《认知神经科学》课程教学资源（参考文献）[Botvinick, M. M., Braver, T. S., Barch, D. M., Carter, C. S., & Cohen, J. D.（2001）]Conflict monitoring and cognitive control

团购合买资源类别：文库，文档格式：PDF，文档页数：29，文件大小：3.57MB

Psychological Review 2001, Vol. 108, No. 3, 624-652 Copyright 2001 by the American Psychological Association, Inc. 0033-295X701/S5.00 DOI: 10.1037//0033-295X.I08.3.624 Conflict Monitoring and Cognitive Control Matthew M. Botvinick Carnegie Mellon University, University of Pittsburgh, and Center for the Neural Basis of Cognition Todd S. Braver and Deanna M. Barch Washington University Cameron S. Carter University of Pittsburgh and Center for the Neural Basis of Cognition Jonathan D. Cohen Princeton University and University of Pittsburgh A neglected question regarding cognitive control is how control processes might detect situations calling for their involvement. The authors propose here that the demand for control may be evaluated in part by monitoring for conflicts in information processing. This hypothesis is supported by data concerning the anterior cingulate cortex, a brain area involved in cognitive control, which also appears to respond to the occurrence of conflict. The present article reports two computational modeling studies, serving to articulate the conflict monitoring hypothesis and examine its implications. The first study tests the sufficiency of the hypothesis to account for brain activation data, applying a measure of conflict to existing models of tasks shown to engage the anterior cingulate. The second study implements a feedback loop connecting conflict monitoring to cognitive control, using this to simulate a number of important behavioral phenomena. A remarkable feature of the human cognitive system is its ability to configure itself for the performance of specific tasks through appropriate adjustments in perceptual selection, response biasing, and the on-line maintenance of contextual information. The processes behind such adaptability, referred to collectively as cognitive control, have been the focus of a growing research program within cognitive psychology. A number of theoretical models have been proposed for how the control of cognition is achieved (Baddeley & Delia Sala, 1996; Cohen, Dunbar, & McClelland, 1990; Norman & Shallice, 1986), and progress has been made toward identifying its neuroanatomical substrates (Cohen, Braver, & O'Reilly, 1996; Cohen & Servan-Schreiber, 1992; Desimone & Duncan, 1995; GoldmanRakic, 1996; Luria, 1973; Posner & Petersen, 1990). Despite the importance of these efforts to characterize the function of cognitive control, most of them share an important limitation in scope. Most current theories focus nearly exclusively on the Matthew M. Botvinick, Department of Psychology, Carnegie Mellon University, Department of Psychiatry, University of Pittsburgh, and Center for the Neural Basis of Cognition, Pittsburgh, Pennsylvania; Todd S. Braver and Deanna M. Barch, Department of Psychology, Washington University; Cameron S. Carter, Departments of Psychiatry and Psychology, University of Pittsburgh, and the Center for the Neural Basis of Cognition, Pittsburgh, Pennsylvania; Jonathan D. Cohen, Department of Psychology and Center for the Study of Mind, Brain, and Behavior, Princeton University, and Department of Psychiatry, University of Pittsburgh. The present work was supported by National Institute of Mental Health Grants MH16804 and MH01306, a grant from the Fetzer Foundation, and a National Alliance for Research on Schizophrenia and Depression Independent Investigator Award. Correspondence concerning this article should be addressed to Matthew M. Botvinick, Center for the Neural Basis of Cognition, 115 Mellon Institute, 4400 Fifth Avenue, Pittsburgh, Pennsylvania 15213. Electronic mail may be sent to mmb@cnbc.cmu.edu. nature of the influence exerted by control. In contrast, very little is yet known about how the intervention of control processes is itself brought about. Existing theories portray the relevant mechanisms as coming into play when their participation is required, but without an account of how the need for intervention is detected or how the intervention itself is triggered. The lack of such an account is problematic, for without it control remains a sort of homunculus that "just knows" when to intercede. For any theory of cognitive control to be complete, it will need to offer an account of how the system determines when control is required. Specifically, it will need to provide answers to the following three questions: 1. On what basis is control recruited? It cannot be the case that one always knows before beginning to act whether a given task will require high levels of control. Kahneman (1973) has argued, to the contrary, that it is often the actual attempt to perform a difficult task that leads to the recruitment of cognitive resources. This appears consistent, for instance, with the finding that participants performing the Stroop task show greater interference on the initial one or two trials in each block than on subsequent trials in the series (Henik, Bibi, Yanai, & Tzelgov, 1997). 2. Once the relevant control processes are engaged in guiding task performance, how is their influence modulated or optimized? There is evidence that adjustments in control do occur on-line, in response to variations in performance. For instance, it is wellestablished that, in speeded response tasks, reaction time and accuracy tend to rise following errors (e.g., Laming, 1968; Rabbitt, 1966). Even in the absence of errors, control appears to adjust to task demands. To take another example from the Stroop literature, participants tend to show less interference on incongruent trials if these are frequent relative to congruent trials than if they are rare (Lindsay & Jacoby, 1994; Logan & Zbrodoff, 1979). What is it that triggers these adjustments? 3. What processes govern how and when control is withdrawn? 624

