The present study affords an explanation for the consistent, but not always statistically significant, pattern showing superior verbal discrimination learning performance for low- as compared to high-frequency words. In a frequency judgment task it was found that relative to high-frequency words, low-frequency words for which subjects (sixth graders) knew the meanings produced apparent frequency measures consistent with superior verbal discrimination learning, while low-frequency words that were known to the children did not. The results, taken together with those based on comparisons of pictures and high-frequency words, lend themselves to a modified Weber's law interpretation of stimulus material differences in discrimination learning. (Author)