isson, T. 1 976. The food of Collocalia swiftlets 
(Aves; Apodidae) at Niah Great .Cave in Borneo. Jour?
nal of 1/ze Bombay Natural History Society 71 :  376-393. 
38. Hc?spenheidc, H.A.  1 975.  Selective predation by two 
swifts and a swallow in  Central America. Ibis 1 17 : 82-99 . 
39. Keskpaik, J. 1 968. Heat production and heat loss of  
swallows and martins during flight. Easti nsv teadusle 
akadeemia toimelised. XVII (Kaoide Biologia) 2: 1 79-
1 90 .  
40. Kuroda, N .  1 936. Birds of !he Islands of .lava 2: 374-385 . 
4 1 . Lack, D. 1 956. Swifts in a Tower. London:  Methuen. 
42. Lack, D .  1 957. The first primary i n  swifts.  A uk 74: 385-
386. 
43. Lack, D. and E. Lack. 1 95 1 .  The breeding behaviour o f  
t he  swift. Hritish Birds 45: 1 85-2 1 5 .  
44. Langham, N.J980. B reeding biology o f  the  Edible-nest 
Swiftlet (Aerodramus fuciphagus). Ibis 122: 447-460. 
-15 .  Lockwood, G . ,  M . P. Lockwood and M . A .  Macdonald 
1 980. Chapin's Spinetai l  Swift Telacalllhura melanopygia 
in Ghana. Bullelin of !he Brilish Ornithologists Club 
lOO: 1 62- 164. 
46. Lyuleeva, D.S. 1 970. Energy of fl ight in swallows and 
swifts. Dok Akad Nauk Tadzhik SSR 190: 1 467- 1 469. 
47.  McLachlan, G.R.  and R. Liversidge. 1 97 8 .  Roberts Birds 
of Sowh Africa. Cape Town: John Voelcker.  
..\8. Mayr, E. 1 945. Hirds of lite Sowh-wesl Pacific. 
Hcrtsfordshire:  Whcldon and Wesley. 
49. Mcdway, Lord. 1 962 . The swiftlcts (Co/localia) of Niah 
Cave. Sarawak part I .  Breeding B iology. Ibis 1.04: 454-
466. 
1 8 4  
84 Tarburton 
50. Morse, R.A. and f.M. Laigo. ! 968. The Phi l ippine 
Spinetailed Swift . Clwetura dubia McGregor as a Honey 
Bee predator. Philippine Entomologist 1 :  1 38- 143 .  
5 1 .  Neufeldt, I .  and A.L  Ianov. 1 960. Some notes on the 
biology of the Needle-tailed Swift in  S iberia.  British 
Birds 53: 43 1 -433. 
52. Palomcque, J., J . l). Rodriqucz, J.D. Palacios and .!. 
Lanas. 1 980. Blood respiratory properties of swifts. 
Comparallve Biochemistry and Physiology 67a: 9 1 -95.  
53. Payne, R.B.  1 969. Overlap of  breeding and moult ing 
schedules i n  a col iection of African birds . Condor 7 1 :  
1 40-1 45.  
54.  Pennycuick, C.J. 1 969. Mechanics of  b ird  rnigr'!tion.  Ibis 
1 1 1 :  525-556. 
55. Pennycuick, C.J. 1 972. Animal Flight. London: Edward 
Arnold. 
56. Rand, A.L. and E.T. Gilliard. 1 967. Handbook of New 
Guinea Birds. London: Weidcnfcld and N icolson. 
57.  Ricklefs, R.E. 1 97 1 .  Foraging behaviour of  Mangrove 
Swallows at Barro Colorado Island. Auk 88: 635-65 1 .  
58. Riplcy, S.D. 1 9SO. A new record of the Sooty Swift let  
Collocalia vanikorensis from New I reland. Bulletin of the 
Brirish Ornithologists Club 100: 238. 
59.  Selander, R.K. 1 955.  Great Swallow-tailed Swift in 
Michoacan,  Mexico. Condor 57: 1 23- 1 25 .  
60. Smyth, D.M. I (}80. Notes on the diet o f  t h e  Grey 
Swiftlet (Aerodramus spodiopygius). Sunbird l l : 20-2 1 .  
(Accepted: 3 January 1 986) 
6 1 . Snow, D.W. 1 962. Note on the biology of some Trinidad 
swifts. Zoulogica 47: 1 29- 1 39.  
62. Stoner, D .  1 936. Studies on  the Bank Swallow (Riparia 
riparia riparia) (Linnaeus), in the Oncida Lake region .  
Rousevelt \Vildlife Annals 4: 1 27-233. 
63 . Stresemann, E. 1 93 1 .  Notes on the systematics and d is?
tribution of some swiftlets (Collocalia) of Malaysia and 
adjacent subregions. Bulletin of 1he 8ajj1es l'vfusewn 6: 
83- 1 0 1 .  
64. Tarburton, M.K. 1 986. The food o f  the White-rumped 
Swiftlet in  Fij i .  Notomi.l 33: 1 - 16 .  
65 . Tucker, V.A. 1 968. Respiratory exchange and evapora?
tive water loss in  the Budgerigar. Journal of Experimen?
tal Biology 48: 67-87. 
66. Utter, J.M. and E.A. Lefebvre. 1 970. Energy expendi?
ture for free flight by the Purple Martin ( Progne subis) . 
Comparalive Biochemistry and Physiology) 35: 7 1 3-71 9 .  
67. Waugh, D.R.  1 978.  Predation strategies in  aerial feeding  
b irds. Unpublished Ph .D.  thesis.  University of  St ir l ing.  
68. Wattling, D. 1 982. Birds of Fiji, T01zga and Samoa. Wel?
l ington:  Mi l lwood Press. 
69. Wells, D.R. 1 975 . The Moss-nest Swiftlct (Colloca/ia 
vanikorensis) Quoy and Gaimard i n  Sumatra . Ardea 
63: 1 48- 1 5 1 .  
70. Wetmore, A .  1 957. Species l imitation i n  certain groups o f  
the swift genus (Chaetura) .  Auk 74: 383-31-:5. 
7 1 .  Zimmer, J.T. 1955. A review of the genus ( /'rogne). 
American Museum Nm?itatt's 1723: 8 .  
THE POPULATION STATUS, LONGEVITY & MORTALITY OF 
THE WHITE-RUMPED SWIFTLET IN .FIJI . 
M. K.  TARBURTON 
Dept . of Botany & Zoology, Massey University, 
Palmerston North, New Zealand . 
Received 
The population size of the white-rumped Swiftlet Aerodramus 
spodiopygius in five Fijian caves has been estimated by up to five 
methods . The averages of these methods for 1974 are : Dry Cave, 413 ; 
1 8 5  
Waterfall Cave, 20 994 ; Ono Cave, west entrance, 345 ; east entrance, 21 
888 ; Waiyala Cave, 8 430 ; Cikobia-i-lau, 210 . The largest annual 
average estimated population was 32 526 for Waterfall Cave in 1975 . 
Although the population of Dry Cave declined between 1974 and 1984 the 
recapture rate of marked birds remained high . These data show an 
average survival rate of 64 per cent, though a survival rate of 73 per 
cent (which is determined when data taken in the abnormal years of 1976 
and 1982 are deleted) , may be more realistic .  The higher rate, which 
gives an adult further life expectancy of 3 . 2  years is higher than most 
passerines , some seabirds and one other species of swift . Higher adult 
life expectancies have been shown for four other swift species . 
Juvenile mortality is shown to be too high to replace the adult 
population and it is reasoned that human interference in Dry cave is 
responsible for the high juvenile mortality . Adult mortality is low and 
arises from predation and accidents caused by conspecifics . The current 
longevity record holder in thi.s study is at least 12 years old . 
INTRODUCTION 
It might be assumed that if a population is large, its survival is 
not t?eatened, but without census data only guesses can be made as to 
its long term viability . 
Knowing something of the stability of the population upon which 
manipulation (Tarburton 1987 ) and longevity studies were made, could 
1 8 6  
help in the interpretation of the results of those experiments . The 
ability of the swiftlets to obtain adequate food for themselves and 
their chicks is the basis of both studies . In the former , the growth 
rate of chicks in artificially enlarged broods of three was compared 
with the growth rates for single chicks and those in broods of two 
young. It was reasoned that a population in balance with its food 
supply would not be able to gather enough food to feed an additional 
chick . On the other hand it has been suggested that for the Lesser 
Black-backed Gull Larus fuscus (Harris & Plumb 1965 ) and the North 
Atlantic Gannet Sula bassana (Nelson 1964) , the ability to raise broods 
artificially larger than normal was at least in part facilitated by an 
increased food supply resulting in a period of considerable population 
growth prior to and during the time the manipulation experiments were 
being conducted . Alternatively , a population that is either stable or 
in decline , as a result of a limiting food supply is unlikely to be able 
to raise a larger brood than that normally produced . 