CONFLICT MONITORING 625 With practice on some initially difficult tasks performance becomes increasingly automatic (e.g., Anderson, 1982; Shiffrin & Schneider, 1977). As this happens, the need for control diminishes. How do control processes evaluate the rate at which their topdown influence can be withdrawn without causing a deterioration in performance? Clearly, in order for the recruitment, modulation, and disengagement of control to occur, control processes need access to information about the functioning of the systems that they modulate. That is, in addition to the regulative dimension of control, by which its top-down influence is exerted, there must also exist an evaluative component that monitors information processing, making an assessment of current demands. If one is to expunge the homunculus from theories of cognitive control, it will be necessary to develop an account of this evaluative function. The Conflict Monitoring Hypothesis In this article we take an initial step toward characterizing the evaluative side of cognitive control, proposing one mechanism by which the demand for control might be gauged. Specifically, we will argue that there exists a system that monitors for the occurrence of conflicts in information processing, a function we refer to as conflict monitoring. By the overall account we will put forth, conflict monitoring serves to translate the occurrence of conflict into compensatory adjustments in control: The conflict monitoring system first evaluates current levels of conflict, then passes this information on to centers responsible for control, triggering them to adjust the strength of their influence on processing. A first goal of the present work is to draw together evidence for the occurrence of conflict monitoring. In particular, we will consider data suggesting that the detection of conflict may be among the functions of a particular area of the human frontal lobe, the anterior cingulate cortex (ACC). Our second objective is to articulate the hypothesis that conflict monitoring serves as a basis for the regulation of control, showing how this idea can be used to explain a set of interesting empirical phenomena. We begin, in the next section, with some theoretical considerations, deriving an initial motivation for the idea of conflict monitoring from current theories of cognitive control. After this, the article is divided into two main sections, corresponding to the objectives identified above. Part 1 examines the possibility that a conflict monitoring function might be reflected in the behavior of the ACC. The section begins by reviewing recent brain activation studies, which together encourage the idea that the ACC may respond specifically to the occurrence of conflict. We then present the results of a first computational modeling study, which serves to articulate our interpretation of the brain activation data and to demonstrate the sufficiency of the theory to account for them. With Part 1 as a foundation, Part 2 turns to the issue of how conflict monitoring might play a role in modulating cognitive control. The section focuses on three behavioral phenomena, quite different from one another in their details, but each involving on-line shifts in control. In a second computational modeling study, we show how linking conflict monitoring to the modulation of control leads naturally to a unifying, mechanistic explanation for these phenomena. Theoretical Background We have suggested that the systems subserving cognitive control are likely to include an evaluative system, which keeps tabs on current demands. This raises the question, what precisely might such a system measure? That is, how does a need for increased control manifest itself within the processing system? One potential answer can be derived directly from current theories of cognitive control, which portray it as serving to prevent the occurrence of conflicts in information processing. Control as Conflict Prevention Given the highly parallel and distributed character of cognitive processing, one of its inherent hazards is crosstalk interference between concurrent processes. A succinct description of the problem is provided by Mozer and Sitton (1998): One can conceive of processing ... as occurring along a certain neural pathway. If the processing pathways for two stimuli are nonoverlapping, then processing can take place in parallel. But if the pathways cross—i.e., they share common resources or hardware—the stimuli will interact or interfere with one another, (p. 342) This sort of interference is perhaps easiest to illustrate in the setting of dual-task performance. According to Navon and Miller (1987), concurrently performed tasks interfere with one another when "each produces outputs, throughputs, or side effects that are harmful to the processing of the other one, in that they change the state of some variable that is relevant for the performance of the concurrent task" (p. 435). A concrete example is provided by Shaffer (1975), who showed that dramatic decrements in performance occur in both typing to dictation and reading aloud when an attempt is made to perform these two tasks simultaneously. The difficulty of this combination can be understood as deriving from crosstalk between the processing pathways activated by auditory and visual inputs, leading to conflicting responses at the level of both speech and typing. The result is a slowing of response times and an increase in the frequency of errors, including so-called crosstalk errors where the participant produces a response in one modality that should have been delivered in the other. Conflict between concurrent processes has been understood as affecting performance in an extremely wide variety of domains. Indeed, it has been credited with placing a central limitation on human information-processing capacity: Allport (1987), in agreement with a number of other researchers (e.g., Cohen et al., 1990; Duncan, 1996; Mozer, 1991; Mozer & Sitton, 1998; Navon, 1985; Navon & Miller, 1987; Schneider & Detweiler, 1987), has argued that "the behavioral phenomena attributed in the past to the limited capacity of a central processor are more appropriately conceptualized ... as the expression of crosstalk interference between parallel processes" (p. 411). This recognition of crosstalk as a ubiquitous pitfall of parallel processing has led to a particular view of cognitive control, according to which one of its central functions is to prevent conflicts. As Allport (1980) put it, "for any distributed system, fundamental issues are raised by the demands of conflict resolution and of controlling undesirable interactions: of keeping separate processes separate" (p. 38). The job of dealing with these problems falls to cognitive control. This view can be discerned in much work on visual search, where the top-down control of visual attention has

