An attempt was made to examine the encoding of visually presented additive and nonadditive digital material and its processing by deaf and hearing Ss. A 2 X 3 X 4 within-Ss design consisting of groups (45 deaf and 45 hearing Ss), numeral sets (additive, lower additive, and additive random), and recall intervals (0, 3, 9, and 12 seconds) was used. Results showed better recall of the additive set in which the third digit was the sum of the first two (e.g., 5-3-8) than other sets. Although the hearing Ss recalled better than the deaf Ss at short delay, performance of both groups deteriorated with increase in retention intervals. Implications of this study are related to the problem of automatic processing in short-term recall and how it is sensitive to the structure of the sequence of a numeral set.