In this paper the population trends in two Fi jian ( 18?S 179?E )  caves 
where manipulation experiments and longevity studies were conducted , are 
determined . The populations of these and three other Fijian caves are 
estimated and the locality of the nesting sites are mapped for four of 
the caves . Together five caves on two islands (Nasinu Caves [Waterfall 
Cave and Dry Cave ] , Ono Cave , and Waiyala Cave are on the island of Viti 
Levu, while Ono-i-Lau Cave is on the island of Ono-i-Lau) , were 
incorporated into the study . The further life expectancy of adults is 
calculated and that measure , survi vorship and longevity are discussed in 
relation to the stability of each population and are compared with those 
for a range of other birds . Finally , causes of mortality are identified 
and their relative importance discussed with a view to preserving these 
South Pacific populations of the White-rumped Swiftlet . 
1 8 7  
METHODS 
The location of the five Fi jian caves visited in this study are 
shown in Figure 1 .  All caves are in limestone and although Ono Cave has 
several levels of development none has reached the stage of collapse 
that allows light to enter internal passages or chambers . The majority 
of nest and roost sites are in the totally dark portions of each cave . 
The cave on Ono-i-lau is the only one without running water and it and 
Dry Cave are the only two that do not pass completely through the base 
of a hill to provide a second entrance . 
Dry and Waterfall Caves were the most frequently visited in the 
course of this and other studies with a total of 59 visits to Waterfall 
Cave and 45 to Dry Cave . Ono cave was visited 10 times while Waiyala 
and the cave on Ono-i-Lau were visited once . 
Five methods of estimating population size have been used in this 
paper . Counting the sleeping birds late at night , once they had ceased 
entering the cave , was the first method of census and is only possible 
with small populations in small caves . Thus this was only feasible in 
Dry Cave Nasinu and the sole cave on Cikobia-i-Lau. This method of 
estimating a population assumes that all birds that roost and breed in a 
particula.J{' cave enter it each night . To check this , estimates made by 
direct counting in Dry Cave were compared with a nest count and two 
capture-recapture analyses ; Jolly ' s Stochastic method and the Modified 
Peterson method (Begon 1979 ) . As the Peterson method does not allow for 
deaths , the number of marked birds estimated by this method was reduced 
at the end of each year by 29 per cent , being the estimated annual 
mortality rate for each of the first two years after banding . 
In the Waterfall cave , very few recaptures were made so the Schnabel 
method as well as those of Petersen and Jolly were applied to the 
recapture data and each result compared with the nest count . 
1 8 8  
In the larger caves the swiftlet populations were estimated by 
counting nests , most of which remained intact and were used from year to 
year . The sampled area in the Waterfall Cave at Nasinu contained just 
over three birds to each nest . An assumption was made that a ratio of 
three birds to each nest , held for all parts of the cave as well as for 
other caves ; the larger ones being censused within a week of completing 
the Waterfall Cave census . 
RESULTS 
CAVE POPULATION AND POPULATION TRENDS 
Dry Cave , Nasinu 
Data from the Dry Cave alone are sufficiently comprehensive to be 
analysed by Jolly ' s  Stochastic method . The results of this probablistic 
method are shown in Table 1 .  The raw capture and recapture data and 
preliminary computations are shown as Appendices 1 , 2 and 3 .  
The four estimates made in 1974 using Jolly ' s  method indicate an 
average population of 397 .?. 51 (x + SE) . Close to this estimate is that 
of the Modified Petersen estimate of 430 + 32 . The two estimates for 
1975 average 339 + 79 which is similar to the Petersen estimate of 346 .?. 
56 . The 1+ recaptures made in 1976 provide an estimated population of 
305 .?. 37 using Jolly ' s  method and 204 + 28 from Petersen ' s  method. 
These estimates differ considerably but both indicate a marked decline . 
This decline has continued , as the nest and bird count data in Table 2 
indicate . Nests declined from 163 in 1974 to 49 in 1983 . Bird counts 
indicated that the population declined from 200 in 1974 to 90 in 1983 
but was maintained at 94 in 1986 . Nest positions in the cave are shown 
in Figure 1 .  
1 8 9  
Table 1 
Population estimates of the Dry Cave 
population 
Jolly ' s  Estimates Modified Petersen 
Sample 
Date Mi n Ni + SE Oi Bi Ni + SE 
8 Aug 74 0 48 
18 Aug 74 54. 75 72 438 +158 0 . 47 233 389 +115 
4 Sep 74 106 . 46 115 515 +1oo 0 . 54 238 522 +118 
27 Oct 74 178 . 2  36 356 + 78 0 . 91 330 387 + 90 
19 Dec 74 143 . 72 34 279 + 57 0 . 90 29 420 +201 
23 Nov 75 155 . 85 58 418 + 78 0 . 93 30 401 +157 
4 Dec 75 155 . 87 19 260 + 55 0 . 96 11 290 +115 
10 Jnn 76 157 . 87 91 138 + 12 0 . 78 36 219 + 70 
13 Jtm. 76 254. 82 91 378 + 44 0 . 90 39 322 + 52 
7 Oct 76 213 . 66 60 214 + 25 0 . 93 16 308 + 59 
14 Oct 76 157 . 86 16 335 + 98 0 . 95 18 340 +154 
17 Oct 76 196 . 96 43 234 + 28 0 . 98 6 132 + 45 
22 Oct 76 174. 62 18 221 + 41 1 . 04 -8 163 + 57 
25 Oct 76 213 . 2  9 237 + 75 1 .00  1 90 + 46 
29 Oct 76 191 . 5  12 244 + 61 0 . 99 3 63 + 34 
7 Nov 76 312 . 0  88 463 + 82 0 . 92 38 153 + 52 
9 Nov 76 421 . 0  15 561 +300 0 . 99 7 235 + 70 
11 Nov 76 271 . 11 43 328 + 94 0 . 98 8 220 + 33 
Key: 
Mi = number of marked birds at risk . 
n = sample size . 
"" 
Ni = population size on day i .  
"" 
Oi =, stochastic survival rate . 
Bi = additions between i and i+l . 
1 9 0  
Table 2 
Population of Dry Cave , Nasinu 
No of Nests Nests Jolly ' s  Estirrate Mod . Petersen Bird 
Nests X 2 X 3 + SD + SD Census 
1974 163 326 489 430 + 51 430 + 32 200 
1975 339 + 79 346 + 56 
1976 142 284 426 305 + 37 204 + 28 88 
1978 91 
1979 94 
1980 82 
1981 61 122 182 
1983 49 98 147 90 
1986 94 
Table 3 
Population of Waterfall Cave 
3 X Modified Average of 
Nests Nests Petersen+SE Schnabel+SE Jolly all estirrates 
1974 3 660 10 980 22 266+ 5 273 26 040+2 121 24 688 ? 20 994+3 428 
1975 45 032+14 868 30 753+2 844 21 792 32 526+6 767 
1976 7 370 22 110 37 045+12 613 32 890+ 965 29 143 30 292+3 170 
1981 7 140 21 420 
, I 
, 
/ 
:430 \ 
" 
WATERFALL 
CAVE 
. I I I 
' 
I 
I 
I 
' 
I 
I 
I 
I 
I 
I .  
NAS INU CAVES 
4M  
1 
N 
7 - J -75 
Pou l tcr  
Tarburton 
DRY CAVE 
Fig .  L 
Nasinu Caves 
1 9 1  
1 9 2  
Waterfall Cave , Nasinu 9 Mile 
Although 2 545 birds were banded in the Waterfall Cave , the 
population is so large that it was the fifth evening visit before any 
recaptures were made. The average number of reca.ptures thereafter was 
. 
5 .  8 from an average sample size of 192 . This recapture rate was so low 
that population estimates using Jolly ' s  method ranged from 11 332 to 408 
165 having standard errors of 5 692 to 445 128 .  The resulting average 
of 91  054 ? 41 796 (+ SE) seems less realistic than the average derived 
by using only those estimates with standard errors less than the means . 
These averaged 25 690 ? 6 078 and are more in line with the Petersen and 
Schnabel estimates shown in Table 3 .  (Computations for Schnabel ' s  method 
are shown in Appendix 4) . The final column of this Table is an average 
of all methods used in estimating the population and indicates a 
population increase between 1974 and 1975 but no significant change into 
1976 . 
As no bands were recovered from this cave after 1976 (not an 
unexpected result with only about 8 per cent of the population marked) ,  
the 1981 nest count is the only method for assessment of the population 
size at the time the manipulation experiment was run (470 birds were 
caught in an effort to make recaptures) . This estimate indicates that 
the 1981 popUlation was similar in size to that of 1976 and 1975 , though 
the occurence of intervening fluctuations cannot be disproved . Nest 
numbers and positions are shown in Figure 1 .  
Ono cave , Wailotua 
The west entrance to this large cave has only a small colony. There 
were 115 nests in 1975 , giving an estimated population of 345 . This is 
much smaller than the 1 211 given by the average of five Modified 
Petersen estimates . This is probably explained by a high level of band 
1 9 3  
loss as a result o? the band being placed on the tarsus during the ?irst 
?ew trips . Bands were subsequently placed on the tibia , once recaptures 
revealed that the hind toe does not prevent bands ?rom sliping down over 
the other three toes and presumably in some cases slipping o?? the leg 
completely. No birds were ever ?ound ?lying the 390 m between the 
closest west and east colonies , but as in the Nasinu Caves , a small 
level o? exchange may have taken place through the separate entrances . 
