Using words of known selection values arranged in a counterbalanced order, it was shown that subjects' ratings of the "imagery value" of each word resulted in error which favored the first position in the series when stimuli were homogeneous. The error did not occur with words which were markedly heterogeneous. (Author)