Performance on paced serial addition tasks indicates an associative network for calculation

Although paced serial addition (PSA) tasks are considered to be tests of general information-processing capacity, recent work suggests that performance on such tasks is influenced by arithmetic-specific variables. We designed two visual PSA experiments to determine whether the performance of normal adults would support predictions derived from the cognitive psychology of calculation. Experiment 1 showed that mixing familiar (Arabic numeral) and less familiar (Roman numeral) stimulus formats reduced scores below the averaged scores for pure Arabic and Roman lists. The Roman-Arabic order of addends was more difficult than the Arabic-Roman order. Experiment 2, which involved only Arabic numerals as addends, showed that performance could be impaired by constraining the trial-to-trial variability of sums. The results of both experiments confirm the importance of arithmetic-specific variables in PSA and provide support for an associative network model of calculation. In addition, the findings implicate interference from extraneous addends and responses as the performance-limiting factor.