Investigated whether conservative human inference which has been attributed to misperception or misaggregation of data may be caused by response biases. In experiments, 26 and 29 male undergraduates revised odds estimates about which 1 of 2 normal distribution data generators was being sampled. An analysis of special sequences and a plot of revised odds against theoretical odds in Exp. I showed a bias in Ss' response functions. They revised odds optimally only over a range of 1.0 log odds. When E set different levels of prior odds, the response functions shifted so that the optimal range centered around the set prior odds. Exp. II showed that the biased functions remained invariate over changes in data generator familiarity and diagnosticity. Of the several explanations offered for these response functions, an odds bias seems the most likely. Whatever the cause of the bias, Ss neither misaggregated nor misperceived data within their optimal range. (PsycINFO Database Record (c) 2006 APA, all rights reserved)