626 BOTVINICK, BRAVER, BARCH, CARTER, AND COHEN been portrayed as helping to prevent the interference that can occur when multiple objects are processed in parallel (e.g., Mozer & Sitton, 1998; Treisman, 1988). It also informs accounts that portray attention as serving to regulate the flow of information through the processing system, favoring flow into selected processing streams while helping to gate off others (Cohen et al., 1990; Desimone & Duncan, 1995; Norman & Shallice, 1986). The idea that control serves to prevent conflicts suggests one answer to the question of how a need for increased control might manifest itself in the processing system. It implies that a need for greater control will typically be indicated by the occurrence of conflict itself. Using Conflict as a Basis for Modulating Control The potential usefulness of conflict as a basis for the regulation of control was recognized early on by Berlyne (e.g., 1960). Working within an information-theoretic framework, Berlyne proposed that the occurrence of conflict often leads to compensatory adjustments in perceptual selection, which in turn serve to alleviate conflict. In the years since Berlyne (1960) made this suggestion, the idea that conflict might be linked to the regulation of cognitive control has resurfaced intermittently, usually in the context of highly theory-driven work. For example, the production system architecture known as Soar (Laird, Newell, & Rosenbloom, 1987) proposes that problem-solving algorithms are triggered by the occurrence of impasses, one important class of which involves conflicts between simultaneously selected but incompatible productions. Conflict appears to play a similar role in the theory of control put forth by Norman and Shallice (1986). Here, a supervisory attentional system (SAS) is understood as monitoring the processes by which action schemas are routinely selected, intervening when these contention-scheduling processes prove inadequate. Although the theory does not explicitly indicate what particular events within contention scheduling serve to trigger SAS intervention, it is emphasized that contention scheduling serves primarily to prevent conflict among potentially relevant schemas (Norman & Shallice, 1986). Thus, the theory seems to imply that control is recruited when conflicts occur that contention-scheduling processes are not able to resolve efficiently. A more explicit instance is provided by the work of Schneider and Detweiler (1987, 1988). This specifies a connectionist-control framework within which a central control module regulates the exchange of information among a number of domain-specific processing modules. In this scheme, input from control is recruited when conflict occurs between messages converging on a single module. In most of this earlier work, conflict monitoring has been adopted as a background assumption, rather than a direct object of scientific inquiry. Moreover, its motivation has typically been almost entirely theoretical; conflict monitoring has been incorporated primarily because it makes sense or because it solves computational problems, rather than because of experimental evidence pointing to its occurrence. However, recent work from cognitive neuroscience has begun to provide evidence that conflict monitoring may in fact play a role in human cognition. Specifically, this work indicates that the occurrence of conflict may trigger activation in a specific area of the brain, the ACC. Part 1: Cognitive Neuroscientific Evidence for the Detection of Conflict Anterior Cingulate Cortex The ACC, situated adjacent to the corpus callosum on the medial surface of the frontal lobe,' is widely believed to play a role in cognitive control (e.g., D'Esposito et al., 1995; LaBerge, 1990; Mesulam, 1981; Posner & DiGirolamo, 1998). Beyond this general claim, however, no consensus exists as to its specific contribution to cognitive processing. ACC engagement has been reported in a remarkably wide variety of cognitive settings, including tasks that involve language, learning and memory, perceptual target detection, imagery, motor control, and dual-task performance, among other capacities (Cabeza & Nyberg, 1997; Paus, Koski, Caramanos, & Westbury, 1998), making it difficult to discern a meaningful common factor that might explain ACC engagement across studies. The notion of conflict monitoring opens up a new possibility here, for the vast majority of data from ACC activation studies appears consistent with the idea that the ACC responds to the occurrence of conflict.2 In the following section, we present an overview of ACC activation studies, dividing them into three categories and suggesting how ACC activation in each of these can be interpreted as reflecting a response to the presence of conflict. In order to make this idea explicit and support its validity, we conducted computer simulations using models of specific tasks drawn from each of the three basic areas of the ACC literature. These studies, presented here as Simulation Study 1, test the consistency of our hypotheses with existing accounts of information processing in these three domains, applying a quantitative measure of conflict to simulate findings from the ACC activation literature. Cognitive Activation of the ACC: Review of Major Findings Empirical research on the role of the ACC in cognition has been conducted using a variety of methodologies, including neuropsychological techniques (e.g., Janer & Pardo, 1991; Turken & Swick, 1999), single-unit recording (e.g., Gabriel, 1993; Niki & Watanabe, 1979), and brain activation techniques including functional neuroimaging and event-related potentials. Although neuropsychological and neurophysiological data have inspired some influential theories of ACC function (e.g., Mesulam, 1981; Vogt, Finch, & Olson, 1992), the vast majority of recent findings and some of 1 Anatomically, the anterior cingulate cortex begins above the callosum, extending forward to wrap around the genu and end inferiorly to it. However, the vast majority of the studies with which we will be concerned involve activation of the portion of the ACC posterior to the genu and superior to the callosum (cf. Bush et al., 1998; Paus et al., 1998, for discussions of functional heterogeneity in the human ACC). 2 As specified in the general discussion, the idea that the ACC responds to conflict is here viewed as part of a more general monitoring function, according to which the ACC responds to a variety of events, all indicating that attentional adjustments are needed to optimize performance or avoid negative outcomes

CONFLICT MONITORING 627 the most consistent results derive from brain activation studies. In what follows, we focus on this literature; however, our conclusions can in many instances be viewed as consistent with established findings from lesion and single-unit recording studies. Although brain activation studies have reported ACC engagement in a wide variety of task settings, the bulk of these studies can be organized into three general types. In one set of experiments, ACC activation has been associated with tasks calling for the overriding of prepotent but task-irrelevant responses; in a second group, it has been associated with tasks requiring the participant to choose among a set of equally permissible responses; and in a third, with tasks that lead to the commission of errors. Here we discuss these three domains in detail, suggesting how in each case ACC activation can be seen as accompanying the occurrence of conflict. Response override. A large number of studies have reported ACC activation in tasks requiring the participant to override relatively automatic but task-inappropriate