Goal-directed behavior is thought to be supported by a class of cognitive processes, collectively referred to as attentional control. In the laboratory, attentional control is frequently investigated, employing conflict task protocols. These protocols, such as the Eriksen flanker task (Eriksen & Eriksen, 1974), require the participants to respond to task-relevant stimulus features while ignoring task-irrelevant stimulus features. In this task, the participants are instructed to respond to a stimulus presented centrally (i.e., target) surrounded by peripheral stimuli (i.e., distractors) which are either assigned to the same response as the target (i.e., congruent trials) or are assigned to a different response (i.e., incongruent trials). Analyzing performance in these protocols typically reveals that responding in congruent trials is, on average, faster and less error-prone than responding in incongruent trials. The Congruency Effect (i.e., the performance difference in congruent and incongruent trials) is broadly considered a measure of the influence of task-irrelevant stimulus features on performance. Specifically, it has been attributed to the co-activation of the distractor-related response in addition to the target-related response. Assuming that target- and distractor-related responses compete for response selection and execution, such co-activation should result in a (response) conflict in incongruent trials (Eriksen & Eriksen, 1974; for an overview, see Eriksen, 1995). Evidence for distractor-induced response activation has, for instance, been found in reach tracking studies in which participants are instructed to make choice responses by moving their hand to one of several target locations. In such studies, prolonged reach curvatures (diverted trajectories towards the distractor-related response’s target position) in incongruent relative to congruent trials have frequently been observed (e.g., Erb et al., 2016; Erb & Marcovitch, 2018; Scherbaum et al., 2010; Scherbaum & Dshemuchadse, 2019).
Analyzing trial sequences in conflict tasks, a large number of studies found that the congruency level of the predecessor trial modulates the congruency effect. Typically, the congruency effect decreases following an incongruent trial relative to following a congruent trial (i.e., congruency sequence effect, CSE; e.g., Gratton et al., 1992; Schmidt & Weissman, 2014; Weissman et al., 2014; Weissman et al., 2015). Consistent with the notion that the congruency effect is mainly brought about by conflict in incongruent trials, these findings suggest that the degree of conflict in incongruent trials is lower after incongruent than after congruent predecessor trials. This modulation of the degree of conflict is commonly interpreted as evidence for enhanced attentional focusing induced by conflict. According to the conflict monitoring account (Botvinick et al., 2001; Botvinick et al., 2004), a monitoring unit computes each trial's conflict level. A higher level of conflict is supposed to lead to a higher degree of processing selectivity via attentional adjustment (i.e., increasing or decreasing the bias in attentional weights regarding processing the target or distractor) compared to a low level of conflict.
Over the years, it has become increasingly clear that the CSE is subject to various confounding factors, which motivated alternative, non-attentional interpretations of the CSE. Although some of these confounds have been thoroughly discussed in the literature, and efficient methods of controlling them have been proposed (see Braem et al., 2019, for an overview), other confounds are still widely neglected in current experimental studies. In the following, we will first present a brief overview of the former and their remedies and then address two less prominent confounds, one of them being confined to a particular (albeit widely used) kind of experimental task, the other one being inherent in all CSE procedures.
