To adapt or not to adapt: The question of domain-general cognitive control
Introduction
Consider these scenarios: (1) a British tourist crossing Manhattan’s Fifth Avenue; and (2) a reader encountering the following sentence in a novel, “At the bank, John got out his fishing pole and cast his line.” Ostensibly, these circumstances are drastically different. One entails a physical act; the other involves word and sentence comprehension. Despite obvious dissimilarities, however, both scenarios share the need for the individuals to overcome routinized responses and adjust their thoughts and actions accordingly. In the first case, the tourist must suppress her well-learned habit of looking rightward before crossing the street, in favor of looking first to the left. In the second case, the reader must abandon the initial, more frequent interpretation of “bank” (financial institution) and recover the less common but context-appropriate meaning (river’s edge) instead. In both situations, individuals must dynamically adapt their information-processing strategies to countermand the “knee-jerk” response and select a competing alternative. This ability is known as “cognitive control,” a term that describes a constellation of mental functions that guide goal-directed behavior consistent with situation-specific requirements; as such, it is particularly important for performing non-routine tasks. Although researchers generally agree that cognitive control is fundamental to complex cognition, its underlying mechanisms remain controversial. In the present research, we test whether cognitive control mechanisms are domain-general in the context of a behavioral phenomenon known as “conflict adaptation.”
According to one prominent theory of cognitive control (the Conflict Monitoring Theory, Botvinick et al., 2001), the detection of conflict—regardless of its source—automatically initiates cognitive control mechanisms (see also Desimone and Duncan, 1995, Miller and Cohen, 2001, Norman and Shallice, 1986; but see Schlaghecken & Martini, 2011). Conflict arises when an individual receives mismatched or ambiguous information about how best to characterize some input (i.e., representational conflict) or how best to respond to that input (i.e., response conflict). Consider the classic Stroop task (Stroop, 1935), where subjects must name the ink color in which color words are printed. Conflict occurs when the ink color and color name are incongruent (e.g., the word RED printed in blue ink). The conflict generated by the stimulus automatically triggers cognitive control, which mediates conflict resolution by overriding dominant but goal-irrelevant information (word meaning) and/or promoting goal-relevant processing (color naming, Botvinick et al., 2004, Egner and Hirsch, 2005, van Veen and Carter, 2006). Furthermore, sustained cognitive control engagement can yield enhanced resolution of subsequent conflict, reflected by better performance on an incongruent trial when preceded by another incongruent trial, versus when it is preceded by a congruent trial. This behavioral savings, called “conflict adaptation” or the “Gratton effect” (Gratton et al., 1992, Mayr et al., 2003, Ullsperger et al., 2005), has been observed within various tasks that elicit representational and/or response conflict (e.g., within Stroop, Flanker, Simon tasks; Botvinick et al., 1999, Hommel et al., 2004, Kerns et al., 2005, Runger et al., 2010).
Although data from within-task conflict adaptation has been instrumental to our understanding of cognitive control, the extent of the behavioral effects of sustained cognitive control—via generalized performance to another task—remains unclear. In other words, is cognitive control “one size fits all,” operating across seemingly disparate situations, or is it idiosyncratic to tasks or domains? Some researchers (Miller and Cohen, 2001, Rajah et al., 2008) argue that cognitive control mechanisms mediating conflict detection and its resolution are domain-general, meaning they operate systematically (and similarly) across conflict types that arise in different domains (e.g., verbal, non-verbal, syntactic, semantic). Others argue for domain-specificity, wherein multiple independent conflict-control systems support conflict processing only within a particular domain (Akcay and Hazeltine, 2011, Egner et al., 2007).1 Since evidence from within-task adaptation does not address the issue of domain-generality, some researchers have adopted cross-task adaptation designs to test how broadly cognitive control operates on a trial-by-trial basis (e.g., across Stroop and Flanker tasks).2 If conflict adaptation reflects sustained engagement of a domain-general system, then conflict adaptation should be observable across tasks and domains, provided that conflict arising in one domain sufficiently activates cognitive control. Conversely, if adaptation is mediated by domain-specific cognitive control, then cross-task (or cross-domain) behavioral adjustments should not occur.
Evidence for cross-task adaptation has been mixed, however: whereas some studies demonstrate adaptation from one task to another (Freitas et al., 2007, Notebaert and Verguts, 2008), others do not (Akcay and Hazeltine, 2011, Funes et al., 2010). These mixed results question the notion that conflict adaptation reflects sustained engagement of domain-general cognitive control. However, such lack of adaptation may be explained by an experimental-design artifact rather than domain-specificity. Most studies failing to find cross-task adaptation have used a paradigm in which subjects encounter a single stimulus that merges two different conflict sources (Akcay and Hazeltine, 2011, Egner et al., 2007, Funes et al., 2010). One example is the lateralized presentation of a Stroop stimulus: whereas stimulus features might induce conflict between color and word representations, stimulus location might simultaneously elicit Simon-like conflict (stimulus location conflicts with response-button location). Because conflict could occur in either task alone or in both tasks concurrently, some incongruent trials may produce less conflict than others. Consider a case where stimulus location conflicts with response-button location (incongruent Simon) but stimulus color and word meaning converge (congruent Stroop). It is possible that the Simon-conflict is opposed by the Stroop-congruency, thus eliciting a relatively weaker conflict signal that may be insufficient to engage sustained cognitive control across trials (Stroop-conflict may also be opposed by Simon-congruency in a similar manner). In other words, some conflict trials in this design may generate weaker conflict signals than those in within-task designs, thereby masking cross-task adaptation effects.
