Competition for priority in processing increases prefrontal cortex’s involvement in top-down control: an event-related fMRI study of the stroop task

Abstract

Prior work indicates that various aspects of task-irrelevant information (e.g. its salience, task-relatedness, emotionality) can increase the involvement of prefrontal cortex (PFC) in top-down attentional control. In light of these findings, we hypothesize that PFC’s involvement increases when task-irrelevant information competes for priority in processing. In an event-related fMRI study using an oddball variant of the Stroop task, we examine the generality of this hypothesis using three manipulations designed to increase the ability of task-irrelevant information to compete for priority in processing. First, we investigated how the frequency of occurrence of task-irrelevant information affects PFC activity. Second, we examined whether conflicting color information (i.e. incongruent trials) increases activity in regions of PFC that are similar to or distinct from those sensitive to infrequently occurring task-irrelevant information. Finally, we examined the impact of the number of levels at which conflict could occur (e.g. non-response only, non-response and response). Activity in posterior–dorsolateral and posterior–inferior PFC increased for infrequently occurring task-irrelevant information, being largest when the task-irrelevant information contained conflicting color-information. In contrast, increases in mid-dorsolateral prefrontal cortex’s activity were only noted when conflicting color information was present, being largest when conflict occurred at multiple levels. The anterior cingulate was primarily sensitive to the occurrence of conflict at the response level with only a small sub-region exhibiting sensitivity to non-response conflict as well. From these findings we suggest that posterior DLPFC and PIPFC are involved in biasing processing in posterior processing systems, mid-DLPFC is involved in biasing the processing of the contents of working memory, and ACC is primarily involved in response-related processes.

Introduction

The ability to select between task-relevant and task-irrelevant information is a basic aspect of attentional function. Neuropsychological and neuroimaging studies indicate a role for prefrontal cortex in this process, as it provides a top-down bias that favors the selection of task-relevant information (e.g. [2], [22], [29], [33], [38]). Here we argue that such a bias is especially important for exerting attentional control when task-irrelevant information can effectively compete with task-relevant information for priority in processing.

As discussed by Frith [17] and others (e.g. [15]), the perceptually-related properties of a task-irrelevant stimulus (e.g. brightness) can increase its salience relative to a task-relevant stimulus, thereby favoring its processing and allowing associated representations to effectively compete for priority in processing. Top-down control by prefrontal cortex is required to overcome such a bias [25]. Another situation in which top-down control is required because task-irrelevant information can effectively compete for priority in processing occurs when task-irrelevant information is processed relatively automatically. In such cases, task-irrelevant representations and/or associated responses are activated to a high degree [12]. A classic example is provided by the Stroop task, in which an individual must identify a word’s ink color while ignoring its identity. Since word-reading is a relatively automatic process, the word activates task-irrelevant information (i.e. semantic and phonological representations as well as their associated responses) to a high degree, despite attempts to ignore it. Because processing of this task-irrelevant information occurs to a high degree, it is extremely effective at interfering with the processing of task-relevant information. As a result, there is an increased need for top-down control by PFC to select the correct information that should be used to guide performance, as demonstrated in numerous studies of the Stroop task and its variants [2], [31], [32]. Finally, other attributes of task-irrelevant information, such as its emotional value, can sometimes increase its ability to compete. For example, emotionally laden information, especially when arousing or negative, is effective at capturing attention [51], [53], leading to an increased need for top-down control by PFC [51], [53].

Two recent studies in our laboratory support the idea that PFC’s involvement in top-down control increases when task-irrelevant information can effectively compete for priority in processing. In one study [31], we found greater PFC activity when task-irrelevant information is related to the task at hand than when it is not. More specifically, in the color-word Stroop task, we found increases in PFC’s activity when the task-irrelevant information introduced by the word was color-related (relative to neutral words), regardless of whether or not it conflicted with task-relevant color information. Thus, we observed greater PFC activity for both congruent (e.g. the word ‘red’ in red ink) and incongruent (the word ‘blue’ in red ink) color-word trials relative to neutral-word trials (e.g. the word ‘bond’ in red ink). The task-irrelevant information introduced by the word is better able to compete for priority in processing on both congruent and incongruent color-word trials than on neutral-word trials, as it is related to the task concept (color). We observed additional PFC activity for incongruent trials relative to congruent and neutral-word trials, probably due to their ability to compromise selection of the correct response.