The population at the east end o? the cave had 3 455 nests giving it 
an estimated size o? 10 365 birds . Nest positions in the cave are shown 
in Figure 2 .  The Modi?ied Petersen estimate was 17 909 ? 7 202 , ?rom 
?our recaptures in a total catch o? 167 . Another estimate when only one 
recapture was made in a catch o? 138 birds was 37 909 + 24 432 . A 
subsequent capture o? 298 birds made no recaptures .  Perhaps this 
disparity is due to some birds staying out at night , though the small 
sample sizes ( <3";G ) , in this case may also i?luence the accuracy. 
Waiyala Cave 
Only one visit was made to this cave (February 1975 ) and the best 
estimate o? _ its population size is made ?rom the count o? 2 800 nests 
giving 8 430 birds . The position o? these nests is shown in Figure 3 .  
Cikobia-i-l.a.u 
' Seventy nests were counted on 7 January 1976 making an estimated 
population o? 210 birds . 
SURVIVAL AND FURTHER LIFE-EXPECTANCY 
It is in the context o? a declining swi?tlet population in Dry Cave 
that the data ?om 502 banded birds have been used to determine 
survivorship and fUrther li?e-expectancy o? adult birds and hence the 
results should be regarded as conservative . 0? the 446 banded adults , 
WAILOTUA CAVE 
IDM 
(-o--N 
2-1 -75 
N. Poulter 
M. Tarburton 
Fig ? 2 ? 
Wailotua . Ono Cave , 
1 9 4  
Fig . 3 .  
Waiyala Cave 
WAIYALA CAVE 
4 M  
1 
13 - 1 -75  
N .  POULTER 
\ M. TAIUlURTON 
' 
' 
' \ 
' 
I 
606' \ I I 
I 
/ 
/ 
, /  
/ 
1 9 5  
' ( 
1 9 6 
261 were subsequently recaptured . This represents a recapture rate of 
59 per cent and contrasts with the nestling recapture rate of 17 per 
cent . Four adults were recaptured 11 times ; one was captured nine 
times ; three , eight times ; four , seven times ; 11 , six times ; 22 , five 
times ; 18 , four times ; 41 , three times ; 42 twice and 115 only once . 
The 788 recoveries of these 261 birds show a range of annual adult 
survival of 41 to 77 per cent and an average of 64 per cent . The lowest 
annual survival was that of 1976 , the year that the Dry Cave population 
had the most visits by me and to my knowledge the most disturbance from 
other people .  Nevertheless none of 99 clutches observed in the 
manipulation experiment (Tarburton 1987 ) were deserted and no chicks 
from the 130 observed in the same experiment died of starvation . 
Even though the two Nasinu caves were only 18 m apart few birds made 
the short transfer . On each banding visit to Dry Cave more than half 
the birds present were captured , yet only three of the 2 545 birds 
banded in Waterfall Cave were found among them. 
Thirty-five of 48 birds taken from Dry Cave and released at Fulton 
College , 21 km to the north , on 10 June 1976 , were subsequently 
recaptured at Dry Cave . This represents a recapture rate of 73 per cent 
which is identical to the average adult recapture rate for this 
population. Hence my handling of birds , whether breeding or 
non-breeding does not cause desertion of the breeding cave . 
When the data for all nine years are used the expected further adult 
life expectancy is determined to be 2 .3 years . When the data for 1976 
and 1982 are excluded , the expected further adult life is estimated to 
be 3 . 2  years . 
DISCUSSION 
There has been some discussion of the accuracy of Lack ' s  method for 
population analysis .  For example Piper et al ( 1981 ) found estimates on 
No 
Table 4 
Survivorship of adult birds in Dry Cave 
as at 1-9-83 
Banded No Still Present 
Birds Present 1 Year Later 
1974 69 52 
1975 119 91 
1976 160 66 
1977 88 65 
1978 65 44 
1979 44 34 
1980 34 23 
1981 23 17 
1982 17 10 
1974-1982 629 402 
% Survival 
75 
76 
41 
74 
68 
77 
68 
74 
59 
64% 
Further adult life expectancy for all years = 2-m = 1 . 64 = 2 . 3  years 
(m = mortality) 2m 0 . 72 
Further life expectancy excluding the poor years of 1976 and 1982 
= 1 . 73 = 3 . 2  years 
0 . 54 
1 9 7  
1 9 8  
vultures using Lack ' s  method were lower than those of Haldane and Piper 
which allow for the incompleteness of data from bands yet to be 
recovered . Yet Seber ( 1972 ) states that it can be shown rrathematically 
that the methods of Haldane and Lack still hold if the recoveries are 
ignored for an initial period of any length . In this longevity study 
banded swiftlets were not counted as present unless they were recovered 
subsequent to the 1 September following their initial recapture . As all 
eggs hatch before the end of February all birds would be at least seven 
months old before being included in the calculations . These data then 
should be reasonably accurate and in any case are comparable with the 
data for the other species mentioned , as the same method was used to 
determine their rnortali ty . As all of the Apodid species have similar 
feeding and breeding ecologies , comparisons within the family should be 
reasonably valid . 
The adult recapture rate of 59 per cent in this study contrasts with 
the recapture rate of 2 . 1  per cent for adult Common Swifts Apus apus in 
Britain between 1909 and 1969 (Spencer 1971 ) . However a Russian study 
on the Common Swift (Kashentseva 1982 ) had better returns than the total 
British banding scheme . Between 1950 and 1979 , 4 . 1  per cent of 
juVeniles and 40. 1  per cent of adults were recaptured from 667 banded 
swifts . Both the Russian study and this study have much higher returns 
than normal , recovery rates of all bird species banded in Britain and 
America, being usually less than 4 per cent (Botkin & Miller 1974 ) . 
That frequent or severe hurran disturbance and low survi va1 rray be 
related is suggested by several authors finding that disturbance causes 
avoidance of the site of capture by up to one third of the Common Swift 
population (Lack 1956 ) ? Lack qualified this statement by adding that 
some birds taken from their nests for banding and measuring deserted but 
when the birds were banded on the nest desertion was very uncommon . 
1 9 9  
It might be reasoned that birds deserting one of the two Nasinu Caves 
as a result of being disturbed would be more likely to go to the other 
cave as the two caves are only 18 m apart , whereas the next nearest cave 
is  10 km away. From the proportion of birds banded in the Waterfall 
Cave and caught in Dry Cave it is estimated that six birds banded in the 
Waterfall Cave would have subsequently been caught in the Dry Cave if 
all birds present had been captured at the time of each visit . If it is 
assumed that the same percentage move in the reverse direction and that 
there is no movement between the caves before the birds are handled for 
banding , then only one of the 502 birds banded in the Dry Cave will have 
transfered to the Waterfall Cave . 
However the observable effect of disturbance in the two caves may not 
be equal . The Waterfall Cave is quite large in cross-section (most 
birds are out of reach of a hand-net ) , is long ( 178 m)  and has two 
entrances . Birds banded and released in the Waterfall Cave may just 
relocate within that cave and would thus be rarely recaptured in the Dry 
Cave . Dry Cave on the other hand is small in cross section (all nests 
and most birds can be reached by the hand-net ) ,  is short ( 90 m) and has 
only one entrance . This means that in Dry Cave a greater percentage of 
birds will be caught or otherwise disturbed at each visit , than is the 
case in the Waterfall Cave . If birds from Dry Cave relocated in the 
Waterfall Cave there would be little chance of their recapture and hence 
little chance of determining whether members of the Dry Cave population 
were deserting more frequently . due to their greater disturbance . 
It is also possible that some birds may change caves periodically 
whether disturbed by humans or not . If this is the case the three birds 
banded ?n Waterfall Cave and retrapped in Dry Cave can be used to 
estimate the total movement of swiftlets from the Waterfall Cave to the 
Dry Cave . The retrapped birds are counted as six to allow for that 
(almost ) half of the birds present each visit that were not caught . The 
2 0 0  
2 545 birds banded in the Waterfall Cave make 8 . 4  per cent of the 
estimated total population of 30 292 . Thus 71 birds are likely to have 
moved from the Waterfall Cave to Dry Cave . However unless a greater 
percentage move in the reverse direction we can still only account for 
one bird moving from the Waterfall Cave as a result of ' random' 
movement , for almost all birds in the Dry Cave were banded by the end of 
1976 . 
The above reasoning assumes random movement , but that this does not 
always exist in swifts is shown by the regular use of two chimneys 
during the northward movements of Chimney Swifts ( Chaetura pelagica) in 
Texas whereas only one of the chimneys is used during their southward 
movements in autumn (Michael & Chao 1973 ) ? That swift lets caught in Dry 
Cave may sometimes sleep elsewhere cannot be discounted . 
