Errors in a recognition memory task are a U-shaped function of word frequency

College students rated verbal items for either familiarity or meaningfulness and were then tested for their recognition memory of rated items from four critical classes: nonwords and words appearing less than 1 time, 1–10, and more than 40 times per million words of written text (Thorndike-Lorge, 1944). Following either rating task, recognition errors were found to be a U-shaped function of word frequency and nonwords were recognized least well. Although a U-shaped function was expected on the basis of what is known about word frequency effects in general, there is significant disagreement as to the mechanism(s) responsible for this relationship.