In standard conflict-task protocols, congruency level repetitions tend to involve a larger proportion of trials associated with a complete repetition or a complete alternation of all discriminative stimulus and response features than congruency level alternations, which, in turn, tend to involve a larger proportion of partial repetitions (i.e., repetition of one stimulus or response feature and alternation of another one). Given that partial repetitions are likely to induce interference with memory episodes of previous trials – as laid out in detail in Hommel’s feature integration account (e.g., Hommel et al., 2004) – obtaining a CSE in such conditions may be unrelated to attentional adjustment. This confound has frequently been controlled by a minimum set of four stimulus objects, used as targets and distractors, mapped to four different responses and excluding all trials involving any repetition of a discriminative stimulus or response feature from the analyses. Although the application of this procedure usually failed to eliminate the CSE (e.g., Akçay & Hazeltine, 2011; Bugg, 2008; Hazeltine et al., 2011; Kunde & Wühr, 2006; Purmann et al., 2009; Tomat et al., 2020; Ullsperger, Bylsma, & Botvinick, 2005), using a larger set of stimuli and responses than in the initial CSE studies (e.g., Gratton et al., 1992; Botvinick et al., 1999) raised another problem. As unconstrained (random) stimulus selection might produce only small amounts of congruent trials (i.e., an expected value of 25% in a task involving a 4:4 S-R mapping), researchers tended to bias stimulus selection in favor of congruent trials to produce a more balanced distribution of congruency levels (and of congruency level sequences). Biasing stimulus selection, however, introduces different distractor-related contingencies for congruent and incongruent stimuli. Precisely, individual distractors, occurring more frequently in congruent than in incongruent trials, coincide more often with the “congruent response” (i.e., the response required if the distractor stimulus was presented as target) than with the other (“incongruent”) responses (Schmidt & de Houwer, 2011). Consequently, distractors presented in congruent trials are more predictive of the correct response compared to distractors presented in incongruent trials. Associative learning of these contingencies should facilitate high contingent/congruent trials relative to low contingent/incongruent trials. This assumption has been supported by a corresponding contingency effect obtained with neutral distractors (i.e., stimuli used as distractors that are not assigned a response and never occur as targets). Specifically, Schmidt et al. (2007) administered a variant of the Stroop task, using four different colors as targets and four neutral words, such as move, each of which was predominantly presented with one of the colors (i.e., high contingency) and only rarely with each of the other colors (i.e., low contingency). Of importance for the interpretation of the CSE, Schmidt et al. (2007) not only observed faster responses for the high-contingency stimuli than for the low-contingency stimuli but also a reduction of this contingency effect after low-contingency trials compared to after high-contingency trials. Given the confound of contingency levels and congruency levels in studies using non-neutral distractors and biased stimulus selection laid out above, this contingency switch cost allows accounting for the CSE under such circumstances as a modulation of associative priming.
To achieve a 50%/50% ratio of congruent and incongruent trials while controlling for feature sequence and distractor contingency effects, one might use a recently established procedure, which we refer to as the split-task method. For this purpose, a four-choice task is divided into two two-choice tasks involving distinct sets of targets and distractors. With this arrangement, contingencies are unbiased when congruent and incongruent trials are administered with a frequency ratio of 50%/50%. Moreover, trial-to-trial feature repetitions can be controlled for by preventing the presentation of the same task in consecutive trials or by confining the analysis to task alternation trials. Several studies reported a CSE with such a task protocol (e.g., Kim & Cho, 2014; Schmidt & Weissman, 2014; Weissman et al., 2014; Weissman et al., 2015). Yet, some experiments using the split-task method failed to obtain a significant CSE (e.g., Weissman et al., 2014), suggesting that at least part of the CSE observed under less controlled conditions is based on confounding factors, leaving only a small effect when these factors are removed.
Despite this progress made in controlling confounds, significant problems have remained. First, we note that the application of the split-task method – as of two-choice tasks in general – might allow for a particular processing strategy that could explain the CSE without reference to attentional adjustment. Specifically, it is conceivable that participants use the distractor to prepare the opposite response than the one the distractor is associated with by instruction (i.e., the response the distractor would require if it were presented as target) after incongruent trials. We refer to this strategy as reversed distractor-response priming (see Wühr & Kunde, 2008, for a discussion of this idea in the context of anticipatory conflict control). Preventing this strategy by using four-choice tasks with random stimulus selection has yielded mixed results. Whereas two studies failed to observe a CSE when data from trials associated with repetitions were excluded from the analyses (Mordkoff, 2012; Schmidt & De Houwer, 2011), Tomat et al. (2020) observed a CSE under such conditions in a temporal flanker task, which involves the consecutive presentation of a distractor and a target (Hazeltine et al., 2011). Noteworthy, however, resembling the vast majority of studies using spatial flanker tasks to investigate the CSE (e.g., Gratton et al., 1992; Schmidt & De Houwer, 2011), their stimulus material consisted of perceptually identical objects (i.e., the letters A, B, C, and D) used as targets and flankers. This property of the procedure results in yet another ambiguity regarding the CSE interpretation, namely a confound of congruency level and perceptual target-distractor identity. Under such conditions, it is conceivable that the CSE is brought about by advantageous perceptual segregation of the target and the distractor after a trial involving the same congruency level, hence the same perceptual requirement. This argument might not be confined to the case of target-distractor conjunctions involving identical elements but generalize to cases in which the targets and distractors are perceptually more similar in congruent than in incongruent trials, a frequent property of priming protocols used to investigate the CSE (e.g., Schmidt & Weissman, 2014; Weissman et al., 2014).