Since the Dry Cave allows for a far higher percentage capture of 
birds present than does the Waterfall Cave , this activity itself 
possibly creates greater disturbance with more birds leaving the cave 
after human activity in it . A small post-disturbance exodus is  
possible ,  as each of  the methods used for estimating the population and 
shown in Table 2 give consistently higher estimates through the 1970 ' s 
than the numbers of birds found to be in the cave . But there are other 
possibilities . 
It is likely that some birds were still to return after the time of 
the visit . I have recorded arrivals as late as 2230 hours , the latest I 
have made observations . The rate of arrival at that time of night is 
however very low,  though it  may continue for some time as Medway ( l961 ) 
has recorded Black-nest Swiftlets Aerodramus maximus returning as late 
as 0310 hours . That birds may delay their return to the roost is 
substantiated by the first reported night feeding for the White-rumped 
Swiftlet . Jim Pierce who is familiar with this swiftlet told me of its 
feeding on insects flying around fluorescent lights at the Williamstown 
mining camp 25-30 km south-west of Mungana , Queensland . These swiftlets 
2 0 1  
were seen feeding amongst the bats for an hour or so after sunset on at 
least two nights in September 1985 . 
Some birds may stay in the field . Several Fi jians have told me 
swiftlets will sleep in the coil of a young banana leaf . Another Fi jian 
whom I consider reliable , once saw a swiftlet enter such a banana leaf 
during the day and leave it a short time later . The Fi jian belief , that 
they sleep in the roll of a young banana leaf may have developed from 
sightings such as this . After all , the Fi jian belief that the swiftlet 
has no legs appears to have developed from the observation that the 
birds never land on tree branches .  
Watling ( 1982 ) suggests some swiftlets probably sleep on the wing but 
no supporting evidence is given . I presume the view is simply a 
transfer of Lack ' s  report that the Common Swift sometimes sleeps on the 
wing . 
That alternative sleeping places may exist does not , however , mean 
that they are used , nor does it mean that they are used more when the 
birds are disturbed frequently at their normal roost site in the cave . 
However such possibilities do allow for the discrepancy between the 
population estimates and the number of birds counted . 
My high recapture rate for all adults ( including those held 
overnight and taken some distance away) , caught in Dry Cave suggests my 
handling of the birds was not causing a significant decline in the 
population . However , disturbance through nest destruction , which is 
quite possible in Dry Cave , (due to other persons visiting it ) could 
cause a population decline in three ways . Birds having to rebuild their 
nests could experience greater physiological stress , resulting in higher 
mortality. Birds losing their nests and clutch or brood may be more 
likely to abandon the colony . This has been suggested to explain the 
persistent decline of Black-nest Swiftlets in Madai Cave ( Sabah , 
Malaysia)  , where the nests are harvested for human consumption (Dal ton 
2 0 2  
1977 ) . Finally, replacement nests may not be as large or as strong as 
the original nest , resulting in higher egg or chick losses from the 
eggs , chicks and/or nests falling to the ground . Replacement nests in 
the Edible-nest Swiftlet Aerodramus fuciphaguS are inferior in this way , 
( Gibson-Hill  1948 ) , though the effect on breeding success is not known. 
If we consider the disturbances of 1976 and 1982 to be abnormally 
excessive and so delete the data for these years , we obtain an average 
adult survival rate of 73 per cent instead of 64 per cent . The 
consequent average mortality of 27 per cent (range 16-32 per cent ) means 
the White-rumped Swiftlet does considerably better than the Barn Swallow 
Hirundo rustica ( 63 per cent ) and 12 other passerines ( 41-72 per cent ) 
but less than the Alpine Swift Apus melba ( 18 per cent ) , and the Common 
Swift ( 20 per cent ) (all in Lack 1954) . Two other studies on the Common 
Swift found mortalities to be between that which Lack found for the 
Common Swift and those found in this study . In the USSR mortality was 
24. 4  per cent (Kashentseva 1982 ) , and in Britain it was 21 per cent . 
Two other swifts also have lower mortality than the White-rumped 
Swiftlet . These are the White-throated Swift Aeronautes saxatalis  of 
the United States , which has an annual mortality of about 20 per cent 
( Collins 1973) and the Chestnut-collared Swift Cypseloides rutilus , 
which in Trinidad has 15-17 per cent mortality ( Collins 1974) . While 
the recapture rate for the White-tipped Swift Aeronautes montivagus in 
Venezuela ( Collins unpub. ) is about 65 per cent , Collins work on this 
species leads him to believe that the real figure is about 82 per cent . 
It appears that both a disturbed bird and its mate are likely to lose 
?;,?r_? ? .  ? 
the nest site and leave the colony , hence avoiding recapture ( Collins , 
pers . comm. ) .  If this is correct , ( as is to be expected , for larger 
birds tend to live longer) ,  the only swift with higher mortality than 
the Whi te-rumped Swiftlet is the Chimney Swift of the United States , 
which has an annual mortality of 38 per cent (Henny 1972 ) . The 
2 0 3  
White-rumped Swiftlet has lower mortality than 1 5  other non-passerines , 
including two seabirds and is only bettered from those non-passerines 
given in Lack ( 1954) by the Royal Albatross Diomedia epomophora and the 
Yellow-eyed Penguin Megadyptes antipodes which have annual mortality 
rates of 3 per cent and 10 per cent respectively . 
Both estimations for further adult life expectancy ( 2 .  3 years and 
3 . 2  years) , appear reasonable , when compared with 1 . 1  years for the Barn 
Swallow, 4 .6  and 5 . 6  years for the Common Swift , (Magnusson & Svardson 
1948 ; Weitnauer 1947 ) . However because of the declining population in 
Dry Cave even the life expectancy of 3 . 2  years should correctly be 
regarded as conservative . 
Just how conservative an adult life expectancy of 3 .  2 years is , can 
be estimated by calculating the number of years it would take for 
parents to replace themselves with breeding offspring at variously 
selected mortality rates . By using the annual fledging success data 
(1 . 1  chicks per pair per year ) from the stable population of the 
Waterfall Cave and the 80 per cent survival rate of adult European and 
American swifts , it would take 2 .3 years for parents to replace 
themselves . .With 74 per cent survival (the average of Fi jian swiftlets 
without the two abnormally poor years) replacement would take 2 .  5 years . 
With 64 per cent survival (the average of all years for f'i jian 
swiftlets ) replacement would take 2 . 9  years . Clearly each of these 
replacement rates could be achieved in the 3 . 2 years of further adult 
life estimated from the declining population of Dry Cave . However as 
juvenile  mortality is  usually higher than adult mortality , lower 
survival rates than those used should be expected . The proven first 
year survival of 21 . 25 per cent (from 74 banded chicks ) in the declining 
Dry Cave population would require 8 .  6 years to replace parents and can 
be regarded as below the minimum of that in a stable population. If we 
raise the juvenile survival to 50 per cent the parents would be replaced 
2 0 4  
in 3 . 6  years , which i s  the average longevity of the adult Common Swift 
in Russia (Kashentseva 1982 ) . 
The Corrmon Swift also has an adult mortality similar to that of the 
White-rumped Swiftlet , and so (assuming no net migration gain or loss ) 
if the same ratio of juvenile to adult mortality holds for the swiftlet , 
50 per cent mortality between fledging and breeding may be realistic . 
If it is realistic , then disturbance of the birds by the suspected 
destruction of their nests and contents in Dry Cave has considerably 
reduced juvenile survival and has led to the decline observed in that 
population . 
Maximun Recorded Longevity 
At the time of writing (March 1986 ) the oldest recorded bird from 
the 502 banded in Dry Cave was 013-69752 banded on 4 September 1974 and 
last recaptured on 27 February 1986 , 137 months having elapsed . As the 
bird was an adult when banded it would have been at least 12 years old 
at the time of recapture . Two other adult birds have been recaptured 
112 months after banding , making them at least nine years old when last 
recorded . 
There are few longevity records for Apodidae with which to compare 
this record for this species . The oldest recorded Chimney Swift was 13 
years (H:ight 1953) . The record for the Alpine Swift is 16 years 
(Rydzewski 1962 ) and 21 years for the Common Swift (Rydzewski 1962 ) . 
The oldest recorded Corrmon Swift in the Russian study is 11 years 
(Kashentseva 1982 ) and in a Czechoslovakian study 12 years 11 months and 
21  days ( Beklova 1976 ) . The records for two swallows in this last study 
show that they only live about half as long as the Apodidae . The 
longest? records for H:irundine longevity are seven years for the Barn 
Swallow and six years five months for the House Martin Delichon urbica. 
2 0 5  
Mortality and its causes 
Because the survivorship of the White-rumped Swiftlet is here shown 
to be greater than most other similarly sized land-birds studied so far , 
it  naturally follows that its mortality will be low compared to theirs . 
Adult mortality was shown to average 27 per cent for seven years . If 
the years ( 1976 & 1982 ) showing abnormally high mortality are included 
the average mortality for the nine years rises to 36 per cent . It can 
be reasoned that the practice of being airborne all day and of roosting 
and breeding in what may be thought of as the safe environment of a cave 
would help reduce mortality . However mortality remains , and some 
observations and discussion regarding its causes will  help clarify 
whether the feeding , roosting and nesting habits do enhance longevity , 
not only for this species but possibly also for other species having 
similar ecological habits . 
