Tests of paced serial addition are used to detect diminished information-pr
ocessing capacity. Nonetheless, performance might be influenced by modality
-specific interference or by variables that specifically affect numerical p
rocessing. In a series of three experiments with normal adults, we manipula
ted, respectively, the modality in which addends were presented, the modali
ty in which responses were produced, and the format in which visual addends
were displayed. Performance was enhanced when stimuli were presented visua
lly and when responses were made manually. When visual addends were used, A
rabic numerals were processed more effectively than number words. Thus, per
formance was influenced by modality-specific interference and by presentati
on format. We conclude that paced serial addition tasks may not provide a p
ure measure of general information-processing capacity.