responses. The most frequently studied of these has been the classic Stroop conflict paradigm (Stroop, 1935; for a review see MacCleod, 1991), in which the participant is asked to name the color in which a color word is displayed. Response times are greater if there is a mismatch between the color the word refers to and the color in which the word is displayed (e.g., red displayed in green) than if the two colors are the same (red displayed in red) or if the stimulus consists of a noncolor word, a series of colored Xs, or merely a color bar. The explanation usually offered for the difficulty of the incongruent condition is that word reading, a strongly automatic process, interferes with color naming. The challenge for the participant is to overcome the word-reading response. ACC activation on the Stroop task was first observed by Pardo, Pardo, Janer, and Raichle (1990). Using positron emission tomography (PET), this study demonstrated increased ACC activation during performance of the incongruent condition when compared with the congruent condition. Increased ACC activation was also shown by Carter, Mintun, and Cohen (1995) in a similar comparison. Several studies have also reported greater ACC activation in association with the incongruent condition when compared with the neutral condition (Bench et al., 1993; Carter et al., 1995; George et al., 1994). The finding of greater ACC activation with incongruent stimuli has been found in variants of the Stroop task as well; Bush et al. (1998) observed ACC activation in a numeric version of the task. Other tasks requiring the overriding of prepotent responses have also been shown to engage the ACC. Taylor, Kornblum, Minoshima, Oliver, and Koeppe, 1994, for example, asked participants in one condition to name the individually displayed letters B, J, Q, and Y. In a second condition, participants were asked to respond with the name of a different letter in the group according to a simple set of rules (e.g., if J is displayed, respond with "Y"). The latter task required them to overcome the temptation to read the letter in order to recover the less stimulus-compatible response dictated by the instructions. In agreement with the Stroop studies, increased ACC activity was observed on the conflict task. A multipart PET study by Paus, Petrides, Evans, and Meyer (1993) showed that the need to override prepotent responses will activate ACC across tasks involving a range of input and output modalities. In one set of experiments, participants first performed according to extensively practiced stimulus-response pairings and later according to a novel mapping. In one version of the experiment, participants were trained to respond to each of three simple visual stimuli with a direction-specific saccade. In the reversal condition, the pairing between the three stimuli and the three saccade responses was changed. Two other versions of the experiment involved mappings from visual stimuli to buttons to be pressed and from heard words to spoken words. In each version, the reversal condition elicited greater ACC activation than the overlearned condition. In a second set of experiments, Paus et al. (1993) asked participants first to produce stimulus-compatible responses, and later to produce responses less congruent with the stimulus. In one version, participants first lifted whichever of two fingers was touched by the experimenter. Later, participants were instructed to raise the opposite finger. In a second version, participants performed a saccade in the direction of either a left-sided or right-sided visual cue, and then later were asked to respond with a saccade in the direction opposite the cue. In a third version, participants responded to the two heard letters "A" and "L" by naming the letter coming next in the alphabet. In the reversal condition, participants responded to "A" with "M" and to "L" with "B." In each version of the experiment, greater ACC activation was once again observed on the task requiring the participant to overcome an ingrained response in favor of a less familiar one. Another instance of ACC activation associated with response override is provided by studies of go/no-go tasks. Using functional magnetic resonance imaging (fMRI), Casey et al. (1997; see also Kawashima et al., 1996) had participants view a series of individually presented letters, pressing a button with each presentation but omitting this response if the presented letter was an X. The majority of trials involved non-X letters, leading the button-press response to be prepotent. In control conditions, the presented letter series contained no Xs. Greater ACC activation was observed in the go/no-go condition. As in other response override tasks, ACC activation is here associated with conditions that require the participant to overcome a prepotent response in order to perform successfully. The finding of ACC engagement in response override tasks provides a first piece of evidence for the view that this brain area responds to the occurrence of conflict. In each of the studies we have reviewed, the strongest ACC activation was observed under conditions where it was necessary for the participant to overcome interference from prepotent but task-irrelevant responses. These circumstances can be understood as involving conflict between processing pathways leading to correct (but otherwise weaker) and incorrect (but prepotent) responses. The mechanisms responsible for this form of crosstalk are considered further in Simulation 1 A. Underdetermined responding. In a second group of studies, ACC activation occurs under conditions requiring the participant to choose from a set of responses, none of which is more obvious or compelling than the others. We describe these tasks as involving underdetermined responding, because the stimulus presented to the participant does not uniquely specify the appropriate response. The first studies to examine brain activation under such task circumstances were reported by Petersen, Fox, Posner, Mintun, and Raichle (1988, 1989). In a series of PET studies, the group asked participants to generate a verb in response to a seen or heard noun, identifying a use for the object named by the stimulus. When activation patterns for this task were compared with those for a condition in which the participant simply repeated or read the presented word, the ACC was found to be consistently engaged

