Separating semantic conflict and response conflict in the Stroop task: A functional MRI study
Introduction
One of the central purposes of our attentional systems is to overcome interference caused by distracting information. The Stroop task (Stroop, 1935) is among the most frequently used tasks in cognitive psychology, clinical neuropsychology, and cognitive neuroscience to study interference and attention (e.g., Kornblum et al., 1999, MacLeod, 1991, MacLeod and MacDonald, 2000, Pardo et al., 1990, Stuss et al., 2001). In this task, participants have to name the ink color of a word that spells a color name. When the color and the word are congruent (e.g., the word “blue” in blue letters), the task is easy; when the color and the word are incongruent (e.g., the word “red” in blue letters), people experience interference. This is thought to occur because word reading is a more practiced and more automatic skill than is the naming of colors, so attentional control is required to overcome the tendency to respond to the word instead of to the color (Cohen et al., 1990, MacLeod, 1991).
Most neuroimaging studies that have investigated the neural basis of these attentional systems by studying the Stroop or Stroop-like tasks have identified the dorsolateral prefrontal cortex (PFC), anterior cingulate cortex (ACC), and posterior parietal cortex (PPC) as being central to overcoming interference (Banich et al., 2000a, Banich et al., 2000b, Barch et al., 2001, Bush et al., 1998, Carter et al., 1995, Carter et al., 2000, Fan et al., 2003, MacDonald et al., 2000, Milham et al., 2001, Milham et al., 2003, Pardo et al., 1990). The dorsolateral PFC is most often thought to select the relevant information by imposing an attentional set, or biasing information in posterior cortices by representing context (Banich et al., 2000a, Miller and Cohen, 2001). The ACC is often thought to detect the presence of conflict and alert other systems to exert control (Botvinick et al., 2001, Van Veen and Carter, 2002a), while the PPC is often thought to represent task-relevant stimulus-response mappings or stimulus-response transformations or to be involved in the visuospatial selection of relevant stimuli (Bunge et al., 2002, Casey et al., 2000, Rushworth et al., 2001).
For decades, it has been debated at which level of processing the conflict that causes Stroop RT interference occurs. While it has been suggested that conflict occurs at the response level (e.g., Cohen et al., 1990, Duncan-Johnson and Kopell, 1981, MacLeod, 1991), it has also been argued that conflict additionally occurs between representations at the level of semantic or conceptual encoding (e.g., Brown and Besner, 2001, Luo, 1999, Seymour, 1977, Treisman and Fearnley, 1969, Zhang and Kornblum, 1998) or conceptual (lemma) selection (Roelofs, 2003). Evidence for the contribution of semantic conflict to Stroop interference comes from various sources; examples include studies that have shown greater RT interference when the word is semantically closer to the color (e.g., Brown and Besner, 2001, Fox et al., 1971, Klein, 1964, Klopfer, 1996) and studies that have shown an interference effect when color and word are incongruent even when the task is simply to note whether color and words are the same or not (e.g., Luo, 1999, Treisman and Fearnley, 1969, Zysset et al., 2001). In contrast, supportive evidence for a contribution of response conflict to Stroop interference comes, among others, from the finding that the latency of the P300 (a component of the event-related potential sometimes assumed to index stimulus-related processing) does not differ between congruent and incongruent stimuli (Duncan-Johnson and Kopell, 1981), and from the finding that interference is larger when the incongruent color word is part of the response set than when it is not (e.g., Milham et al., 2001, West et al., 2004).
