The aim of the present study is to analyze the characteristics of recognition errors of letters by halfly-seeing children in comparison with those by children with normal eye-sight. This study consists of two experiments. In the first experiment, the experimental subjects of halfly-seeing children were fourteen children in the first grade of an elementary school, while controlled subjects were thirteen with normal eye-sight. The recognition errors of letters were tested by cancellation tests by work limit method. The procedure was the same as that of the second experiment. Nine target letters were some of Japanese characters: う (u), は (ha), る (ru), わ (wa), い (i), そ (so), へ (he), め (me) and お (o). Below each target letter, the same letters and other letters of a similar form, both ten in number, were arranged at random. The subjects cancelled the letters which they recognized as the same as the target letter. According to non-parametric U-test, they had significantly more confusion of recognition between は (ha) and ほ (ho), る (ru) and ろ (ro), め (me) and お (o) and お (o) and あ (a) than the controlled subjects. In the types of errors, however, there was no significant difference between the two groups. Through factor analysis by the complete centroid method and Varimax rotation, four factors were extracted which were Xo, Yo, uncancelling and failure of recognition of details. In this analysis the subjects were four halfly-seeing children and seventy-nine four-and five-year-old children with normal eye-sight. In the case of the four-year-old children only with normal eye-sight, the extracted factors were reversal errors, Xo, confusion of recognition in the case of the letters of a similar form as a while, failure of recognition of details and uncancelling. In the second experiment by one-group method, subjects were eight halfly-seeing children in the first grade of an elementary school. Target letters were all forty-six Japanese characters. In each experimental situation, all forty-six characters includinding the target letter and four more target letters were arranged below every target letter. The recognition errors were fewer than those of the first experiment, but compared with the other types of errors, there were more uncancelling error and recognition errors of the letters of a similar form as a whole.
