Beyond feature binding: interference from episodic context binding creates the bivalency effect in task-switching

Abstract

When switching between different tasks and bivalent stimuli occur only occasionally on one of them, performance is slowed on subsequent univalent trials even if they have no overlapping features with the bivalent stimulus. This phenomenon has been labeled the "bivalency effect." Recent evidence has revealed that this effect is robust, general, and enduring. Moreover, it challenges current theories of task-switching and cognitive control. Here, we review these theories and propose a new, episodic context binding account. According to this account, binding does not only occur between stimuli, responses, and tasks, but also for the more general context in which the stimuli occur. The result of this binding process is a complex representation that includes each of these components. When bivalent stimuli occur, the resulting conflict is associated with the general context, creating a new conflict-loaded representation. The reactivation of this representation causes interference on subsequent trials, that is, the bivalency effect. We evaluate this account in light of the empirical evidence.

Keywords: anterior cingulate cortex; binding; bivalent stimuli; cognitive control; univalent stimuli.

Figure 1

(A) Example of the basic paradigm used to test the influence of occasionally occurring bivalent stimuli. A task-triplet comprised a parity decision, a color decision, and a case decision. On a bivalent task-triplet (not pictured here), the letters were presented in color (either in blue or red). (B) Experiment structure with three blocks of task-switching; bivalent stimuli occur only in block 2 on 20% of case decisions. (C) Decision times for univalent stimuli across blocks (blocks 1 and 3 pure, block 2 mixed), on parity decisions (circles), color decisions (squares), and case decisions (triangles). (D) Bivalency effect, expressed as performance difference on univalent stimuli between block 2 and blocks 1 and 3 averaged. Error bars represent standard errors. Adapted from Woodward et al. (2003).

Figure 2

Endurance of the bivalency effect: Mean decision times for task-triplets following a bivalent case decision in the mixed block (closed circles) compared with the corresponding task-triplets in the pure block. Error bars represent standard errors. Task-triplet N refers to the task-triplet containing a bivalent stimulus in the mixed block; subsequent task-triplets (represented here) are labeled N + 1, N + 2, N + 3, and N + 4, respectively. Results adapted from Meier et al. (2009), averaged across experiments, experimental conditions, and tasks.

Figure 3

Invariant bivalency effect across response-set conditions depicted as DT difference between univalent stimuli from the mixed block and the average of the pure blocks. Results adapted from Rey-Mermet and Meier (2012a) averaged across experiments.

Figure 4

Conflict specificity of the bivalency effect (task 1 refers to the task with overlapping stimulus features, task 2 refers to the task with no overlapping stimulus features, and task 3 refers to the task that occasionally involved bivalent stimuli). (A) Decision time data (i.e., performance on univalent stimuli for switch and repetition trials in pure and mixed blocks). (B) Bivalency effect (i.e., difference between univalent trials from the pure block and those from the mixed block). Results adapted from Rey-Mermet and Meier (2012b) averaged across experiments and tasks.

Figure 5

Neural structures underlying the bivalency effect. (A) Dorsal anterior cingulate cortex signals the requirement to adjust cognitive control (cf. Woodward et al., 2008). (B) Hippocampus (and other memory-related structures not depicted here) are required for episodic binding and the reactivation of the episodic context (cf. Meier et al., submitted).