Prefrontal regions play a predominant role in imposing an attentional ‘set’: evidence from fMRI

Abstract

fMRI was used to determine whether prefrontal regions play a predominant role in imposing an attentional ‘set’ that drives selection of task-relevant information. While monitoring for an atypical item, individuals viewed Stroop stimuli that were either colored words or colored objects. Attentional demands were varied, being greater when the stimuli contained two distinct and incongruent sources of information about the task-relevant attribute (e.g., when attending to color, seeing the word ‘blue’ in red ink) as compared to only one source (e.g., seeing the word ‘late’ in red ink). Prefrontal but not anterior cingulate regions exhibited greater activation on incongruent than neutral trials, suggesting that prefrontal cortex has a major role in imposing an attentional ‘set’. In addition, we found that prefrontal activation is most likely to occur when that attentional set is difficult to impose.

Introduction

Traditionally, frontal regions of the brain have been considered to be responsible for ‘executive control’ (see [1], Chapter 10, for a review). Although there is no agreement or precise definition of executive control, it has been used to subsume a wide variety of functions including the ability to create a plan or goal, to monitor progress towards that goal, to oversee the organization of action, and to override stereotypical responses. It has been suggested that frontal regions play such an executive function for a wide variety of cognitive abilities including memory [42] and attention [6], [29]. In this paper, our goal is to discern the area(s) of frontal cortex associated with a specific executive aspect of attentional control, namely the ability to create an attentional set that allows the brain to hone in on sources of information that are task-relevant [17].

Our prior neuroimaging work provides evidence that frontal regions, including dorsolateral prefrontal and cingulate cortex, are important for executive aspects of attention that aid in the selection of task-relevant information [2]. In these studies we used variants of the Stroop task. This task is commonly employed in studies of attentional selection [18], [19], and has been found to be sensitive to damage in prefrontal regions [26], [15], [41]. In the standard color-word Stroop task, individuals must attend to and identify the ink color in which a word is written, while inhibiting the more automatic response of reading the word. The need for attentional selection is high in the incongruent condition in which the word’s identity (i.e., the semantic information and its associated responses) conflicts with the color in which it is written (e.g., the word ‘BLUE’ written in green). Less attentional selection is required in the neutral condition in which the word’s identity is unrelated to color (e.g., the word ‘DOOR’ written in green).

We found that patterns (topography) of activity within frontal regions produced by an increased need for attentional selection are dependent on the nature of information that is task relevant (i.e., color, spatial), suggesting a special organization of this brain region for selecting task-relevant information. Second, we demonstrated that patterns of activity within frontal regions are relatively independent of the type of information that is task-irrelevant. Hence, frontal areas appear to have a special sensitivity to task-relevant information, making them a leading candidate for the function of imposing an attentional set.

Our evidence for these assertions came from two experiments in our laboratory. In the first experiment, we employed two Stroop tasks that differed in the nature of the task-relevant dimension, but shared the same type of task-irrelevant dimension (i.e., a word’s identity that needed to be ignored). By holding the task-irrelevant dimension constant across the two tasks, we could determine whether attentional effects within a given brain region are dependent upon the nature of the task-relevant dimension. In the second experiment we took the converse approach: we varied the nature of the task-irrelevant dimension across two tasks, but kept the task-relevant dimension constant (i.e., required a decision based on color). In this case, we were able to determine which brain regions are modulated during attentional selection by the task-irrelevant information.

These experiments yielded three important results. First, we verified a role for frontal regions in attentional control, as activation in these areas was greater in both experiments when there was a greater need for attentional selection (i.e., during incongruent as compared to neutral trials). Second, we observed that depending on the task-relevant attribute, different subregions within both dorsolateral prefrontal cortex (DLPFC) and anterior cingulate cortex (ACC) became activated with an increased need for attentional selection. Activation was centered in more ventral regions when individuals had to attend to color, whereas it was centered in more dorsal regions when they had to attend to the spatial relationship between items. Third, we found that when individuals had to attend to color, similar regions of the DLPFC and ACC became activated regardless of whether the task involved ignoring the identity of a word or an object. These findings confirm that the subregion of DLPFC and ACC that becomes active during attentional selection is dependent on the nature of task-relevant information but relatively independent of the nature of task-irrelevant information.

