Prior research into expectations about counselling has assumed that failing to meet client expectations will have a detrimental effect on process variables such as state anxiety and adherence to treatment. However, the empirical support for this is equivocal. Both self-regulation theory and the attentional-bias model suggest that experiencing the confirmation of accurate, but negative expectations will result in an increase in state anxiety. Therefore, instead of focussing only on improving accuracy of client expectations it is suggested that the affective valence of the expectations must also be acknowledged. The aim of the present study was to investigate the differential effects of confirmation/disconfirmation of positive or negative expectations on anxiety about seeing a counsellor. Thirty-nine adult clients attending their first session at a university counselling centre completed pre- and post-session measures which assessed their expectations about counselling, and state and trait anxiety. As hypothesised, the effects of disconfirmation of expectations on state anxiety were moderated by the valence of the expectations. However, contrary to what was predicted, those client's who experienced confirmation of negative expectations did not display greater state anxiety than those with positive expectations, and there was no significant relationship between trait anxiety and negative expectations. Simple exposure to therapy resulted in a decrease in state anxiety for all clients regardless of confirmation/disconfirmation and expectation valence. Finally, those clients who had their negative expectations confirmed did not drop out of treatment more than any other group. A preliminary investigation of the validity of using the Expectations About Counseling questionnaire (EAC) to measure positive and negative expectations about counselling revealed that it was not as effective as had been suggested by previous researchers. It was concluded that this factor and a lack of power contributed to the paucity of significant results. The results are discussed in the context of self-regulation theory and the attentional-bias model and an argument is made for the continued use of these two paradigms in future research.