Congruency sequence effects and previous response times: conflict adaptation or temporal learning?

Abstract

In the present study, we followed up on a recent report of two experiments in which the congruency sequence effect -the reduction of the congruency effect after incongruent relative to congruent trials in Stroop-like tasks -was observed without feature repetition or contingency learning confounds. Specifically, we further scrutinized these data to determine the plausibility of a temporal learning account as an alternative to the popular conflict adaptation account. To this end, we employed a linear mixed effects model to investigate the role of previous response time in producing the congruency sequence effect, because previous response time is thought to influence temporal learning. Interestingly, slower previous response times were associated with a reduced current-trial congruency effect, but only when the previous trial was congruent. An adapted version of the Parallel Episodic Processing (PEP) model was able to fit these data if it was additionally assumed that attention "wanders" during different parts of the experiment (e.g., due to fatigue or other factors). Consistent with this assumption, the magnitude of the congruency effect was correlated across small blocks of trials. These findings demonstrate that a temporal learning mechanism provides a plausible account of the congruency sequence effect.

The Parallel Episodic Processing (PEP) model 2.0: A single computational model of stimulus-response binding, contingency learning, power curves, and mixing costs

“…If the rhythm is broken (e.g., not enough evidence for a response when the threshold is temporarily lowered), then responding is slower. In previous work, it was demonstrated that this mechanism is important for simulating several findings in the cognitive control literature (see Schmidt, 2013c;Schmidt & Weissman, 2016) and also proves important for mixing costs. No changes were made to this mechanism in the present work.…”

“…Typically entirely new models or variations of old models are used for each simulated dataset. Related to our fixed-model philosophy, we also ensured that the current version of the model is able to simulate the findings that the older versions of the model were able to simulate, namely item-specific proportion congruent effects (Schmidt, 2013a), list-level proportion congruent effects (Schmidt, 2013c), congruency sequence (or Gratton) effects (Schmidt & Weissman, 2016), and asymmetric list shifting effects (Schmidt, 2016a). For brevity, these results are presented in Appendix B, but further details can be obtained in the original papers or from the lead author on request.…”

“…First, representations of individual events (e.g., trials in an experiment) are recorded into individual memories in episodic memory. These episodes might contain information about the stimuli experienced and the response that was executed (Hommel, 1998), in addition to timing information such as response times (Kinoshita, Forster, & Mozer, 2008;Kinoshita, Mozer, & Forster, 2011;Mozer, Kinoshita, & Davis, 2004;Schmidt, 2013cSchmidt, , 2014Schmidt, Lemercier, & De Houwer, 2014;Schmidt & Weissman, 2016), and perhaps even more complex information such as task sets (Waszak, Hommel, & Allport, 2003) or cognitive control settings (Egner, 2014). Second, on subsequent events episodic memories are retrieved to aid in performance.…”

“…In addition to practice curves, episodic memory models have been proposed for a broad range of other performance phenomena, including contingency learning (Schmidt, Crump, Cheesman, & Besner, 2007;Schmidt & De Houwer, 2012a, 2012c, 2012d, 2016aSchmidt, De Houwer, & Besner, 2010), feature binding effects (Frings, Rothermund, & Wentura, 2007;Hommel, 1998), negative priming (Rothermund, Wentura, & De Houwer, 2005), proportion congruent effects (Schmidt, 2013a(Schmidt, , 2013b, congruency sequence effects (Hommel et al, 2004;Mayr, Awh, & Laurey, 2003;Mordkoff, 2012;Schmidt & De Houwer, 2011), evaluative conditioning (Schmidt & De Houwer, 2012b), task switch costs (Logan & Bundesen, 2003Logan & Schneider, 2006a, 2006bLogan, Schneider, & Bundesen, 2007;Schmidt & Liefooghe, 2016;, rhythmic responding (Kinoshita et al, 2008;Kinoshita et al, 2011;Mozer et al, 2004;Schmidt, 2013cSchmidt, , 2014Schmidt, Lemercier, et al, 2014;Schmidt & Weissman, 2016), and various other phenomena.…”

Time-out for conflict monitoring theory: Preventing rhythmic biases eliminates the list-level proportion congruent effect.

“…One potential interpretation of this effect is conflict adaptation: following a conflicting incongruent trial, attention to the word is diminished (e.g., Botvinick et al, 2001). Another potential interpretation is temporal learning: congruency effects are larger following a fast congruent LIST-LEVEL PROPORTION CONGRUENT EFFECT 25 trial than following a slow incongruent trial (see Schmidt & Weissman, 2016). The current paradigm might thus be adjusted to pit these two accounts against one another.…”