Competitor Rule Priming: Evidence for priming of task rules in task switching

Abstract

In task-switching experiments, participants switch between task rules, and each task rule describes how responses are mapped to stimulus information. Importantly, task rules do not pertain to any specific response but to all possible responses. This work examined the hypothesis that task rules, as wholes, rather than (just) specific responses are primed by their execution, such that, in the following trial, response conflicts are exacerbated when the competing responses are generated by these recently primed rules, and performance becomes relatively poor. This hypothesis was supported in two task-switching experiments and re-analyses of additional three published experiments, thus indicating Competitor Rule Priming. Importantly, the Competitor Rule-Priming effect was independent of response repetition vs. switch, suggesting that it reflects the priming of the entire task rule rather than the priming (or suppression) of specific responses. Moreover, this effect was obtained regardless of Backward Inhibition, suggesting these effects are unrelated.

Appendix: Selection criteria

The straightforward Competitor Rule-Priming analysis included many confounds, which involved the degree of interference in Trial N (Meiran & Kessler, 2008), the degree of interference in Trial N–1 (Botvinick et al., 2001), Competitor Rule Suppression, and competitor repetition (Meiran et al., 2011). In this section, we briefly explain how we controlled for these confounds. CRP− can only occur in trials with zero, one or two competing rules, but not with three competing rules. This is because if there are three competing rules, one of them must have been the relevant rule in Trial N–1 (given that there are four tasks). Furthermore, CRP+ can only occur in trials with one, two or three competing rules, but not with zero competing rules, because to obtain Competitor Rule Priming there must be a competitor rule to be primed. Given these considerations, we excluded trials with zero or three competing rules to remove this particular confound.

Moreover, we wanted to assure that CRP− and CRP+ trials are equally influenced by conflict in Trial N–1 (Botvinick et al., 2001) and conflict in Trial N (Meiran & Kessler, 2008). Specifically, according to conflict adaptation theory (Botvinick et al., 2001), control is recruited after a conflict. This increased control mode is carried into the following trial leading to more cautious responding, which slows down the otherwise automatic response. We, therefore, wanted to assure that CRP+ and CRP− trials are equally present following 0, 1, 2 and 3 competing rules. Similarly, because trials with two competitor rules are slower than trials with one competitor rule, we wanted to equate CRP+ and CRP− trials in this respect as well.

Additionally, since competitor rule repetition should be seen in reduced interference because the competitor rule in Trial N has already been suppressed in Trial N–1 (see Meiran et al., 2011, p. 301, for support), we wanted to assure that CRP− and CRP+ trials were equated also on competitor repetitions. Finally, we also controlled for Competitor Rule Suppression (Meiran et al., 2010). To control for the Competitor Rule Suppression confound, we excluded trials if their Trial N−1 to Trial N transition included CRS+ (or CRS−) only in CRP+ trials but not in CRP− trials, and vice versa.

As Table 4 shows, controlling for all the possible confounds when considering Trial N–1 to Trial N transitions, reduces immensely the amount of valid trials. The final number of trials is detailed in the manuscript.

Finally, note that the exclusions removed all stimulus repetitions. This resulted from a combination of several factors. Consider that we had 100 % switch trials. Also, only stimuli that had one or two competitors could repeat, because only trials with one or two competitors in Trial N were included in the analysis. Note that stimuli with one competitor were stimuli with three competitors in Trial N–1 if the competitor in Trial N was the relevant in Trial N–1. Nevertheless, trials with three competitors in Trial N–1 were removed (see Table 4). The remaining trials with one competitor always had the same competitor in Trial N–1, and therefore could only be CRP− trials with competitor repetition. Because CRP+ trials could not have competitor repetition if they had only one competitor in Trial N (see Table 4), these stimulus repetition trials were removed from the analysis. As for trials with two competitors in Trial N, these trials can involve stimulus repetition only if there were also two competitors in Trial N-1. Note that we included trials with two competitors in Trial N and two competitors in Trial N-1 only if they were CRS+ trials with competitor repetition (Table 4). In this scenario one of the competitors necessarily remained the competitor rule and the other necessarily became the relevant rule. Note that the correct responses for these two rules had to be equal in Trial N-1 (because they were both competitors). However, in Trial N, these two rules were necessarily mapped to different keys (because one rule remained the competitor rule and the other became the relevant rule), and therefore there could not be stimulus repetition in this scenario.

Table 4 Selection criteria Competitor Rule Priming