Binding Error-Induced Control States

Abstract

Binding and retrieval of stimulus features, response features, and their attentional weighting tune cognitive processing to situational demands. The two mechanisms promote successful actions, especially in situations in which such actions depend on controlled processing. Here we explored binding and retrieval of attentional control states that follow from erroneous actions. By definition, such errors are characterized by insufficient cognitive control but at the same time, error detection has been shown to trigger corresponding adjustments to prevent future failures. We reanalyzed existing datasets and conducted a novel experiment to investigate whether error-induced control states become bound to task-relevant stimuli. Results point towards a binding and retrieval of error-induced control states; however, the effect appears to be less reliable than for binding and retrieval of specific stimulus and response features. We discuss potential implications and alternative interpretations in terms of a mediating impact of error-induced control.

Keywords: binding and retrieval; cognitive control; error processing.

Figure 1

Exemplary trial sequence with proposed control state changes, and hypothesized binding and retrieval processes. We examined sequences of three trials to probe for binding and retrieval of error-induced control states. On each trial, participants responded with a left or right response to a target letter that was surrounded by irrelevant letters. The arrangement of letters corresponds to the design of the Confirmatory Experiment. Stimuli are not drawn to scale for legibility. Assumed representations of the features of an action episode are shown as circles. Binding is illustrated through lines between feature representations and retrieval through arrows. In trial n–2, a commission error (orange) should lead to increased control (Ctrl) in comparison to correct responses (grey). The resulting control state then binds to the presented stimuli (S) in trial n–1 as do features of the executed correct response (R). A repetition of the relevant stimulus in trial n is assumed to retrieve the previously bound control state and response (top), whereas retrieval does not take place for stimulus changes (bottom).

Figure 2

Mean response times of Reanalysis 1 and 2. Mean response times as a function of the response in n–2 (correct in black, error in bright orange) and whether correct response repetitions had to be executed for target repetitions or target changes in (A) Reanalysis 1 and (B) Reanalysis 2 for which we averaged over feedback conditions here. Error bars represent the standard error of the differences, separately computed for correct and erroneous responses in n–2.

Figure 3

Mean response times of the Confirmatory Experiment. Mean response times as a function of the response in n–2 (correct in black, error in bright orange) and whether correct response repetitions had to be executed for target repetitions or target changes in the Confirmatory Experiment for (A) unmatched data and (B) data that was matched for effects of the response in n–2 on RT_n–1. Error bars represent the standard error of the differences, separately computed for correct and erroneous responses in n–2.