We investigate the influence of biased feedback on decision and learning pr
ocesses in a vernier discrimination task. Subjects adjust their decision cr
iteria and hence their responses according to biased external feedback. How
ever, they do not use learning processes to encode incorrectly classified s
timuli. As soon as correct feedback is restored observers regain their orig
inal performance indicating an involvement of internal criteria. If the ext
ernal feedback is switched off instead of being corrected, the rebound is l
ess vigorous. The findings contradict predictions of supervised neural netw
ork models. (C) 1999 Elsevier Science Ltd. All rights reserved.