Working memory’s workload capacity
DOI: 10.3758/s13421-015-0526-2
archive: archived pipeline: cataloged verified
Summary
Lab study (Heathcote, Coleman, Eidels, Watson, Houpt, Strayer; U. Tasmania + U. Newcastle + U. Utah + Wright State; Mem Cogn 2015, doi:10.3758/s13421-015-0526-2) developing and validating a novel dual two-back task (auditory + visual sequences with redundant-target responses) to measure working memory's workload capacity using Townsend & Nozawa's (1995) systems-factorial-technology framework. Participants (N=311 after exclusions, recruited online) performed three practice blocks plus 16 experimental blocks of 27 trials, comparing single (auditory-only or visual-only) and dual (both modalities, redundant target) two-back performance with 50% target rate. The redundant-target design enabled formal computation of the workload-capacity coefficient C(t). Decrements in dual-modality two-back accuracy and speed were mediated by limited shared processing capacity, with high and constant proactive interference from a small stimulus set yielding more reliable performance metrics than standard n-back. A response-bias-minimized version produced stronger correlations with operation-span (OSPAN) than has been reported for other n-back variants.
Key finding
Working memory shows clear capacity sharing between simultaneously maintained auditory and visual two-back streams: the new redundant-target dual two-back yields formal Townsend-Nozawa workload-capacity coefficients indicating sub-unlimited capacity, more reliable performance metrics (due to deliberately-induced high proactive interference), and stronger convergent validity with complex-span measures than standard n-back paradigms — making it a candidate cognitive-load benchmark.
Methodology
Lab cognitive task. 311 participants (after exclusions from 372 due to software errors) recruited online; 147 in 50%-target group and 164 in another. Within-subject single-modality and dual-modality two-back conditions. Townsend-Nozawa systems-factorial-technology applied to compute workload-capacity coefficient C(t). Convergent validity assessed against operation-span (OSPAN).
Sample size: N=311 (final sample after exclusions; 147 in 50% group, 164 in other).
Quality score: 5 / 5