The findings of a second study by our laboratory are also consistent with our hypothesis, as they suggest that activity within PFC increases with the number of levels at which a stimulus can compete for priority in processing [32]. Using a manual version of the color-word Stroop task, we examined increases in neural activity associated with two different types of incongruent trials (response-eligible, response-ineligible) relative to that associated with neutral trials. On response-eligible incongruent trials, the incongruent color-word names a possible response (e.g. the word ‘blue’ in yellow ink, when blue, yellow, and green are the possible ink colors). For these trials, task-irrelevant information is present and can compete for priority in processing at both the response and non-response levels. On response-ineligible trials, the incongruent color-word does not name a possible response (e.g. the word ‘purple’ in yellow ink, when blue, yellow, and green are the possible ink colors), and hence can only compete at non-response levels. Consistent with our hypothesis, DLPFC activity was greater, relative to neutral trials, for response-eligible trials, which could introduce competing information at both response and non-response levels of processing, than for response-ineligible trials, which introduce competing information at just non-response levels. Interestingly, we noted increases in activity (relative to neutral trials) within anterior cingulate cortex only for response-eligible incongruent trials, not response-ineligible trials, implicating this prefrontal structure in response-related processes, not top-down control (see [31], [37], and [36] for additional evidence of the cingulate’s activity being linked specifically to response-related processes).

Here, we further explore the possibility that PFC’s involvement in control is dependent upon the ability of task-irrelevant information to compete for priority by examining another variable, frequency of occurrence. Although fMRI studies using priming and oddball paradigms have consistently demonstrated greater neural activity in prefrontal regions for the processing of relatively novel or infrequently occurring information when it is task-relevant, such investigations have not focused on information that is task-irrelevant. When task-irrelevant information occurs infrequently it should be better able to compete for priority in processing, as it is relatively novel and hence more salient [7].

To test our hypothesis that PFC’s involvement in control increases in response to the ability of task-irrelevant information to compete for priority in processing, we determined if PFC’s activity is greater for infrequently occurring task-irrelevant information than frequently occurring task-irrelevant information. To accomplish this, we employed an oddball variant of the Stroop task that made use of two sets of neutral words—the first was baseline neutral trials, which comprised the majority of trials (86%), and the second was oddball neutral trials, which occurred with a much lower frequency. If as we suggested, PFC’s involvement in top-down control is dependent upon the ability of task-irrelevant information to compete for priority in processing, increases in activity should be noted for oddball neutral trials relative to baseline neutral trials.

The second goal of the present study was to determine if the subregions of PFC sensitive to infrequently occurring task-irrelevant information overlap with those engaged when task-irrelevant information conflicts with task-relevant information, as on incongruent trials in the Stroop task. To accomplish this, we included oddball incongruent trials as well as oddball neutral trials. Task-irrelevant information introduced by incongruent color-words has a unique ability to compete for priority in processing, as it is both related to the task at hand and able to interfere with selecting the correct representation on which a response should be based. Hence, we examined whether the additional ability of task-irrelevant information to compete with task-relevant information on incongruent trials leads to further increases in activity within the same regions of PFC as activated by neutral oddball trials, or whether activation is observed in additional and distinct subregions of the PFC.

The third goal of our study was to re-examine the impact of response eligibility on increases in neural activity within prefrontal regions. A prior study in our laboratory [32] indicated that the ACC was impervious to the occurrence of conflict at non-response levels. Not only did we find that it exhibited greater activity during response-eligible than response-ineligible trials, but we found no significant difference in activity for response-ineligible and neutral trials. Thus, as long as the word did not name a possible response, no ACC activity was noted. However, other studies have indicated that the ACC is most responsive to novel or infrequent events [8], [54]. Thus, we reasoned that if the ACC has any sensitivity to non-response related conflict, it might be detectable with an oddball design in which the trial types of interest occur with low frequency.

To review, we made use of an oddball design, in which the majority of trials were neutral trials (referred to as baseline neutrals), with oddball trials occurring once every six to eight trials. Three oddball trial types were included: (1) neutral word trials; (2) response-ineligible incongruent trials; and (3) response-eligible incongruent trials. While we expected that all three oddball types would produce increases in neural activity due to their low frequency of occurrence, we expected activations to be more extensive for interference trials, due to their ability to activate representations of conflicting color information, which would provide them with a greater ability to compete for priority in processing. Finally, we expected response-eligible trials to produce greater increases in neural activity than response-ineligible trials, as they can compete with task-relevant information at more levels of processing.