628 BOTVINICK, BRAVER, BARCH, CARTER, AND COHEN The finding has been replicated in a number of studies from other laboratories (e.g., Andreason et al., 1995; Barch, Sabb, Braver, & Noll, 2000; Thompson-Schill, D'Esposito, Aguirre, & Farah, 1997), in some cases with verb generation carried out silently (Warburton et al., 1996; Wise et al., 1991). In the related letter fluency (or FAS) task, participants are asked to list words beginning with a given letter (Spreen & Benton, 1969). Here again, the participant selects freely among a number of potential responses. Letter fluency has been repeatedly shown to activate ACC, in comparisons with simply repeating the lettername cue (Friston, Frith, Liddle, & Frackowiak, 1993), repeating heard words (Frith, Friston, Liddle, & Frackowiak, 199la), or performing a lexical decision task (Frith, Friston, Liddle, & Frackowiak, 1991b). Yetkin et al. (1995), using fMRI, found ACC activation even if participants generated letter fluency responses without voicing them aloud. Semantic fluency, in which the task is to name members of a given category, also activates ACC (Yetkin et al., 1995), as has stem completion, another task involving underdetermined responding (Buckner et al., 1995). Activation of ACC under conditions of underdetermined responding is not limited to verbal tasks. Frith and colleagues found it when participants were asked to lift either of two fingers, chosen at random, when one of the fingers was tapped, in comparison with a condition where participants were instructed to lift the tapped finger (Frith, Friston, Liddle, & Frackowiak, 199la). Deiber et al. (1991) compared PET activation patterns when participants were asked to move a joystick randomly in any of four directions with a condition in which they moved it repeatedly in only one specified direction, finding relative ACC activation in the free selection condition, a finding replicated by Playford et al. (1992) and (with button presses) Jeuptner, Frith, Brooks, Frackowiak, and Passingham (1997). As in response override tasks, ACC activation in underdetermined responding is consistent with the view that the ACC is engaged by the occurrence of conflict. Because the stimuli involved in underdetermined responding tasks are each associated with a number of legal responses, stimulus presentation may lead to the parallel activation of multiple incompatible response pathways, resulting in crosstalk during the period between stimulus presentation and response delivery. In support of this interpretation (examined more critically in Simulation IB below), Raichle et al. (1994) showed that the verb generation task no longer produced detectable ACC activation once participants had encountered the same list of nouns several times and their responses had become well rehearsed. Activation was restored when a new list of nouns was later presented, once again placing the participant in the position of generating underdetermined responses. Similarly, in the Deiber et al. (1991) joystick movement study, no difference in ACC activation was observed between the single-direction condition and conditions where participants moved the joystick according to a previously learned sequence or on the basis of a direction-specifying tone. Again, increased ACC activation was noted only when the stimulus is likely to have activated pathways to multiple, mutually interfering response representations. Error commission. In a third group of studies, ACC activity has been observed in association with the commission of errors. In contrast to the work discussed so far, using PET or fMRI, indications of a connection between ACC activity and errors comes primarily from studies of event-related potentials in electroencephalographic (EEG) recordings (Rugg & Coles, 1995). The term error-related negativity (ERN) refers to a discrete event-related potential that has been described as accompanying the commission of errors in a number of speeded response tasks (e.g., Falkenstein, Hohnsbein, & Hoorman, 1995). The potential, independently discovered by two laboratories in 1989 and 1990 (Gehring, Coles, Meyer, & Donchin, 1990; Hohnsbein, Falkenstein, & Hoorman, 1989), is best seen in response-aligned averages over error trials, where it usually appears with the onset of response-related electromyographic (EMG) activity, peaking 100- 150 msec later. The ERN (also designated as Ne) has been demonstrated in a variety of task settings. Gehring and colleagues (Gehring et al., 1990; Gehring, Coles, Meyer, & Donchin, 1995; Gehring, Goss, Coles, Meyer, & Donchin, 1993) used versions of the Eriksen flanker and Sternberg memory search tasks and a category judgment task requiring participants to indicate whether one of two displayed words represented an exemplar of the class named by the other. Falkenstein and colleagues used two- (Falkenstein, Hohnsbein, Hoorman, & Blanke, 1991) and four-way (Falkenstein et al., 1995) forced-choice letter discrimination tasks (cf. Bernstein, Scheffers, & Coles, 1995). Dahaene, Posner, and Tucker (1994) have also observed the ERN in a task requiring participants to indicate with a rapid keypress whether viewed numbers (displayed either as an Arabic numeral or in word form) were greater or less than 5, and in another task whether viewed words denoted animals. The ERN has also been observed in association with errors of commission in go/no-go tasks of varying design (Falkenstein et al., 1995; Scheffers, Coles, Bernstein, Gehring, & Donchin, 1996). The generator of the ERN has consistently been localized to a medial frontal region. Dahaene et al. (1994), applying a dipole localization technique to EEG data, judged the source of the potential to lie in the ACC. Given the limited spatial resolution of the technique, however, a localization in supplementary motor cortex could not be ruled out. Carter et al. (1998), in a study discussed in more detail below, used fMRI to evaluate regional activity associated with incorrect versus correct responses in a version of the Continuous Performance Test, confirming that error responses were accompanied by temporally and anatomically specific activation of ACC. As discussed in Simulation Study 1C, it appears likely that errors are associated with conflict due to interference between the pathways leading to correct and incorrect responses. Behavioral data indicates that errors in speeded response tasks frequently represent premature responses delivered while stimulus analysis is still incomplete (Gratton, Coles, Sirevaag, Eriksen, & Donchin, 1988). Even as such impulsive errors are executed, stimulus evaluation can continue, leading to activation of the correct response (Rabbitt & Vyas, 1981). The very short latency of error-correcting movements confirms that activation of the correct pathway can take place even while an incorrect response is being delivered (Cooke & Diggles, 1984; Rabbitt & Rodgers, 1977). This makes it seem likely that errors will frequently be associated with conflict between the coactivated pathways leading to correct and incorrect responses. More direct evidence for this idea is provided by a recent study by Gehring and Fencsik (1999). Participants in this study performed the flanker task, responding using the left hand for one target and the right hand for the other. EMG was used to measure

