This study was designed to explore direct and vicarious reinforcement of cognitive skill. Cognitive skill, in this case, was defined by specific types of answering behaviour. In developmental stages of the study, a taxonomy of educational objectives was found unsatisfactory for classifying answers given by subjects. For this reason, a classification scheme was developed for use in the present study. This scheme supplied six categories of cognitive skill. Subjects' answers were classified according to which of the six cognitive skills they represented. Developmental work also explored a variety of types of questions (completion, multiple-choice, and open-ended) in an attempt to determine which were efficient at eliciting answers representing a variety of cognitive skills. Open-ended questions appeared most satisfactory because they allowed free manifestation of a variety of types of answers rather than predetermining the cognitive skill represented by answers, as did completion and multiple-choice items. However, with open-ended questions, there was an accompanying difficulty of eliciting answers which represented specific categories of response so that consistent stimuli could be provided for observing subjects. Since some categories had low operant levels, a videotape was used to provide standardized and consistent stimuli for observers. The modelling situation did not permit the exploration of direct reinforcement so the emphasis of the study turned to vicarious reinforcement. In an experimental situation, subjects (girls aged 15 to 16 years) watched a model giving "answers" which corresponded to all six categories of the classification scheme. Some of the model's "answers" were praised and others were not. Subsequent measures of the observing subjects' responses tested the effect of vicarious praise. Measures were also taken to determine the effect of awareness of contingencies of reinforcement on subjects' responses. Furthermore, alterations in the experimental design were used to explore the effects of increased intensity and relative position of praise in the modelled sequence. Exact probabilities were used to analyze the data where possible. However, where there was more than one degree of freedom, a chi square statistic was used to determine the significance of results. The results showed that vicarious praise did influence the types of answers given by observing subjects. That is, the praise was an effective vicarious reinforcer. However, the order of the model's "answers" and awareness on the part of subjects were unrelated to responses given by subjects. Moreover, increased intensity and change in position of the praise did not produce greater treatment effects. Alternate methods of classifying the data were explored but not found to be more efficient than the original scheme at detecting a treatment effect. Also, an unsuccessful attempt was made to identify characteristics of subjects which affected their answering behaviour. The results are discussed as they relate to both education and psychology, and implication for both fields are considered.