That man has little direct effect on the mortality of this bird 
outside the caves is  evidenced by the fact that whereas numerous bands 
from those I have placed on similar numbers of other bird species in 
Fiji have been returned , none of the 4 554 swiftlets I banded in Fiji 
were ever recovered away from the caves they use for nocturnal roosting 
and breeding . The small size of this swiftlet means man is not 
interested in it as a food source . Though some indigenous Fi jians are 
very good at collecting birds by throwing stones at them I have heard 
only once of their collecting a swiftlet in this manner . The extreme 
difficulty I had in trying to mist-net this swiftlet in the field also 
demonstrated that their keen eye sight and rapid manoeuvrability make 
them much harder for man to capture th:m most land bi rds . 
Even the caves offer protection from man when the birds are roosting. 
Indians are reticent to enter caves for fear of snakes and Fi jians 
rarely enter alone , giving as their reason that the devH lives there . 
However when a group of Fijians do go in to catch the free-tailed 
2 0 6  
fruit-bat for food , numbers of swiftlets may also perish . In Ono Cave 
at Wailotua village in the Wainibuka Valley , bamboo is burnt to drive 
the bats (and birds) into small dead end passages and if this is  done 
when swiftlets are present , numbers of swiftlets may also perish. 
Apart from man the White-rurnped Swiftlet in Fiji has few predators . 
A Pacific Python Enygrus bibranii , a little over a metre in length was 
found sleeping on a rock below nests in the eastern end of Ono Cave . It 
would be unreasonable to expect that pythons would feed on anything but 
chicks and eggs that fell from nests . American Cockroaches Blattaria 
americanus and large fresh-water eels do the same , though the 
cockroaches also feed on the saliva that glues the nests to the wall . 
Although cats are reported ( Clunie 1984) , to bat them down when flying 
low,  most birds feed over the forest where such a fate is not likely. 
The Barn Owl Tyto alba does take adult and nestling swiftlets . It 
was said to be responsible for the abandonment of several score of nests 
placed in the twilight zone of the upper entrance ( south end) of the 
Waiyala Cave ( see map) ? Many eggs lay on the guano below the nests and 
villagers from Waiyala said they had seen the Barn Owl chasing swiftlets 
in this entrance . I have found a Barn Owl feather in the entrance to 
Waterfall Cave and Clunie ( 1972 ) has seen a Barn Owl catching swiftlets 
at the entrance to a cave in Navosa. I have picked up several freshly 
dead and concussed birds from the stream in the entrance of Waterfall 
Cave . However , they were probably victims of head-on collisions in the 
zone where the birds fly their fastest , though the possibility that they 
were struck by a Barn Owl could not be ruled out . Even in Europe where 
there are abundant data for avian predation on the Common Swift the 
diurnal predators take few swifts compared to other species . The Common 
Swift forms only one percent of the prey of the Sparrowhawk Accipiter 
nisus , one and a half percent of the prey of the Peregrine Falcon Falco 
peregrinus and two and a quarter percent of the prey of the Hobby Falco 
2 0 7  
subbuteo (Lack 1956 ) . Clunie ( 1972b , 1976 ) has shown that swiftlets 
comprise only a small portion of the diet of tne Fi jian Peregrine 
Falcon. 
In short , the low mortality rate of the White-rumped Swiftlet results 
from the inability of terrestrial predators to reach them and the 
limited effect of aerial predators on their numbers . (Neither Barn Owls 
nor Peregrine Falcons flock at cave entrances ) .  So apart from periodic 
interference from man, the availability of food appears to be the main 
regulator of Fijian populations of the Whi te-rumped Swiftlet . As no 
chicks starved in the manipulation experiment , even in the artificially 
enlarged broods of three (Tarburton 1987 ) , pressure from a lack of food 
does not appear to be critical in a normal brood situation and one is 
left with the likelihood that a period when available food is  low such 
as in a prolonged cyclone , and/or a period of excessive human 
interference may individualy or in unison increase mortality .  
It  is probably predatory pressure from Barn Owls that has encouraged 
the majority of swiftlets to nest beyond the twilight zone in the five 
Fijian caves I have examined . That swiftlets increase their speed at 
cave entrances (Tarburton 1986 ) , supports this view. This view is 
contrary to Watling ' s  ( 1982) statement that most nests are built in the 
twilight zone of caves . In Waiyala only 4 per cent of nests were in the 
twilight zone , the rest in total darkness . In Dry cave only 1 per cent 
and in Waterfall cave only about 27 per cent were in the twilight zone . 
At Ono Cave , 36 per cent of nests were in the twilight zone but none of 
the nests in the cave on Cikobia-i-Lau were in the twilight zone . The 
position where twilight gives way to total darkness is  shown as a dotted 
line across the passage on each map . 
Death may result from the activity of conspecifics . I found five 
adults dead at their nests . Their wings had been glued by saliva to 
their neighbours nest . This presumably happened while they slept but 
2 0 8  
the hardened saliva held them suspended in the air when they attempted 
to fly , thus preventing them from feeding . Two other birds had not been 
long in the same predicament and were rescued . This problem is  clearly 
caused by high density nesting . 
CONCLUSIONS 
Population size of the White-rumped Swiftlet in Fiji correlates with 
island size , except that a small colony may be found even on the largest 
island if it is close to another colony . Because most nests are in 
total darkness and on overhanging cave walls and roofs , brooding birds , 
eggs and chicks are safe from most natural predators . Handling of the 
birds or their young does not cause desertion , but the marked decline in 
the population of Dry Cave , where all nests can be reached by humans is 
thought to have occured as the result of willful destruction of nests 
eggs , chicks and possibly adults by man . The other small colony in this 
study is not under threat as the villagers on Ono-i-Lau protect the site 
and the birds . The large colonies are not considered to be in danger 
either , as most nests are out of easy reach and there is little interest 
in catching such small birds . 
The factors that have made the Dry Cave population vulnerable to 
human predation have also brought higher percentages of band retraps in 
this longevity study than in all other studies on Apodids . It is  
concluded that my activity in  collecting the data that show an expected 
further adult life of 3 . 2  years has not significantly reduced the birds 
survival and that the estimate is close to reality. However the lower 
than expected juvenile survival is attributed to the destruction of eggs 
and young by other visitors to the cave . 
2 0 9  
ACKNOWLEI:X}EMENTS 
The helpful advice and criticism of earlier drafts of this paper given 
by Dr.  Ed .  Minot of Massey University is gratefully acknowledged . I also 
wish to thank the following persons for valuable assistance in the field : M. 
& A. Ahkoy , A .  Currie , C .  Grieve , P. Hoffman , T. Kabu , R. Litster , T .  
Osborne and J .  Wells . 
REFERENCES 
Begon , M .  ( 1979 ) . Investigating Animal Ab.mdance - Capture-recapture for 
Biologists . Edward Arnold , London . 
Beklova , M .  ( 1976 ) . Contribution to the characteristics of population 
dynamics of certain hemisynanthropic species of birds in Czechoslovakia. 
Zool . Listy 25 : 147-155 . 
Botkin , D . B .  and Miller , R . S .  ( 1974) , Mortality rates and survival of birds . 
Arner. Nat . 108 : 181-192 . 
Clunie , F .  ( 1972a ) . Fi jian Birds of Prey. Fi ji Museum , Suva. 
Clunie , F .  ( 1972b) . A contribution to the natural history of the Fi jian 
Peregrine . Notornis 19 : 302-322 .  
Clunie , F .  ( 1976 ) . A Fi ji Peregrine (Falco peregrinus ) in an urban-marine 
environment . Notornis 23 : 8-28 . 
Clunie , F .  ( 1984 ) . Birds of the Fiji bush Fi ji Museum. Suva. 
Collins , C . T .  ( 1973 ) . Notes on survival and band wear in White-throated 
Swifts . West . Bird Bander 48 : 20-21 . 
Collins , C . T .  ( 1974) . Survival rate of the Chestnut-collared Swift . West . 
Bird Bander 49 : 10-13 . 
Dalton , K .  ( 1977 ) . Letter from Kuching. Far East Econ. Rev. 21 , Oct . p .62 . 
Gi bson-Hill , C .A.  ( 1948) ? The M3.laysian Swiftl ets . M3.layan Nat . J .  3 :  
190-200. 
2 1 0  
Har.ris ,  M.P.  and Plumb , W. J .  ( 1965 ) . Experiments on the ability of Herring 
Gulls (Larus argentatus ) and Lesser Black-backed Gulls (L .  fuscus ) to 
raise larger than normal broods . Ibis 107 : 256-257 . 
Henny , C . J .  ( 1972 ) . An analysis of the population dynamics of selected avian 
species with special references to changes during the modern pesticide 
era. Bureau of Sport , Fish . and Wildl . ,  Res .  Rep . 1 :  1-99 . 