Several studies have shown evidence for both kinds of conflict contributing to the Stroop interference effect within the same experiment. First, there have been studies that have manipulated response eligibility by using incongruent color words that either were (“eligible”) or were not (“ineligible”) part of the response set, and compared these to the effects of neutral, non-color words (e.g., Milham et al., 2001, Milham et al., 2003, West et al., 2004). These studies have found longer RTs for incongruent-ineligible trials compared to neutral trials, presumably reflecting semantic conflict, and still longer RTs for incongruent-eligible trials, presumably reflecting response conflict. Milham et al. (2001) found response conflict activity in ACC and right PFC and non-response conflict activity in left PFC and PPC. In a subsequent study, Milham et al. (2003) used a similar design but presented incongruent-eligible and incongruent-ineligible trials in only a small proportion of trials, a manipulation that usually increases interference (Logan, 1985, Tzelgov et al., 1992) and ACC activity (Carter et al., 2000). In addition, Milham et al. used oddball neutral trials, in which the irrelevant neutral word occurred as infrequently as did the color words of the incongruent trials. In this study, semantic conflict engaged bilateral superior parietal cortex, bilateral inferior and dorsolateral PFC, a small area of the ACC, and other regions; response conflict engaged bilateral inferior and dorsolateral PFC, ACC, and inferior parietal cortex. Note that in this study, semantic conflict only activated a small area of the ACC, whereas response conflict engaged a much larger area of this structure, encompassing the semantic conflict area. A similar study using ERPs yielded slightly different results. West et al. (2004) used a numeric version of the Stroop task with neutral, incongruent-eligible and incongruent-ineligible trials, and found that the N450 of the event-related potential was enhanced to both types of conflict but did not differ between incongruent-eligible and incongruent-ineligible trials. They modeled this component as having ACC and right PFC generators, suggesting that these regions responded to semantic conflict but did not distinguish between response eligibility.
A different behavioral study that showed evidence that both forms of conflict contribute to the Stroop interference effect was conducted by De Houwer (2003). Based on the dimensional overlap model of Kornblum and colleagues (e.g., Kornblum et al., 1990, Kornblum et al., 1999, Zhang et al., 1999), De Houwer used a two-choice button-press version of the paradigm, with two colors associated with each response hand. This manipulation allowed for stimuli to be either congruent (CO; word and color are the same), incongruent at only the semantic level (SI; word and color are different, but mapped onto the same response hand), or incongruent at both semantic and response levels (RI; word and color are different and mapped onto opposite response hands). De Houwer (2003) showed that, with this version of the Stroop task, RTs are longest to the RI condition, somewhat faster to the SI condition, and fastest to the CO condition. Note that this design assumes “subtractive” logic; the RI condition is assumed to contain both semantic and response conflict, the SI condition only semantic conflict, and the CO condition neither. Therefore, the RI–SI comparison should reflect the effects of response conflict, and the SI–CO comparison should reflect the effects of semantic conflict.
In determining how the brain processes and deals with these two different kinds of conflict, a study based on the experimental design used by De Houwer (2003) might have an advantage over studies manipulating response eligibility. In each trial type of De Houwer's design, the distracting word is part of the response set, thus controlling for response eligibility. In the studies manipulating response eligibility, the distracting word stimuli were not always part of the target set, and it is possible that this might have biased the results. There is some evidence suggesting that stimuli that are not part of the target set are processed differently in the brain. For instance, event-related potential studies of the Stroop task have shown that the brain distinguishes neutral from congruent and incongruent stimuli before it distinguishes congruent from incongruent stimuli (West and Alain, 1999). This might suggest that the brain attenuates the processing of such task-irrelevant stimuli quite early in the processing stream. Another study has found that congruent words, compared to neutral words, elicit increased activation in bilateral PPC and PFC (Milham et al., 2002). It is possible that color word stimuli elicit competition between entire task sets (i.e., color naming versus words reading), and neutral words that that are not part of the target set might do less so (cf. Monsell et al., 2001, Posner and DiGirolamo, 1998). In the design of the present study, the distracter is always part of the target set for each trial type, thereby eliminating such possible confounds.
Therefore, in the present study, we used this version of the Stroop task to investigate what brain networks underlie the detection and resolution of these two different forms of conflict in the Stroop task, and to what degree these networks might overlap. Areas selective to the detection and resolution of semantic conflict should display greater activity for SI and RI than for CO, but not differ between SI and RI. Areas selective to response conflict should display similar activity to the CO and SI conditions, and be increased only to the RI condition. Finally, areas involved in general difficulty or the detection and resolution of general conflict (i.e., conflict regardless of its nature or source) should display a stepwise increase in activity; activity should be greater to the SI than to the CO condition, and greater still to the RI condition.