CONFLICT MONITORING 629 the strength of the response made with each hand. Participants very frequently reversed errors, and the EMG results indicated clearly that when this occurred, there was typically temporal overlap between the error and error-correcting responses. This same study provides evidence consistent with the idea that this transient response conflict is n critical factor in the generation of the ERN; EEG data indicated that the ERN coincided with the period of response overlap on error trials. A number of other findings corroborate the connection between response conflict and the ERN. First, error trials associated with the largest ERN amplitudes more frequently involve response reversals than do trials with smaller ERNs (Gehring et al., 1993). Thus, the largest ERNs are associated with error trials where there is the strongest evidence for belated activation of the correct response. Second, an ERN appears, even in association with correct responses, if these are subsequently reversed (Gehring et al., 1993).3 Third, in a study where participants were asked to withhold their responses until 2 s after stimulus presentation, no ERN was observed in association with errors (Dahaene et al., 1994). Given such a delay, any transient competition between processing pathways is likely to have resolved by the time of response delivery. Thus, incorrect responding in the Dahaene et al. study is unlikely to have been accompanied by the sort of crosstalk to which we attribute the ERN. Residual studies. The close association between conflict and ACC activation in the studies we have reviewed is reinforced by the fact that conflict also appears to play a role in ACC activation studies that do not fall into any of the three categories we laid out above. For example, D'Esposito et al. (1995) used fMRI to compare ACC activity during two simple tasks performed either singly or concurrently, observing greater activation in the latter condition. On the basis of the earlier discussion of the role of crosstalk in dual-task performance, it is clear how ACC activation in this study can be explained as a response to conflict. In another divided attention study, Corbetta, Miezin, Dobmeyer, Shulman, and Petersen (1991) measured brain activity with PET while participants monitored forms in a visual display for subtle changes along the dimensions of color, shape, and direction of movement. In a focused attention condition, participants monitored only one of these dimensions. In a divided attention condition, participants searched for changes in any of the three dimensions. Greater ACC activation was associated with the divided attention condition. Participants made more errors in this condition, and so it may be possible to attribute ACC activation in this study to errors. However, another interesting (and closely related) possibility is that the parallel evaluation of different stimulus dimensions led on some trials to crosstalk between pathways supporting "same" and "different" responses. Although the published data do not allow a conclusive evaluation of this possibility, it is consistent with the reported higher frequency of misses (incorrect "same" judgments) in the divided attention condition. In another study, Baker et al. (1996) found ACC activation in association with performance of the Tower of London task. Because the solution to this task is rarely immediately apparent to the unpracticed participant, competition or conflict among alternative actions seems likely to be involved. As we have already noted, certain computational models accord such competition a pivotal functional role in triggering problem-solving behavior (e.g., Laird et al., 1987), a point to which we will return in later discussion. Finally, Grasby et al. (1993) had participants listen to and immediately repeat word lists from 2 to 13 items long. Using PET, they found that ACC activation increased with list length. As in the Corbetta et al. (1991) study, one option is to attribute this finding to errors, for the frequency of errors rose along with list length. However, there are again other potential explanations that involve conflict. One is that, as list length grew, greater response competition occurred during the retrieval process. Even more intriguing is the possibility that ACC activation may have been related to interference among lexical representations being maintained in working memory. One way of examining this latter possibility is provided by the phonological similarity effect, the fact that participants asked to repeat a list of words shows relatively poor memory for the list if it is composed of similar-sounding entries (Baddeley, 1966). If, as appears reasonable, this phenomenon can be assumed to derive from interference among representations being held in working memory, then a potentially informative experiment might be to measure ACC activity during retention of short word lists, comparing activity levels during maintenance of phonologically similar and dissimilar items. If the ACC is responsive to conflict among representations in working memory, then greater activation should be seen in the condition using phonologically similar words. Accounting for ACC Activation: Simulation Study 1 In connection with each group of studies reviewed above, we briefly proposed how ACC activation can be understood as a response to the occurrence of conflicts in information processing. In the present section, we describe a set of three computer simulations in which this interpretation of the ACC literature is more fully articulated. The objectives of these modeling studies were threefold. A first goal was to make the account we have presented so far more explicit, providing a precise indication of what we intend by such terms as crosstalk, conflict, top-down control, and conflict monitoring. A second goal was to confirm the sufficiency of these constructs, as we have used them in the conflict monitoring hypothesis, to account for the results of ACC activation experiments. The third goal was to lay the groundwork for further modeling work, reported in Part 2, that examines the entire feedback loop running from conflict monitoring to cognitive control. Each of the present simulations makes use of a previously and independently implemented computational model of a single task from one of the three primary domains in which ACC activation has been reported. To examine the role of conflict in response override, we consider a model of the Stroop task (Cohen & Huston, 1994); for underdetermined responding, a model of stem completion (McClelland & Rumelhart, 1981); and to evaluate the relation between conflict and error commission, we examine a model of the Eriksen task (Servan-Schreiber, 1990), the task most frequently used in studies of the ERN. To each of these models, we apply a quantitative measure of conflict, allowing the models to be used in simulations of the 3 Although trials involving overt reversals are likely to involve the strongest coactivation of correct and incorrect responses, the account we are proposing does not require that any actual reversal occur, only that activation of the correct pathway occur while activation of the incorrect pathway is still present