Consequently, we revisit the question of domain-general cognitive control using a cross-task adaptation design that circumvents the aforementioned methodological issue (but see Egner, 2008). Moreover, our study addresses limitations in prior studies that have observed cross-task adaptation. For instance, most studies have tested adaptation effects across Stroop, Flanker, and Simon tasks, which share the need to overcome stimulus–response conflict. Given this similarity, it remains unclear how far conflict adaptation extends. Here we ask, can behavioral adjustments be observed across syntactic and lexical domains and across perceptual and verbal domains, where task demands and stimulus characteristics are wildly different?
We hypothesize that two tasks sharing conflict-control demands, regardless of apparent dissimilarities, should yield conflict adaptation—the engagement of online adjustments that generalize from one task to another. Demonstrating that conflict detection in one task facilitates conflict resolution in a different task would provide strong evidence for domain-general cognitive control, since cross-task adaptation should occur only as a result of the activation of cognitive control resources that operate across both task-domains. We selected well-established tasks known to elicit conflict: in Experiment 1, reading syntactically ambiguous sentences that generate temporary misinterpretation (Garnsey, Pearlmutter, Myers, & Lotocky, 1997), and in Experiment 2, passively viewing a bistable figure (Necker cube) that induces perceptual ambiguity (Kornmeier and Bach, 2005, Long and Toppino, 2004). We measured adaptation effects on the Stroop task because its behavioral profile is well understood, and it is widely used in studies of cognitive control and conflict adaptation (see Section 2.4 for additional information). Across these experiments, we tested if conflicts arising in syntactic and perceptual domains influence performance, via conflict adaptation, in a non-syntactic, verbal domain (i.e., Stroop).
Section snippets
Overview
All subjects performed two tasks: a color–word Stroop task and either a sentence-processing (Experiment 1) or perceptual-processing task (Experiments 2 and 3). Furthermore, all tasks included both no-conflict (congruent or unambiguous) and high-conflict (incongruent or ambiguous) trials. To match the terminology used for Stroop trials, we describe both unambiguous sentences and unambiguous Necker stimuli as “congruent” and both ambiguous sentences and ambiguous Necker stimuli as “incongruent.”
Experiment 1
The design of Experiment 1 was motivated by data in the neurocognitive literature demonstrating the following findings: (1) co-localized BOLD activity in regions within left ventrolateral prefrontal cortex (VLPFC) during healthy adults’ performance of both a color–word Stroop task and a language comprehension task involving syntactic conflict (January et al., 2009, Ye and Zhou, 2009), and (2) concomitant deficits in a left VLPFC patient, who showed inflated susceptibility to conflict during
Experiment 2
If conflict adaptation is mediated by domain-general cognitive control, then ambiguity experienced in a perceptual domain should trigger cognitive control and initiate behavioral adjustments in the verbal domain. Thus, we tested conflict adaptation from a non-verbal task that induces perceptual ambiguity—passively viewing a bistable Necker Cube (Necker, 1832)—to the Stroop task. We chose the Necker stimulus for three reasons. First, as shown in Fig. 3, the set of overlapping squares can induce
Materials and procedure
Experiment 3 was procedurally identical to Experiment 2, except that incongruent Necker trials in Experiment 2 were replaced with 90 s of alternating unambiguous Necker cubes (with right-downward Necker and left-upward Necker randomly intermixed). Subjects pressed a button whenever a stimulus change occurred. Importantly, these alternating Necker trials were constructed to mimic the distribution of self-reported reversals in Experiment 2. That is, the duration of each unambiguous stimulus varied
General discussion
Across these studies, we provide strong evidence for domain-general cognitive control, revealed by cross-task behavioral adjustments. Importantly, conflict adaptation findings from a syntactic to non-syntactic domain, and from a perceptual to verbal domain, are novel. Furthermore, in Experiment 2, the extent to which individuals experienced perceptual ambiguity predicted the magnitude of their adaptation on the subsequent Stroop task, suggesting that the extent of cognitive control engagement
Acknowledgments
We thank Anna Beeman for assistance with data collection. This work was supported in part by NSF-IGERT training grant DGE-0801465.
References (59)
- et al.
Conflict monitoring and anterior cingulate cortex: An update
Trends in Cognitive Sciences
(2004) - et al.
Thematic roles assigned along the garden path linger
Cognitive Psychology
(2001) Multiple conflict-driven control mechanisms in the human brain
Trends in Cognitive Sciences
(2008)- et al.
Separate conflict-specific cognitive control mechanisms in the human brain
Neuroimage
(2007) - et al.
The independence of syntactic processing
Journal of Memory and Language
(1986) - et al.
The contributions of verb bias and plausibility to the comprehension of temporarily ambiguous sentences
Journal of Memory and Language
(1997) - et al.
The Necker cube—An ambiguous figure disambiguated in early visual processing
Vision Research
(2005) - et al.
Interdimensional interference in the Stroop effect: Uncovering the cognitive and neural anatomy of attention
Trends in Cognitive Sciences
(2000) - et al.
The relative involvement of anterior cingulate and prefrontal cortex in attentional control depends on nature of conflict
Cognitive Brain Research
(2001) Task switching
Trends in Cognitive Sciences
(2003)