Hight , G . L .  ( 1953 ) . Longevity record for Chimney Swift ( Chaetura pelagica) . 
Bird Banding 24: 109-110 . 
Kashentseva , T .A. ( 1982 ) . Age-related structure of the Swift (Apus apus ) 
population in Oka reserve , Russian S . F . S .R . , U . S . S .R .  Vestn , Zool . 0(3 ) : 
44-48 . 
Lack, D .  ( 1954) . The Natural Regulation of Animal Numbers .  Oxford University 
Press , London. 
Lack, D .  ( 1956 ) . Swifts in a Tower . Methuen , London. 
Lack , D .  and Am, H .  ( 1947) , Die bedeutung der gelegrobe beim Alpensegler . 
Orn. Beob . 44: 188-210 . 
Ma.gnusson, M. and Svardson , G .  ( 1948 ) , Livslangd hostornsvalor (Micropus 
? ( L) ) .  [Age of Swifts (Apus apus ) ] .  Var Fagelvard 7 :  129-144 . 
Medway, Lord. ( 1962 ) . The swiftlets ( Collocalia) of Niah cave , Sarawak part 
2 .  Ecology and the regulation of breeding . Ibis 104: 228-245 . 
Michael , E . D .  and Chao W.H.  ( 1973 ) . Migration and roosting of Chimney Swifts 
in east Texas . Auk 90: 100-105 . 
Piper , S . E . , Mundy , P . J .  and Ledger , J .A. ( 1981 ) . Estimates of survival in 
the Cape Vulture ( Gyps coprotheres ) . J .  Anim. Ecol . 50: 815-825 . 
Rydzewski , W. ( 1962 ) . Longevity of ringed birds . ? 3 :  147-152 . 
Seber , G .A. F .  ( 1972 ) . Estimating survival rates from bird band returns . J .  
Wildl . Manag. 36 : 405-413 . 
Spencer , R .  ( 1971 ) . Report on bird ringing for 1969 . Brit . Birds , 65 : 
137-186 . 
2 1 1  
Tarburton , M .K. ( 1986 ) . A comparison of the flight behaviour of the 
White-rumped Swiftlet and the Welcome Swallow.  Bird Behaviour 7 : 72-84. 
Tarburton, M.K.  ( 1987 ) . An experimental manipulation of clutch and brood 
size of White-rumped Swiftlets Aerodramus spodiopygius of Fiji . Ibis 
129 : 107-114. 
Watling , D. ( 1982 ) . Birds of Fi ji , Tonga and Samoa. Milwood press , 
Wellington . 
Weitnauer , E .  ( 1947 ) . Am neste des mauerseglers , (Apus apus apus) (L . ) .  
Ornith .  Beobachter 44:  133-182 . 
Appendix 1 .  
Raw Data for use in Jolly' s Population Estimate of the Dry Cave Population 
Time of release of marked birds ( j ) 
18 4th 27 19 23 4th 10 13 7th 14 17 22  25 29 7th 9th 11 20 
Aug Sep Oct Dec Nov Dec Jun Jun Oct Oct Oct Oct Oct Oct Nov Nov Nov May 
Date Dayi n .  r .  ' 74 ' 74 ' 74 ' 74 ' 75 ' 75 ' 76 ' 76 ' 76 ' 76 ' 76 ' 76 ' 76 ' 76 ' 76 ' 76 ' 76 ' 78 l l 
8 Aug 74 1 - 48 Number of Marked Birds Reca2tured (m . . ) 
18 Aug 74 2 71  71 8 l J  
4 Sep 74 3 115 115 8 15 
27 Oct 74 4 37 36 2 6 10 
19 Dec 74 5 34 34 3 1 11 2 
23 Nov 75 6 58 58 3 5 5 4 4 
14 Dec 75 7 19 19 0 1 4 0 3 3 
10 Jun 76 8 90 90 1 2 9 8 6 10 7 
13 Jun 76 9 91 89 6 4 7 1 2 11 5 25 
7 Oct 76 10 60 60 0 2 3 0 2 2 1 15 17 
4 Oct 76 11 16 16 0 0 0 0 0 1 0 3 1 2 
17 Oct 76 12 43 41 2 2 2 2 2 3 1 6 3 10 3 
22 Oct 76 13 18 18 0 1 1 0 1 1 0 1 2 1 2 4 
25 Oct 76 14 9 9 1 0 0 0 1 0 0 1 3 0 1 0 1 
29 Oct 76 15 13 12 0 0 0 0 0 1 0 2 0 1 2 2 1 1 
7 Nov 76 16 88 88 0 0 3 1 1 2 0 3 8 15 3 9 6 2 6 
9 Nov 76 17 15 15 0 0 0 0 0 0 0 0 2 1 0 2 1 0 0 5 
11 Nov 76 18 45 43 1 1 6 1 1 0 0 4 5 2 3 3 2 0 1 6 0 
20 May 78 19 92 92 0 1 1 1 0 5 2 5 3 4 1 5 2 2 1 13 3 9 
KEY 
n.  = number captured on day i ,  r .  = number marked & released on day i ,  m .  = number of marked recaptures on day i .  l l l 
N 
f-' N 
KEY 
ri 
m.  l 
z .  l 
yi 
Appendix 2 
Preliminary Computations for Jolly ' s  Estimate 
of the Dry Cave Population 
Day , i  r .  m.  yi z .  l l l 
1 48 35 
2 71 8 41 27 
3 115 23 62 45 
4 36 18 20 89 
5 34 17 22 93 
6 58 21 40 93 
7 19 11 16 122 
8 90 43 76 97 
9 89 61 45 98 
10 60 42 36 103 
11 16 7 14 132 
12 41 36 27 106 
13 18 14 13 116 
14 9 8 5 114 
15 12 10 8 121 
16 88 59 24 69 
17 15 11 3 82 
18 43 37 9 49 
19 92 58 
= number of marked birds released on day i (Appendix 1 )  
= number of marked birds recaptured on day i .  
= number of birds marked before day i ,  but not recaptured 
until after day i . 
= number of ri individuals subsequently recaptured . 
2 1 3  
KEY 
Appendix 3 .  
Results Using Jolly ' s  Estimate of the Dry Cave Population. 
Sample M. N.  SE Ni 0 .  B. l l l l 
1 0 
2 54. 76 438 158 0 . 47 233 
3 106 . 45 515 100 0 . 54 238 
4 178 . 20 356 78 0 . 91 330 
5 143 . 72 279 57 0 . 90 29 
6 155 . 85 418 78 0 . 93 30 
7 155 . 87 260 55 0 . 96 11 
8 157 . 87 158 12 0 . 78 36 
9 254 . 82 378 44 0 . 90 39 
10 213 . 66 214 25 0 . 93 16 
11 157 . 86 333 98 0 . 95 18 
12 196 . 96 234 28 0 . 98 6 
13 174 . 62 221 41 1 . 04 -8 
14 213 . 20 237 75 1 .00  1 
15 191 . 50 244 61 0 . 99 3 
16 312 .00 463 82 0 . 92 38 
17 421 . 00 561 300 0 . 99 7 
18 271 . 11 328 94 0 . 98 8 
Mi = the estimated number of marked birds at risk on day i .  
Ni = the estimated population using the modified Petersen 
Oi 
Bi 
forrrula 
N. l 
? 
= M. (n. +1 )  l l 
(m. +1) l 
= the estimated stochastic survival rate 
= the estimated additions between i and i + 1 
SE Ni = the standard error of the estimate , for method of 
estimation see Begon ( 1979 ) . 
2 1 4  
Appendix 4 
Computation for estimates of Population in Waterfall Cave - Schnabel ' s  Method. 
A B c Estimate of 
No . Newly No . Already No . Population 
No . Marked & Marked in Recap- (A X B)  
Date Trapped Released Population A x B (A X B) -tures c --c 
7 APR 74 27 27 0 
8 AUG 74 83 83 27 
18 AUG 74 165 165 110 
4 AUG 74 129 129 275 
27 OCT 74 121 121 404 48 884 48 884 2 2 24 442 
26 DEC 74 264 264 525 138 600 187 484 6 8 23 436 
29 DEC 74 299 299 789 235 911 423 395 6 14 30 243 
26 040+2 121 
22 JUN 75 164 164 796 130 544 448 090 7 18 24 894 
16 JUL 74 235 235 1031 242 285 690 375 6 24 28 766 
23 NOV 75 75 75 1106 82 950 773 325 1 25 30 933 
4 DEC 75 202 201 1307 264 014 1 037 339 2 27 38 420 
30 753+2 844 
-
10 JUN 76 237 237 1270 300 990 1 089 368 8 31 35 141 
13 JUN 76 208 208 1478 307 424 1 396 792 13 44 31 745 
23 SEP 76 182 182 1660 302 120 1 698 912 11 55 30 889 
7 OCT 76 127 126 1786 226 822 1 925 734 2 57 33 785 
32 890+ 965 
NOTE This corrmonly used method estimates the size of a population by taking the sum of all birds captured (A) , mutiplied by the 
number of birds already marked (B) , and dividing that by the sum of the number of marked birds captured . Like many methods 
Schnabel ' s  assumes that the population is stable . N I-' Ul 
2 1 6  
APPEN D I X  7 
VOCALIZATIONS OF THE WHITE-RUMPED SWIFTLET IN FIJI 
INTRODUCTION 
Following the discovery that some swiftlets have the ability to 
echonavigate in complete darkness (Medway 1959 , Novick 1959 ) , those 
species with the ability ,  were separated from the genus Collocalia 
(Medway & Pye 1977 ) , to form what has become the larger genus 
Aerodramus. The White-rumped Swiftlet A .  spodiopygius was placed in 
this genus , as Pecotich ( 1974) , had recorded that A. s .  terraereginae 
and A.  s .  chillagoensis produced the echolocatory "click" call . 
