Sequential dependencies in the Eriksen flanker task: A direct comparison of two competing accounts

Abstract

The Eriksen flanker task (Eriksen & Eriksen, 1974) has been frequently used to investigate attentional control processes. The basic task requires a fast response to a centrally presented target stimulus, which is flanked by several distractor stimuli that also activate response channels. A congruent trial is one in which the flankers are associated with the same response as the target (e.g., ), whereas in an incongruent trial, the flankers are associated with a competing response (e.g., ). Reaction times (RTs) are faster for congruent than for incongruent trials. The dominant explanation for this finding is that the presentation of flankers results in automatic activation of the response channel associated with the flanker stimuli (Gratton, Coles, & Donchin, 1992;Gratton, Coles, Sirevaag, Eriksen, & Donchin, 1988), leading to fast correct responses in the congruent condition (i.e., the flankers are identical to the target and facilitate a correct and fast response). In the incongruent condition, the automatic activation leads to fast incorrect responses, necessitating attentional control processes to overrule the incorrect automatic activation, resulting in slow correct responses.This generally accepted view has been implemented in sophisticated computational models to investigate the dynamics of the control processes. One such model is the conflict-monitoring theory of Botvinick, Braver, Barch, Carter, and Cohen (2001), in which the automatic activation of the incorrect response channel competes with activation of the correct response channel at the level of response selection. Resolving this response competition requires time and thus underlies the main effect of congruency-the flanker effect.One of the important contributions made by the Botvinick et al. (2001) model is that the amount of response competition, or conflict, is monitored and used to adjust top-down attentional control-the more conflict, the more control. In the case of the flanker task, increased control results in an increased attentional focus to the central target stimulus. Botvinick et al. proposed that conflict monitored on the previous trial is used to modulate cognitive control on the current trial and that this sequential dependency across trials underlies the so-called Gratton effect (Gratton et al., 1992). This effect is an interaction between the trial type (congruent vs. incongruent) of the current trial and the trial type of the previous trial (see Figure 1). In this article, we will use the notation xX, where the first letter-in lowercase-indicates the congruency of the previous trial and the second letter-in uppercaseindicates the congruency of the current trial. The Gratton effect is a pattern of slower RTs for congruent trials that follow incongruent trials (iC), as compared with those that follow congruent trials (cC), and faster RTs for incongruent trials that follow incongruent trials (iI), as compared with those that follow congruent trials (cI). To prime the reader, the RT difference between cI and iI trials tends to be la...

No conflict control in the absence of awareness

“…According to domain-general theories of conflict control, conflict in one dimension (or task) can elicit control of conflict in an alternative dimension (or task, cf. Botvinick, Cohen, & Carter, 2004;Kunde & Wühr, 2006; but see, e.g., Davelaar & Stevens, 2009;Kiesel, Kunde & Hoffmann, 2006). Here, we tested whether conflict elicited in one domain by relations between the visible target's meaning and the responses affected the regulation of the conflict in an alternative domain, that is, between invisible prime and target.…”

Now you see it, now you don't: Controlling for contingencies and stimulus repetitions eliminates the Gratton effect

“…An overview of the literature suggests that most Gratton experiments are confounded by either stimulus binding effects (e.g., Akçay & Hazeltine, 2007Davelaar & Stevens, 2009;Egner, Delano, & Hirsch, 2007;Funes, Lupiáñez, & Humphreys, 2010b;Kerns et al, 2004;Notebaert et al, 2006;Ullsperger, Bylsma, & Botvinick, 2005, Experiment 1;Wendt, Kluwe, & Peters, 2006) or contingency biases (e.g., Akçay & Hazeltine, 2007;Mayr & Awh, 2009;Ullsperger et al, 2005, Experiment 2;Verbruggen et al, 2006). Contingency biases are less common in the (typically two-choice) flanker tasks, but stimulus binding effects are much more difficult to control for in these two-choice flanker tasks, for reasons we will discuss later.…”

“…We do note, however, that 30% congruent items in a six-choice task is still well above chance (16.7%), thus not eliminating all contingency biases. Some studies do present congruent trials no more often than expected by chance, particularly in two-choice tasks (e.g., Davelaar & Stevens, 2009). However, these studies do not control for stimulus bindings (and cannot do so with a two-choice task).…”

Context, not conflict, drives cognitive control.

“…Studies using various other response conflict paradigms found Gratton effects even when controlling for stimulus/response repetitions (e.g., Akcay & Hazeltine, 2007;Kerns et al, 2004;Kunde & Wühr, 2006;Ullsperger, Bylsma, & Botvinick, 2005). To accommodate these and other diverging results, more recent models of behavioral control and context adaptation favor a hybrid approach, combining associative bottom-up learning mechanisms with conflict-driven top-down control (e.g., Davelaar & Stevens, 2009;Verguts & Notebaert, 2009). An entirely different model has been proposed by Scherbaum, Dshemuchadse, Fischer, and Goschke (2010), who suggest that conflict adaptation reflects not a proactive mechanism, but the conflict-triggered reactivation of previouslyused conflict resolution processes.…”