Because similar patterns were found for both the DLPFC and the ACC we wished to investigate whether the ACC acts in concert with DLPFC to impose an attentional set, or whether each of these frontal regions is involved in different aspects of attentional selection. At present, the data available in the literature do not allow one to disambiguate between these two possibilities, as there is evidence for both viewpoints. Studies of both animals and humans provide evidence suggesting that prefrontal regions play a role in imposing an attentional set. In monkeys, the activity of at least a subset of neurons in lateral prefrontal cortex appears to prospectively code for the expectancy of an anticipated choice in a delayed paired associate task [32]. In humans, switching between two or more different tasks in which individuals must systematically change their expectancy about task demands [33] appears to rely on dorsolateral prefrontal cortex (DLPFC) and associated areas. Neuroimaging data from PET [20] reveal that the left DLPFC is more active when participants are required to alternate between shape and color discriminations as compared to making just one of these decisions.

Although suggestive, none of the above data definitively supports the idea that prefrontal regions play a principle role in creating or maintaining an attentional set. Whereas prospective processing of information about objects occurs for cells in prefrontal cortex, this phenomenon has also been observed in other brain regions, such as inferior temporal cortex [34]. With regards to task switching in humans, it is not clear if these frontal areas are responsible for setting and maintaining biases about task requirements or whether they are specifically linked to the ability to switch between different task expectancies.

There is also evidence that the cingulate may play an important role in imposing an attentional set. Early neuroimaging studies of the Stroop task were more consistent in yielding activation of the ACC as compared to the prefrontal cortex [24], [3], [13], [4]. These results have contributed to some researchers characterizing the ACC as playing a pre-eminent role in executive aspects of attention [29]. Furthermore, in numerous neuroimaging studies of executive function, co-activation of the ACC and DLPFC is observed [8], [10], [13], [20], [27], [30].

Yet there are a variety of sources of data that suggest that the cingulate’s role in attentional selection may be quite distinct from that involved in imposing an attentional set. First, there is evidence suggesting that activity of the ACC does not always accompany processing of incongruent Stroop stimuli. Taylor et al. [38] found prefrontal but not cingulate activity when taboo words or false fonts rather than neutral words were used as a baseline against which to compare activation for incongruent color words. Second, there are numerous alternative theories about the role of the ACC in attention ranging from response selection [25], to anticipation [23], to detecting and compensating for error [21], to the detection of conflict [5]. These findings raise the possibility that ACC activity in our prior studies was generated for reasons other than creating and maintaining an attentional set.

To disentangle the role that the DLPFC plays in imposing an attentional set as compared to the ACC, we considered the design of our prior studies. We realized that incongruent as compared to neutral trials engender a high degree of response competition [25] and potential for error [5], two factors previously linked to cingulate activity. Hence, we decided to change our task such that these factors (response competition, potential for error) were held as constant as possible across conditions differing with respect to the need for attentional selection (i.e., incongruent vs. neutral trials).

To minimize the effects of response competition and potential for error, we modified the instructions provided to subjects, but retained use of the same stimuli and comparison (incongruent vs. neutral) as in our previous work. Rather than identifying in which of three colors an item appeared, we had participants monitor for the appearance of an atypical item, such as a purple stimulus, when items were presented in a limited set of colors, such as red, green, and orange. Participants were told that such an atypical item might or might not occur once during the course of each run, and that they would have to indicate whether or not such an item did appear. Under such task instructions, the likelihood of an error or response conflict is much more equivalent for incongruent and neutral items. For example, the word ‘RED’ displayed in yellow is no more likely to be considered purple than will the word ‘LOT’ displayed in yellow.

Despite these task modifications, the region that is mainly responsible for imposing an attentional set should still yield greater activity on incongruent than neutral trials. On incongruent trials, both the task-relevant dimension and the task-irrelevant dimension contain information related to the attentional set (e.g., both the word ‘red’ and its ink color blue are related to the attentional set for color). Proper task performance requires a separation of these two sources of color information to determine which one is task-relevant. In contrast, on neutral trials, only the task-relevant dimension, the item’s ink color, contains information related to the attentional set (i.e., the word ‘lot’ has no relevance with regards to color), and no disambiguation between the sources of color information is required. Hence, any area that is mainly involved in imposing an attentional set should be more engaged on incongruent than on neutral trials. We predict that the DLPFC will exhibit such a pattern.

In contrast, any brain region that is related to response conflict or error detection rather than the imposition of an attentional set, should yield little or no difference in activation on incongruent versus neutral trials. One of the strengths of the current experiment is that we have obtained increased activity in both ACC and DLPFC on incongruent as compared to neutral stimuli in our prior study which used the same stimulus set and attentional contrast as well as a similar statistical approach [2]. Hence, a lack of increased activation on incongruent versus neutral trials in the current study for either the ACC or DLPFC would be quite telling. We predict that the ACC will not show a difference in activity on incongruent as compared to neutral trials, which would implicate a role in response conflict and/or error detection, rather than a role in imposing an attentional set.