Subsequently this call and two other calls used by chillagoensis were 
described (Roberts et al 1976 ) .  
This paper records four new calls for the species . Three of these 
are from A .  s .  assimilis and one from chillagoensis .  It is confirmed 
that two of the previously described calls from chillagoensis are given 
by assimilis . Although .66 visits totalling over 300 hours , were made to 
five separate cave nesting colonies , the call given by Queensland birds 
at cave entrances ( Roberts et al 1976 ) was not heard . The "chirrup" 
call is probably the "twittering" calls currently being described in the 
common literature on the birds of Fi ji (Mercer 1966 ; Watling 1982 ; 
Clunie 1984) . 
Three of the calls described here have recognizably different 
functions , while two infrequently used calls will  be more difficult to 
ascribe a function. Two of the calls were heard outside of the cave as 
well  as inside . Three calls have only been heard inside caves and one 
has only been heard outside . 
METHODS 
Onometopoeic transcriptions of the calls were made during four 
separate days of continuous watching to determine chick feeding rates , 
2 1 7  
and during ten hours of watching adult feeding behaviour . Additionally 
two one hour visits were made to tape-record calls and similar 
transcriptions with qualifying descriptions were made from these . 
Narrow band sonagrams were made from these recordings using a digital 
Sona-graph 7800 (Kay Elemetrics Corp . ) .  Dry Cave ( in which the tape 
recordings were made ) and Waterfall Cave are at Nasinu 9 Mile , nine 
miles north of Suva . Ono Cave is  in the Wainibuka Valley , 40 miles 
north-west of Suva and Waiyala Cave is ? in the Sigatoka Valley 25 miles 
north of Sigatoka. 
RESULTS 
Echolocatory "click" call 
The most commonly produced call of the White-rumped Swiftlet is a 
distinctive "click" . This call is  used frequently in the twilight zone 
of the cave they use for roosting and breeding . The call is used 
continuously by birds flying in total darkness , as the call is used for 
echonavigation. The "click" call is rarely given by birds while 
clinging to the wall or being handled , and even then , it is  accompanied 
by a flapping of the wings , as Harrisson ( 1966 ) found in the Mossy-nest 
Swiftlet (Aerodrarrn.ls vanikorensis ) . This call is  characterised by a 
sharp wave front and rapid decay. Most energy is concentrated between 
four and five kHz, though it  ranges from one to eight kHz and on one 
sonagrarn a harmonic was visible at 14 to 15 kHz. 
The repetition rate of this call varies . From sonagrams for two 
birds the time from the commencement of one "click" to the begining of 
the next varied between 0 .09 seconds and 0 . 173 seconds . Further 
evidence of variation comes from counting calls from live birds and 
recorded calls . The mean rate was 4 .0  ? 0 . 23 per second ( x  ? SE , n = 
22 , range 1 . 4 - 5 . 9 ) . Birds flying swiftly along the narrow passage of 
Dry Cave , Nasinu , or those just frightened from their roost , appeared to 
call at a faster rate than those making repeated short flights from the 
F i gu r e 1 S on agrams  .o f a d u l t  c a l ls 
0 
0 0 .2  0 .4  0 . 6  0 .8  . 1 .0  1 . 2  1. 4 
8 
7 
6 
5 
? 4  
? 
3 
2 
1 
0 0 
8 
7 
6 
5 
? 4  
? 
3 
2 
1 
0 0 
se c onds 
# t\ r:?? ' p I I 
0.2 0 .4 
.? 
0.2 0 .4  
"Ch i r ru ?" c a l l 
0 .6  0 .8  
s e conds 
,,tl f 
"Cherp" c a l l  
0 .6  g.s 
secon s 
1 .0 1. 2 1 .4 
?? 
1 .0 1 . 2  1 .4 
.,. 
2 1 8  
wall in an effort to locate their nest . In Dry Cave it was evident that 
there was variation between individuals in the pitch of the call . 
In the larger caves , such as Ono , Waiyala , and Waterfall Caves ,  the 
cacophony just after sunset , resulting from many hundreds of flying 
birds giving the "click" call , is so great that one wonders how the call 
could ever have echolocatory usefulness . However , even when several 
thousand roosting birds were put to flight , the birds still managed to 
move through the cave , though movement .was slower . 
The "click" call of a fledgling on i ts first flight is higher in 
pitch and noticeably quieter than that of the adult .  However this 
"thinner" call is adequate in preventing juveniles from colliding with 
the cave wall and the sound of adult birds appears to help guide them 
towards the entrance instead of into other sections of the cave . 
The "chirrup" call 
The second adult call gives the impression of being more highly 
pitched than the "click" call . However the sonagram shows that the 
frequencies of' this call do not reach any higher than those of the 
"click" call . 
The most common form of the call is "chirrup , chirrup" , though I 
have heard "giddy up , giddy up" and "gar-p" . The call is used much less 
frequently than the "click" call and occurs most often upon the arrival 
of a bird at the nest . It is normally uttered by the bird already at 
the nest , as if in agonistic challenge or threat , to the new arrival and 
yet at the same time given to help identify itself as an aid to the 
orientation of the incoming bird . 
If the incoming bird replies and both birds vocalize this call 
simultaneously , the result is a sharp vocal exchange or squabble , best 
described as a screech (shown on the right of the sonagram) ,  which 
sometimes leads to a brief fight and the new arrival flying away. The 
"chirrup" call is not used often when feeding . I heard it only once , 
when a second bird joined in the 25 m diameter feeding circuit of the 
original bird . The call is used by birds wheeling around , above the 
entrance to a cave and is usually associated with a pair of birds 
involved in an aerial chase . 
2 1 9  
In Queensland the call used by chillagoensis in such aerial pursuits 
is a long "tweet-tweet-tweet-tweet-tweet-tweet" ,  "peer-peer-tweet" or a 
"tweer-tweer" that sounds like the scream that the Fi jian "chirrup" call 
sometimes ends with. 
The "shree-ee" call 
A call more shrill than any other call was given by chillagoensis 
when a flock had been dispersed from above a cave entrance by a Brown 
Goshawk Accipiter fasciatus . Some minutes later the flock reassembled 
above the goshawk occasionally giving a high pitched "shree-ee" call 
which I have not heard at any other time . I do not have a recording of 
this call . 
The "cherp" call 
This is the least common of the three adult calls given by assimilis . 
I have only heard it in the total darkness of a cave and then only 
rarely. It was given by a roosting or recently landed bird . The call 
was a mellow yet fast "cherp , cherp , cherp" , much softer than the 
"chirrup" call . 
The chick begging call 
Starting as a plaintive whisper then becoming a demanding "cheep" 
and culminating in a full fledged raspy call , each of the stages in this 
call only become louder and more harsh wi th age . Al ternatively , the 
call may start suddenly , close to full  volume . 
Begging , which accompanies this call is contagious and is often 
triggered by the begging of nearby chicks , the landing of an adult near 
by , or the close passage of an echolocating bird . 
8 
7 
6 
5 
? 4  
? 
3 
2 
1 
0 0 
8 
7 
6 
5 
N r. 4  
? 
3 
2 
1 
0 0 
F igu re  2 S o n ag ra m s  o f  ch i ck c a l l s  
,. ? .-.. -. ! h/ ' " I t 
I 
Begg_Lrrg c a l l  
0.2 0 .4 
??J\ r?? I I 
Pre s e nc e  c a l l 
0.2 0.4 0.6 0.8 
se con d s  
2 2 0  
The chick "presence" call 
This call is a single sof't "cheep" , "chip" or "peep" that is given 
(or at least noticeable )  on occasion when the cave is quiet . Whether it 
is some f'orm of' sub-song or just making one ' s  presence f'elt ,  I do not 
know. The pattern on the sonagram shows its similarity to the " cherp" 
call of' adults . 
DISCUSSION 
Echolocatory "click" call 
The echolocatory "click" call is termed the "rattle call" by Medway 
( 1959 , 1962a ,b ,  1966 , 1967 ) and by Medway and Wells 1969 ) , and has been 
proposed as the basis f'or separation of' those swif'tlets with the call 
into the genus Aerodramus Oberholser (Medway & Pye 1977 ) . 
