Software estimation research is normally concerned with designing models and techniques that help estimators reach accurate effort calculations. However, since software estimation involves human judgment, we should also consider cognitive and social factors that have been observed in psychological studies on judgment. A potentially relevant factor is the anchoring and adjustment cognitive bias. It takes place if, when attempting to respond a complex question, the respondent is given a possible---though quite likely incorrect---answer. The respondent adjusts it internally to reach a more plausible answer, but the adjustment is frequently insufficient.This thesis presents the results of an experiment about anchoring and adjustment in software estimation. Results show that anchoring and adjustment changes the outcome of software estimation processes. They also suggest that estimators tend to have too much confidence in their own estimations.