Congruency level and perceptual stimulus identity of target-distractor pairs are naturally deconfounded in Stroop-like tasks, that is, in tasks, in which targets and distractors belong to perceptually different stimulus categories such as colors and words for which no systematic relationship of perceptual similarity has been established. Although Blais et al. (2014) reported a CSE in a Stroop task while controlling for stimulus feature sequences and distractor-response contingencies (using a 4:4 S-R mapping and varying the Proportion Congruency conditions), Schmidt (2014) criticized accounting for these results in terms of attentional adjustment based on concerns regarding statistical power and transfer of biased contingencies between blocks of trials. (This criticism might also apply to the study of Tomat et al., 2020.) Moreover, it has recently been suggested that the CSE in the Stroop task is brought about by a different (priming) mechanism than in other conflict protocols, such as flanker and Simon tasks (Aschenbrenner & Balota, 2017), raising doubts about reliance on Stroop task data as the sole source of CSE theorizing.
In light of these developments, further investigation of the CSE, deconfounding congruency, and target-distractor similarity in Stroop tasks or non-Stroop tasks, seems a valuable endeavor. Prime-probe procedures like the temporal flanker task may play a prominent role in this for at least two reasons. First, the availability of distractor information before target information might be a relevant precondition for obtaining a robust CSE in confound-minimized procedures (Weissman et al., 2014). Second, manipulating the length of the distractor-target stimulus-onset asynchrony (SOA) may yield results of theoretical interest. In this connection, the finding of a CSE with a prime-probe (i.e., distractor-target) interval of 1000 ms (Weissman et al., 2015) deserves attention. Intriguingly, in this condition, the congruency effect as such was no longer significant, yielding a CSE in the form of a regular congruency effect after congruent trials and a reversed congruency effect after incongruent trials. As the latter cannot be accounted for in terms of attentional focusing (which would result in the absence of a congruency effect, at most), this finding raises new doubts about the attentional adjustment interpretation of the CSE. In fact, the authors favored a (non-attentional) response modulation account, which attributes the CSE to inhibition of response(s) activated by the distractor, assuming that this inhibition is more efficient if it was involved in the preceding trial. Consequently, the response activated by the distracter is assumed to be inhibited more efficiently if the preceding trial was incongruent compared to if the preceding trial was congruent (cf. Ridderinkhof, 2002). However, the reversal of the congruency effect after incongruent trials might also reflect reversed distractor-response priming after conflict or perceptual facilitation in congruency level repetition trials, as it occurred under conditions of a two-choice task and higher perceptual target-distractor similarity in congruent than in incongruent trials. Nevertheless, the possibility of investigating the CSE in the absence of a congruency main effect offers the possibility of critically testing the attentional adjustment account. Considering the fact that the congruency effect in the temporal flanker task tends to decrease strongly even in four-choice tasks when the prime-probe interval increases (Gillich et al., 2019), this procedure could likely provide such a condition.
At this point, we would like to point out that all the CSE-related confounds hitherto mentioned are a consequence of the fact that the presumed attentional adjustment is assessed in terms of a congruency level repetition advantage. This might allow for more non-attentional explanations than currently identified. In Experiment 3 of the current study, we turn to a new method of assessing conflict-induced attentional adjustment, which arguably controls for all confounds associated with the sequence of congruency levels (cf. Tomat et al., 2020; Wendt et al., 2012; Wendt et al., 2014). As this method has its difficulties – which we discuss in the introduction of Experiment 3 – we first present two conventional CSE experiments in which we examined the CSE in the temporal flanker task under conditions of control of feature sequences, distractor-related contingencies, and perceptual target-distractor similarity. This was achieved by applying an 8:4 S-R mapping in a temporal flanker task. More precisely, the digits 1, 2, 3, and 4, as well as the letters A, B, C, and D were used as stimuli, and digit-letter pairs (i.e., 1-A, 2-B, 3-C, and 4-D) were assigned to four different response keys. Because on each trial, the target and the distractor were drawn from different stimulus categories (i.e., letters vs. digits), congruent trials never involved perceptually identical stimuli.