630 BOTVINICK, BRAVER, BARCH, CARTER, AND COHEN behavior of the ACC as it has been observed in brain activation experiments. Each study provides an explicit account of the mechanisms that give rise to conflict, comparing their role across task conditions that have been shown to engage the ACC to different degrees. General Methods Selection of models. Although the models we consider are examined in a novel context, they are not themselves new. Each has been drawn from the published literature and is considered here in its original form. The fact that these models were formulated independently of present hypotheses allows us to approach them quasi-empirically, using them to test the consistency of the conflict monitoring hypothesis with current theories of information processing in specific tasks known to engage the ACC. Leaving the models' original parameters intact and using the same simple computation to determine conflict across all three studies reduces the number of free parameters associated with our simulation of ACC activation to zero. Of course, these points should not be taken to imply that the models used were selected in a disinterested or theory-neutral fashion. On the contrary, the three models implement a shared set of basic assumptions about information processing that also form part of the background for the conflict monitoring hypothesis. Specifically, they assume that information processing is parallel, distributed, and interactive. These assumptions are captured in the connectionist framework, within which all three models were conceived (McClelland, 1992; Rumelhart & McClelland, 1986). Like other connectionist models, the ones we will consider here are composed of identical processing units, each carrying a realvalued activity level, which excite and inhibit one another through weighted connections. When external input is applied to a subset of the units, information propagates through the network, resulting in a final output activation pattern. Information processing relies upon the strength of the network's connections, which can either be set by hand or by a number of training algorithms. Again, these values were set in previous studies and used unmodified in our current simulations. Implementing conflict monitoring. In each simulation, the underlying model adopted from the literature is extended by the addition of a single conflict-monitoring unit (see Figures 1-3). This unit takes input from the basic network and computes the current amount of conflict prevailing there. This component of the simulations raises the important question of how conflict might be measured. As a first step toward defining a method for accomplishing this, conflict may be operationally defined as the simultaneous activation of incompatible representations. In the models we will consider here, incompatible representations (e.g., representations of alternative responses) correspond to units interconnected by inhibitory weights. Thus, conflict can here be defined as the simultaneous activation of mutually inhibiting units. Although this makes it clear what conflict involves at a qualitative level, it is a more difficult question how conflict should be quantified. Berlyne (1957, 1960), who discusses this issue at length, offers a list of desiderata for a measure of conflict: (a) It should increase with the absolute activation of the competing representations; (b) it should increase with the number of competing representations; and (c) it should be maximal when the activations of competing representations are equal. Although it is an empirical question how conflict might be measured by the brain (a point we consider further in the General Discussion), Berlyne's criteria provide a reasonable starting point for considering alternative possibilities. Berlyne noted that there are many potential measures of conflict that would meet his specifications. He himself adopted one based on the information-theoretic expression for entropy (specifically, this involves multiplying entropy by the average activation in the set of competing representations). In the present context, this approach carries the technical disadvantage that it requires activation levels to be translated into probability values, a step that in turn requires peripheral assumptions. In the present studies, we chose a different measure of conflict— Hopfield energy—which satisfies Berlyne's criteria while being based on values specified directly by the models we examine. Hopfield (1982) defined the energy in a recurrent neural network as (1) where a indicates unit activity and both subscripts are indexed over all units in the set of interest (related measures are discussed by Rumelhart, Smolensky, McClelland, & Hinton, 1986, and Smolensky, 1986). To see how energy reflects conflict, consider a single pair of mutually inhibiting (incompatible) units. When both are inactive, energy is zero, consistent with the absence of conflict. Energy remains at zero if only one of the units becomes active, once again mirroring the level of conflict. Energy rises above this level only if both units are active. The particular value for energy then depends on the activation values of the two units, becoming largest when both units are maximally active and thus most strongly in conflict.4 Note that his implementation of conflict does not involve any additional parameters, and this preserves the zero-parameter nature of our simulations. As in the cognitive system, conflicts between representations in connectionist networks can occur at a variety of levels of processing, including stimulus evaluation, memory and set representation, and response selection. In the simulation studies presented here, we focus exclusively on the role of response conflict, measuring energy over units in the output layer of each model. We were motivated in this choice by the commonality of response selection processes among tasks that involve response override, underdetermined responding, and error commission, which led us to hypothesize that ACC activation in these domains might be accounted for in terms of conflict at this level of processing. Although this is the hypothesis addressed in the simulation studies, there are reasons for leaving open the possibility that conflict at other levels of processing might also be relevant to ACC activation, a point to which we return in later discussion. Simulation procedure. In each study, the underlying model is used to simulate information processing in conditions that have been reported to engage the ACC to different degrees. In the 4 For completeness, it is worth noting that concurrent activation of units interconnected by excitatory links causes a reduction in energy. In this regard, energy is more than a measure of conflict; it measures compatibility or consistency. This interesting aspect of Hopfield's (1982) formula does not come into play in the simulations to be reported here, where units share only inhibitory connections

