A shift to proactive interference with increased set size for dual-task working memory consolidation

Working memory (WM) dual-task studies report either retroactive interference of a later decision task (T2) on WM consolidation (e.g., Nieuwenstein & Wyble, Journal of Experimental Psychology: General, 2014), or proactive interference of WM consolidation on T2 (e.g., Jolicoeur & Dell’Acqua, Cognitive Psychology, 1998). The factors mediating between obtaining proactive versus retroactive interference in such WM paradigms are poorly understood. Recently, we replicated the retroactive interference effect in a series of tasks based upon that of Nieuwenstein and Wyble (2014). Here, we closely replicated a task that typically yields retroactive interference, with the simple modification of an increased memory set size. Increasing the WM sample from 4 to 6 visually-presented letters abolished retroactive interference but led to clear proactive interference. We also manipulated task priority across participants. Counterintuitively, retroactive interference was observed when the WM task was prioritized, while a Bayesian ANOVA yielded evidence against retroactive interference when T2 was prioritized. However, the presence of retroactive alongside proactive interference in the WM-priority group must be interpreted cautiously because the interaction of retroactive interference and priority group did not itself achieve statistical significance. Taken together, these results suggest an account for when one can expect proactive versus retroactive interference. Specifically, when demands on processes underlying consolidation of perceptual items into WM are high, insufficient slack capacity is available for a subsequent task to be processed sufficiently to interrupt consolidation. Instead, T2 decision processes are deferred, leading to proactive interference. However, when WM encoding demands are low, sufficient slack capacity exists for a high-priority (speeded) T2 decision to be processed sufficiently to override ongoing consolidation processes, leading to retroactive interference. Regardless of the specific account, high WM set sizes yield primarily proactive interference on T2 rather than retroactive interference on WM encoding.