If we were to find different areas of the PFC to be engaged by semantic and response conflict, this finding would provide partial support for an “organization-by-material” view of the PFC, according to which different subregions of this area are involved with controlling different representational domains (e.g., Smith and Jonides, 1999). Although research on this area has focused mostly on different stimulus or working memory domains, such as verbal versus spatial, it has recently been suggested that more dorsal areas of the PFC are more sensitive to the stimulus demands of a task, whereas more ventral areas are more sensitive to the response demands (Casey et al., 2001). Thus, this view would predict that the SI–CO contrast would result in more dorsal activation in the PFC, whereas the RI–SI contrast would result in more ventral activation. In contrast, it is also possible that the SI–CO contrast would result in activation of the left ventrolateral PFC (BA 47/45), as this area has repeatedly been implicated in selecting between competing alternatives from semantic memory or the control of semantic retrieval (e.g., Badre and Wagner, 2002, Kan and Thompson-Schill, 2004, Thompson-Schill et al., 1997, Wagner et al., 2001). And, if the findings by Milham et al. (2001) were to replicate, we would find left-lateralized activation with semantic conflict and right-lateralized activation with response conflict.
In addition, since the PPC has often been implied in stimulus-response mapping (e.g., Bunge et al., 2002), we expect it to be engaged by the SI–CO contrast. The ACC has often been associated with response conflict (e.g., Milham et al., 2001, Van Veen et al., 2001) so we expect it to be engaged by the RI–SI contrast; however, it has sometimes been associated with conflict between semantic or stimulus representations (e.g., Milham et al., 2003, Weissman et al., 2003, Zysset et al., 2001) and it is therefore possible that it will be engaged by the SI–CO contrast.
Section snippets
Research participants and task
Informed consent was obtained from 14 right-handed participants (6 women, 8 men) with an age range of 19–28 years old (M = 21.4, SD = 2.2). They were instructed to make a left index finger button press if the color of the presented word was red or yellow, or a right index finger button press when the color of the word was blue or green. For congruent stimuli (CO), color and word were the same (e.g., “red” printed in red); for semantically incongruent stimuli (SI), color and word were different
Performance data
Average performance data are summarized in Fig. 1. Mean RTs for the CO, SI, and RI conditions were 598 ms (SD = 146), 643 ms (SD = 164), and 719 ms (SD = 203), respectively, which were significantly different (F(2,13) = 38.98, P < 0.001). Planned contrasts indicated that RTs to SI trials were slower than to CO trials (F(1,13) = 25.64, P < 0.001), and RTs to RI trials were again slower than to SI trials (F(1,13) = 39.82, P < 0.001). Erroneous responses tended to be fast (M = 570 ms, SD = 188);
Discussion
The performance data replicated those by De Houwer (2003): RT increases with SI compared to CO, reflecting a contribution of semantic conflict (without response conflict), and increases again with RI compared to SI, indicating response conflict. This provides strong evidence for a contribution of both kinds of conflict to the overall Stroop effect (cf. Brown and Besner, 2001, Luo, 1999, Milham et al., 2001, Roelofs, 2003, Seymour, 1977, Zhang and Kornblum, 1998). In addition, accuracy was
Acknowledgments
This work was supported by MH64190 from the NIMH and a Translational Clinical Scientist Award from by the Burroughs-Wellcome Foundation to Dr Carter. We thank Mylaina L. Gordon for assistance with data collection, and Stefan Ursu and Raymond Y. Cho for helpful discussions.
References (74)
- et al.
Selection, integration, and conflict monitoring: assessing the nature and generality of prefrontal cognitive control mechanisms
Neuron
(2004) Prefrontal regions play a predominant role in imposing an attentional ‘set’: evidence from fMRI
Cogn. Brain Res.
