Forty third-graders and an equal number of sixth-graders listened to a list of words and for each word had to indicate, by saying "old" or "new," whether it had appeared before on the list or not. The subjects gave more erroneous "old" responses to words which were semantically or phonetically related to previously heard words than to control words. Furthermore, they took longer to respond correctly ("new") to the experimental words than to the controls. Experimental words which were highly associated to the preceding words did not produce more errors than words with little association to the preceding words. The semantic relations were relatively less effective in producing errors for the third-graders than for the sixth-graders. These findings are interpreted to favor the hypothesis advanced by Anisfeld and Knapp (ED 019 639) that false recognition results can be more adequately and directly explained by a logical analysis of the semantic and phonetic features common to words than by free association norms. The relatively lesser effectiveness of semantic relations in producing false recognition errors in the third-grade subjects than in the sixth-grade subjects suggests that at the younger age the semantic features for interrelating vocabulary items are not yet as prominent as the more superficial phonetic features. (Author/JD)