While preventing perceptually identical target-distractor pairs, drawing targets and distractors from separate stimulus categories also introduced a new possibility of conflict adjustment. Specifically, rather than modulation of the processing of stimulus information presented at a particular point in time, processing adjustment might also be targeted at the stimulus categories (i.e., amplification of the stimulus category from which the target was drawn and/or inhibition of the stimulus category from which the distractor was drawn). Although such adjustment should reveal itself in a CSE if the assignment of target and distractor to the stimulus categories repeats from the preceding trial (e.g., target: digit/distractor: letter ➔ target: digit/distractor: letter), in case the assignment is switched (e.g., target: letter/distractor: digit ➔ target: digit/distractor: letter), performance should be generally impaired after conflict by inhibited processing of the target or increased competition from the distractor. This hypothesized pattern of results is reminiscent of findings observed in task-switching experiments. Task-switching experiments (for an overview, see, e.g., Kiesel et al., 2010; Monsell et al. 2003; Vandierendonck et al., 2010) require participants to execute different tasks in varying sequences, allowing the classification of a given trial as a task repetition (i.e., the same task was executed on the directly preceding trial) or a task switch (i.e., a different task was executed in the directly preceding trial).Footnote 1 In addition to task-switch costs (i.e., impaired performance in task switch trials relative to task repetition trials), task switching experiments have also demonstrated competition between the currently relevant stimulus-response rules and the currently irrelevant task. Specifically, using the same set of motor responses for both tasks, performance tends to be better in trials, in which the same response is called for by both tasks (i.e., congruent trials), than in trials, in which the tasks call for different responses (i.e., incongruent trials). Consistent with the notion of conflict-induced biasing of task readiness (i.e., amplification of the relevant task’s mental set and/or inhibition of the competitor task’s mental set in incongruent trials), analyzing task performance as a function of the congruency level of the preceding trial demonstrated a CSE when the task repeated but not when the task alternated (Brown et al., 2007; Kiesel et al., 2006; Schneider, 2015; Wendt et al., 2013), as well as larger task switch costs after incongruent, compared to congruent trials (Brown et al., 2007; Goschke, 2000; Kiesel et al., 2006; Schneider, 2015; Wendt et al., 2013).
In light of these previously observed effects in task-switching studies and our considerations concerning the conflict-induced adjustment of target and distractor categories, it seems useful to analyze the CSE as a function of repetition versus alternation of the sequence of stimulus categories (i.e., digit first/letter second or letter first/digit second) or, if conceived of as distinct (digit and letter identification) tasks, of task repetition versus alternation. Whereas adjustment concerning the processing of stimuli presented at particular points in time within a trial (temporal attentional adjustment; i.e., amplification of processing the second [target] stimulus and/or inhibition of processing of the first [distractor] stimulus after conflict) should be effective for both repetition and alternation trials, thus eliciting a CSE in both cases, adjustment concerning the processing of category-specific stimuli (i.e., digits, letters) or task-sets (digit identification, letter identification) should not. More precisely, inhibition of the distractor category or amplification of the target category should only reduce the distractor's influence on performance (thus producing a CSE) if the assignment of stimulus categories to the target and the distractor is maintained in the subsequent trial. By contrast, switching this assignment would require responding to a stimulus of the previously inhibited category and/or suffering interference from a stimulus or the previously amplified category. Thus, we would expect a CSE in repetition trials and performance decrement in alternation trials after an incongruent compared to after a congruent stimulus.
In summary, despite almost three decades of research on the CSE in flanker tasks, unequivocal evidence for increased attentional focusing on target-related information, evoked by response conflict, is still lacking. In Experiments 1 and 2 of the current study, we controlled for previously identified confounds with stimulus and response feature sequences, distractor-response contingencies, and perceptual target-distractor similarity, preventing a strategy of conflict-induced reversal of distractor-response priming, by use of a temporal flanker task protocol including eight stimuli (four digits and four letters), mapped onto four responses.