A second objective of the current study was to determine whether prefrontal activation is linked to the difficulty of imposing an attentional set. Many theories of frontal lobe function emphasize that this brain region is involved when a non-routine (i.e., novel) situation is encountered and guided control over behavior is imperative. For example, the theory of Stuss and Benson [36] suggests that the frontal lobes are highly operative when behavior must be tightly regulated via conscious control, when information must be organized to reach a goal, or when the situation is atypical. Similarly, Shallice [35] assumes that the frontal lobes are important for controlling behavior via the supervisory attentional system. This system is invoked when there is no pre-existing linkage between a stimulus and behavior, when a task is difficult, when problem solving is required, and when the typical tendency to respond must be overcome or inhibited. In both of these models, the frontal lobes become more critical as tasks become less routine and controlled attention must be exerted to ensure correct performance.

Coupling this work with our prior neuroimaging results [2], we wished to investigate whether prefrontal involvement would vary with the degree to which attentional control must be exerted to process task-relevant information. In particular, we predicted that increased prefrontal activation on incongruent as compared to neutral trials would be most likely to be observed when processing of the task-relevant dimension requires more attentional control than processing of the task-irrelevant information as compared to vice versa. If processing of the task-irrelevant information occurs with little attentional control (e.g., a word is read relatively automatically), then it will be more readily available to influence performance than will the task-relevant information (e.g., the word’s ink color). On incongruent trials, the task-irrelevant dimension contains information relevant to the attentional set. In such a situation, prefrontal mechanisms will need to select and prioritize the task-relevant information over the task-irrelevant information (e.g., a set must be imposed to monitor for a word depicted in purple, not the word ‘purple’). In contrast, if the more automatically processed dimension is task-relevant, then it will not be as difficult to impose the correct attentional set.

We used color-word and color-object Stroop stimuli to explore this hypothesis. Like the color-word Stroop stimuli, color-object stimuli consist of colored items. For an incongruent color-object stimulus, an object is depicted in a color different than the one with which it is strongly associated (e.g., a blue strawberry when strawberries are strongly associated with the color red). Neutral trials are those in which the object is depicted in one of many colors with which it is associated (e.g., a blue car).

With these color-word and color-object stimuli, we were able to contrast conditions in which the processing of color is either more attentionally demanding or less attentionally demanding than that of the other dimension. When viewing color-word Stroop stimuli, processing of color is more demanding (i.e., less automatic) than that of the word [7]. In contrast, for color-object Stroop stimuli, color is a fundamental feature of visual processing [40] and hence it is less demanding to direct attention to this feature as compared to directing attention to an object, which involves higher-order visual processing areas [39].

If prefrontal activation is linked to the difficulty of imposing an attentional set, then different patterns of prefrontal activation should be observed when attending to color with the color-word stimuli as compared to the color-object stimuli. To investigate this issue, we varied, for each type of stimulus, whether the individual was instructed to attend to color or was instructed to attend to the item (i.e., the word or the object). For the color-word Stroop stimuli, we expected that increased attentional demands would result in more prefrontal activity when attending to color than to the word because with these stimuli it is more attentionally demanding to attend to the color than the word. In contrast, for the color-object Stroop stimuli, we predicted that increased attentional demands would produce more prefrontal activity when attending to the object than the color because in this task object processing is more attentionally demanding than color processing. Notice that such a contrast provides an extremely strong test of our hypothesis because for both types of stimuli we have a condition in which an individual is attending to color. If our hypothesis is wrong and the results are driven solely by the nature of the task-relevant dimension (e.g., color vs. identity), then we should obtain similar patterns for both types of stimuli when attention is directed to color. Also notice that such an approach provides another method for differentiating the role of the prefrontal and cingulate regions. If prefrontal regions play a predominant role in imposing an attentional set, then prefrontal activity should co-vary with how demanding it is to direct attention to the task-relevant dimension, but that of the cingulate should not.

In summary, the objectives of our study were twofold. First, we wished to investigate the hypothesis that prefrontal areas play a prominent role in creating and maintaining an attentional set. Second, we wished to test the hypothesis that the degree to which prefrontal areas are involved in imposing such a set depends on how attentionally demanding it is to direct attention to the task-relevant dimension.