(2000)Investigations of the functional anatomy of attention using the Stroop test
Neuropsychologia
(1993)Dissociable contributions of prefrontal and parietal cortices to response selection
NeuroImage
(2002)Interference and facilitation effects during selective attention: an H215O PET study of Stroop task performance
NeuroImage
(1995)Cognitive and brain consequences of conflict
NeuroImage
(2003)- et al.
Executive control emerging from dynamic interactions between brain systems mediating language, working memory and attentional processes
Acta Psychol.
(2004) The thalamic matrix and thalamocortical synchrony
Trends Neurosci.
(2001)Reading anomalous sentences: an event-related fMRI study of semantic processing
NeuroImage
(2002)Executive control of thought and action
Acta Psychol.
(1985)
Interdimensional interference in the Stroop effect: uncovering the cognitive and neural anatomy of attention
Trends Cogn. Sci.
The relative involvement of anterior cingulate and prefrontal cortex in attentional control depends on nature of conflict
Cogn. Brain Res.
Attentional control in the aging brain: insights from an fMRI study of the Stroop task
Brain Cognition
Competition for priority in processing increases prefrontal cortex's involvement in top-down control: an event-related fMRI study of the Stroop task
Cogn. Brain Res.
An investigation of the value of spin-echo-based fMRI using a Stroop color-word matching task and EPI at 3T
NeuroImage
Stroop performance in focal lesion patients: dissociation of processes and frontal lobe lesion location
Neuropsychologia
The anterior cingulate as a conflict monitor: fMRI and ERP studies
Physiol. Behav.
Anterior cingulate cortex, conflict monitoring, and levels of processing
NeuroImage
Errors without conflict: implications for performance monitoring theories of anterior cingulate cortex
Brain Cogn.
Recovering meaning: left prefrontal cortex guides controlled semantic retrieval
Neuron
Conflict monitoring in the human anterior cingulate cortex during selective attention to global and local object features
NeuroImage
Event-related neural activity associated with the Stroop task
Cogn. Brain Res.
A parallel distributed processing model of stimulus–stimulus and stimulus–response compatibility
Cogn. Psychol.
Color-word matching Stroop task: separating interference and response conflict
NeuroImage
Semantic retrieval, mnemonic control, and prefrontal cortex
Behav. Cogn. Neurosci. Rev.
fMRI studies of Stroop tasks reveal unique roles of anterior and posterior brain systems in attentional selection
J. Cogn. Neurosci.
Anterior cortex and response conflict: effects of response modality and processing domain
Cereb. Cortex
Conflict monitoring versus selection-for-action in anterior cingulate cortex
Nature
Conflict monitoring and cognitive control
Psychol. Rev.
On a variant of Stroop's paradigm: which cognitions press your buttons?
Mem. Cogn.
The counting Stroop: an interference task specialized for functional neuroimaging: validation study with functional MRI
Hum. Brain Mapp.
Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex
Proc. Natl. Acad. Sci. U. S. A.
Dissociation of response conflict, attentional selection, and expectancy with functional magnetic resonance imaging
Proc. Natl. Acad. Sci. U. S. A.
Sensitivity of prefrontal cortex to changes in target probability: a functional MRI study
Hum. Brain Mapp.
On the control of automatic processes: a parallel distributed processing account of the Stroop effect
Psychol. Rev.
A psychophysiological investigation of the continuous flow model of human information processing
J. Exp. Psychol. Hum. Percept. Perform.
On the role of stimulus–stimulus and stimulus–response compatibility in the Stroop effect
Mem. Cogn.
Cited by (275)
Enhancement of visual dominance effects at the response level in children with attention-deficit/hyperactivity disorder
2024, Journal of Experimental Child PsychologyDisentangling three different conflicts at non-response/semantic and response levels in a new stroop-matching task
2024, Learning and MotivationShort-term PM<inf>2.5</inf> exposure and cognitive function: Association and neurophysiological mechanisms
2022, Environment International