With these calls , White-rumped Swif'tlets are able to detect rods 
down to 6 . 3  mm diameter (Griffin & Thompson ( 1982 , Smyth & Roberts 
1983) . These authors agree that the "click" call is not sufficiently 
perceptive to be used f'or nocturnal feeding , as the majority of swiftlet 
prey is below the size the birds can detect with the low frequency of 
these calls . The advantage of low frequency calls is that they do not 
attenuate as quickly as high frequency calls (Pye 1983) . These findings 
do not mean that swiftlets cannot f'eed at night . In situations where 
artificial illumination is sufficient , The White-rumped Swif'tlet 
(Tarburton 1987 ) , the Indian Edible-nest Swif'tlet Aerodramus unicolor 
(Ali & Dillon 1970) , the Alpine Swift Apus melba (Freeman 1981 ) and the 
Chimney Swif't Chaetura pelagica (Cottam 1932 ) , have each been observed 
feeding around lights at night . 
The ability of' the White-rumped Swif'tlet to echonavigate in the dark 
. .  ' ::?? . 
not only means that the birds may nest in relatively safe sites , but it 
also means they can feed in rich feeding areas far f'rom their nests 
during dawn and dusk , the times when aerial insects are the most 
abundant (Hespenheide 1975 ; Tarburton 1986 ) . I have observed birds 
2 2 1  
probably 20 km from their roosting caves, feeding actively until last 
light. I have also recorded birds coming into their caves as late as 
2230 hours. That most of the birds had left the caves by 0430 hours 
that morning means that their echonavigational ability allows them to 
make the most of the tropical day. The ability to echonavigate saves 
White-rumped Swiftlets from the problem that Common Swifts ( Apus apus ) 
have of not being able to reach their nests or chicks if they arrive at 
the nest site a few minutes after dark . (Lack 1956 ) . 
The echolocatory call is not only used for echolocatory purposes. 
It is sometimes given by adults feeding in bright light and so is 
presumed to have additional function. I have never heard a lone, 
feeding bird use it, but it may be heard soon after one feeding bird 
starts to chase another. This context suggests the call has a 
communication function and the differences I could detect between birds 
in the quietness of Dry Cave suggests that identification of individuals 
may be one function. It may be this call, that some mistake for the 
"clapping" or clicking of the birds bill as it persues insects (Mercer 
1966 , Watling 1982 , Clunie 1984) . 
The "click" call is a double click similar to that in terraereginae 
( Roberts et al. 1976 , Smyth & Roberts 1983 ) . There is a pulse at each 
end of the "click" so that it sounds "cli-ik", but is so fast that the 
two pulses are barely perceptible. The syringeal procedure for making 
this double click has now been determined , (Suthers & Hector 1982 ) . In 
effect the birds transform a longer squeek-like vocalization into two 
brief clicks by momentarily closing the syrinx in the centre of the 
call. The authors suggest that by so generating the brief clicks the 
swiftlets have created a sonal signal with a greater bandwidth , which 
having abrupt rise-decay times should improve target range determination 
based on measurement of the pulse-echo interval. 
Swiftlets using paired clicks , 20 ms or so apart , gain information 
additional to determining target range, for upon reflection not only the 
2 2 2  
pitch but also the time interval between the clicks will  be changed by 
the doppler shift , allowing target velocity to be determined (Pye 1983 ) . 
This should enable the birds to avoid other birds or ffits that might be 
flying in the same airspace . 
The importance of this call can be seen in the location of the 
majority of nests . Most nests and roost sites in Fiji and Chillagoe are 
protected from a number of potential predators by being in total 
darkness . 
The "chirrup" call 
It is evident that , in saying "their cry is a high pitched 
twittering which also appears to act as a type of sonar when the birds 
are flying in and out of the dark caves" , Mercer ( 1966 ) , is confusing 
the "chirrup" call with the function of the "click" call . Watling 
( 1982 ) also talks of a high pitched twitter , most commonly heard when 
used for echolocation in the dark caves . At least Clunie ( 1984) , leaves 
out the high pitched portion of the description when he writes of their 
using a twittering call to echo-sound in the caves . 
The "chirrup" call appears to serve as a greeting , helping to 
identify individuals .  If the incoming bird is not the mate , the call 
may develop into a threatening scream. In the Common Swift the threat 
and greeting calls are also very similar , though Lack ( 1956 ) found that 
the position of the head during vocalization was distinctive . 
The "cherp" call 
It is possible that this call corresponds with the plaintive piping 
call of the Common Swift , given by the loser of a fight over a nest site 
or a mate (Lack 1956 ) . 
The chick begging call 
Similar calls in the Chimney Swift ( Chaetura pelagica) are triggered 
by any disturbance (Amadon 1936 ) . 
2 2 3  
ACKNOWLEJ:X}EMENTS 
The helpful advice and criticism given by Drs . Edward Minot of 
Massey University and Hugh Ford of the University of New England in 
preparing this conmunication is gratefully acknowledged . Special thanks 
are also due to the Wildlife Division of C . S . I . R . O .  Australia for 
running the sonagrams shown in figures 1 and 2 ? 
REFERENCES 
ALI , S . ; DILLON , R .S . 1970 . Handbook of the birds of India & Pakistan : 
4 .  frogmouths to pi tta ' s .  Bombay. 
AMADON, D .  1936 . Chimney Swifts nesting in a barn . Auk 53 : 216-217 . 
CLUNIE , F .  1984. Birds of the Fiji bush . Suva. Fi ji Museum. 
COTTAM, C .  1932 . Nocturnal habits of the Chimney Swift . Auk 49 : 479-481 . 
FREEMAN, H .J .  1981 . Alpine Swifts feeding by artificial light at night . 
British Birds 74 : 149 . 
GRIFFIN , D .R . ; THOMPSON, D .  1982 . Echolocation by cave swiftlets . Behav. 
Ecol . Sociobiol 10 : 119-123 . 
HARRISSON, T .  1966 . Onset of echolocation clicking in Collocalia 
swiftlets . Nature 212 : 530-531 . 
HESPENHEIDE, H.A. 1975 . Selective predation by two swifts & a swallow in 
Central America. Ibis 117 : 82-99 . 
LACK, D .  1956 . Swifts in a tower . London. Muthuen . 
MEDWAY , L .  1959 . Echo-location among (Collocalia) . Nature 184 : 
1352-1353 . 
MEDWAY, L .  1962a. The swiftlets (Collocalia) of Niah Cave , Sarawak part 
1 .  Breeding Biology. Ibis 104: 45-66 . 
MEDWAY, L .  1962b. The swiftlets (Collocalia) of Java & their 
relationships . J .  Bombay Nat . Hist . Soc . 59 : 1-8 . 
MEDWAY, L .  1966 . Field characters as a guide to the specific relations 
of swiftlets . Proc . Linn. Soc . Lon . 177 : 151-177 . 
2 2 4  
MEDWAY, L .  1967 . The function of echonavigation among swiftlets . Anim.  
Behav . 15 : 416-420 . 
MEDWAY, L . ; Wells , D .R .  1969 . Dark orientation by the Giant Swiftlet 
(Collocalia gigas) .  Ibis 111 : 609-611 . 
MEDWAY, L . ; PYE , J . D .  1977 . Echolocation & the systematics of swiftlets . 
In Stonehouse , B .  & C .  Perrins . Evolutioary Ecology. London . 
Ma.cmillan. 
MERCER, R. 1966 . A field guide to Fiji birds . Suva. Fi ji Museum. 
NOVICK , A.  1959 . Acoustic orientation in the cave swiftlet . Biol . Bull . 
117 : 497-503 . 
PECOTICH, L .  1974. Grey swiftlets in the Tully River Gorge and Chillagoe 
caves . Sunbird 5 :  16-21 . 
PYE , J . D .  1983 . Echolocation and countermeasures ; In Lewis , B .  
Bioacoustics , comparative approach . pp . 407-429 . London . Academic 
Press . 
ROBERTS , J .R . ; SMYTH, D .M . ; SPAIN, A.V. 1976 . Preliminary studies on 
acoustic orientation in Collocalia spodiopygia.  Search 7 :  97-98 . 
SMYTH , D .M. ; ROBERTS ,J .R .  1983 . The sensitivity of echolocation by the 
Grey Swiftlet (Aerodramus spodiopygius) .  Ibis 125 : 339-345 . 
SUTHERS, R.A. ; HECTOR , D.H.  1982 . Mechanisms for the production of 
echolocating clicks by the Grey Swiftlet , ( Collocalia spoctiopygia) . 
J .  Comp. Physiol . ;  a sens Neural Behav. Physiol . 148 : 457-470. 
TARBURTON, M.K.  1986 . The food of the Whi te-rumped Swiftlet (Aerodramus 
spodiopygius) in Fiji . Notornis 33 : 1-16 . 
TARBURTON, M .K .  1987 . The population Status , longevity and mortality of 
the White-rumped Swiftlet in Fi ji . Corella 11 :  in press . 
WATLING,  D .  1982 . Birds of Fiji , Tonga & Samoa . Wellington. Millwood 
Press . 
M.K.  TARBURTON. Department of Botany & Zoology, Massey University ,  
Private Bag , Palmerston North , New Zealand .