This experiment tested subjects' memory for verbal (Kanji and Kana) and figural displays by using a reaction time procedure. There were three conditions: stimulus (figure, Kanji, Kana), configuration (1-dimension, 2-dimensions), and type (identity, transformed, different). After a 2-dimensional array of three stimulus components was presented as an original stimulus, subjects were presented with one of six (2 configurations×3 types) comparison arrays. Subject's task was to verify that the comparison array contained the same elements as the original stimulus. The factor of the configuration had a significant effect on RTs for the figural stimulus, but not for the verbal stimulus. The transformed array had significant effects for both figural and verbal stimuli except for the figural stimulus in 1-dimension condition. There was no difference between Kanji and Kana in all of the conditions. The results demonstrated that short-term verbal representations are both sequentially and spatially constrained, and that representations of figures preserve the spatial relationships of components and are not so constrained by sequential ordering. The data are discussed in terms of a dual-coding model.