There are at least two processes in leaning to name letters: learning to differentiate graphic features of letters(perceptual discrimination learning)and learning to code letters to their sounds(associative learning). In the present study two experiments were reported which focused on the process of perceptual discrimination learning. Experiment I explored the relationships among the following three factors:the discrimination ability of letter-like figures, that of Japanese kaktakana letters, and the rate of naming katakana letters. Experiment II was designed to examine effects of discrimination training of figures and letters upon learning to name katakana letters. Experiment I The Ss were 92 nursery school children with a mean age of 4 years, 9 months. They could not name any of the letters used in the experiment. The figure and letter discrimination tasks were provided, each of which consisted of four sets of four figures or four letters(FIG.1). In each set one of the four stimuli was put on the upper center of a white card and the four stimuli were put on the bottom half in a row(FIG.2). The S was given these 32 cards one by one in a mixed series of the figure and letter sets, and was required to point to one which was just the same as the upper one, out of the bottom four stimuli. No information was given about the S's responses. The letter-naming task consisted of four two-letter words, each of which was written on a white card. In the study trial the S was given these words one by one with the E naming the letters and was required to repeat after the E. In the test trial the S was given the letters without the E naming them and was required to name them. Two study trials and a test trial were given seven times alternatively. The correlation coefficients were statistically significant for all comparisons(TABLE 2). This suggests that S's ability to discriminate figures and letters was positively related to the rate of naming letters. The stimuli of Set 1 were more discriminable than those of Set 4 for the figure and letter sets(TABLE 3). Experiment II The Ss were 60 nursery school children with a mean age of 5 years, 3 months. They could not name any of the katakana letters used in the experimant. Letter Set I, Letter Set 4, and Figure Set 4 were used as the discrimination tasks. The S was trained on one of these three sets with the E's information of "Hit" and "Miss". After reaching a criterion of eight successive correct choices, the S was given a letter-naming task which consisted of the four katakana letters employed in Letter Set 4. The procedure for learning to name letters was nearly the same as that in Experiment I. Letter Set 1 was learned faster than the remaining two sets and the latter did not differ significantly(left half of TABLE 4). This indicates that discriminability was positively related to the rate of discrimination among stimuli learning. The facilitative effect of discrimintion upon naming letters was more conspicuous when the S had been trained on the less discriminable sets than when the S had been trained on the more discriminable set(right half of TABLE 4). The fact that there was no differetial effect of Figure Set 4 and Letter Set 4 upon naming letters can be explained by assuming that Ss acquired some general set to attend to and discriminate stimuli through the discrimination training of less discriminable figures.