CONFLICT MONITORING 631 Stroop model, congruent, neutral, and incongruent trial conditions are compared; in the stem completion model, the stem completion task is compared with word reading; and in the Eriksen model, correct responses are compared with errors. Except where explicitly noted, simulations are run according to the procedure originally used for each of the basic models as reported in the literature. With each step of processing, the conflict monitoring unit assumes an activity equal to the current level of energy in the output layer of the underlying model. In each study, the activity of this unit is compared across conditions, with the prediction that the greatest activity will be observed in the condition associated with the strongest ACC activation. Simulation 1A: The Stroop Task In this first simulation study, we introduce the basic elements of the proposed framework by considering the origins of ACC activation—and, by implication, the role of conflict monitoring—in a response override task. Method. Stroop performance was simulated using a model proposed by Cohen and Huston (1994), shown in Figure 1 (left). This model is based on an earlier feed-forward model (Cohen et al., 1990), revised to include recurrent connections and interactive processing (both of which are amenable to our measurement of conflict). The model includes input units for display color and word identity. The appropriate units in each group connect reciprocally via excitatory weights to an output layer with units representing potential responses. In addition, the model includes a task demand layer with units standing for word reading and color naming, respectively. The task demand units serve to bias activation in the model so that either word or color inputs may dominate response activation. As shown in Figure 1, units within every layer are interconnected by symmetrical negative weights. The procedure used in simulating a trial is detailed in Cohen and Huston (1994). Briefly, one of the task units is activated during an initial priming interval, during which the output units are inhibited to prevent premature responses. The input pattern is then applied and the response-layer inhibition removed. As illustrated in Figure 1, the current simulation added a conflict monitoring unit that takes inputs from the response layer of the underlying model, taking on an activation level equal to the energy in that layer on the current cycle of processing. In order to account for the findings regarding ACC activation in neuroimaging studies of the Stroop task, the activation of the conflict monitoring unit was evaluated during simulation of incongruent, congruent, and neutral conditions in the color-naming task. Results. Results are shown in Figure 1 (right). As with ACC activation in neuroimaging studies of the Stroop task, activation of the conflict monitoring unit was higher in the incongruent condition than in the congruent or neutral conditions. As shown in the figure, activation rose rapidly in all three conditions; this is because both output units move toward their (nonzero) resting activity levels once the inhibition they receive prior to stimulus presentation is removed. Differences in energy across conditions soon appeared, however, with incongruent trials associated with the highest degree of energy. The increased activity of the conflict monitoring unit on incongruent trials reflects the occurrence of crosstalk within the Cohen and Huston (1994) model. On incongruent trials, word and color inputs each activate a different set of units in their corresponding pathways. The intersection of these two pathways in the output layer (in addition to the other sectors of the model) causes conflict between the response units, and this in turn raises the activity of the conflict monitoring unit. Discussion. Response override tasks have been repeatedly observed to engage the ACC. This first simulation provides an Conflict ffitofiftorinQ RMDORM Ink Color Task Dwnand Word 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Cycte Figure 1. Left panel: Illustration of the Stroop model discussed in Simulation Study 1. From "Progress in the Use of Interactive Models for Understanding Attention and Performance," by J. D. Cohen and T. A. Huston, 1994, in C. Umilta and M. Moscovitch, Attention and Performance XV (Figure 18.8, p. 462), by J. D. Cohen and T. A. Huston, 1994, Cambridge, MA: MIT Press. Copyright 1994 by MIT Press. Adapted with permission. Eq. = equation; R = red; G = green; C = color-naming; W = word-reading; N = neutral. Right panel: Energy as measured in the response layer of the Cohen and Huston (1994) model during simulation of congruent, neutral, and incongruent trials in the Stroop task. Arrows indicate average response times. Energy was recorded for each cycle of processing, and the data shown are means based on 100 trials in each condition

632 BOTVINICK, BRAVER, BARCH, CARTER, AND COHEN illustration of how the idea of conflict monitoring can be used to explain this finding. When an element is added to an existing model of a typical response override task, acting to transform the occurrence of conflict into an activation-based signal, a pattern is observed across conditions that parallels that observed in ACC activation studies. The color-naming condition of the Stroop task is a classic example of controlled information processing, and in fact, the Cohen and Huston (1994) model was originally proposed as a basic model of control function. Control is implemented here through the color-naming and word-reading units, insofar as these units bias information flow through the rest of the system in accordance with task demands. It is interesting that varying the control signal coming from these units impacts the degree to which conflict occurs during stimulus processing. In simulations of the color-naming task, specifically, weakening the input from the color-naming unit on incompatible trials leads to increasing interference between color and word inputs and, thus, to higher peak energy.5 This aspect of the model's behavior fits well with the idea that conflict might serve as an indicator of insufficient control, as it means that conflict is most likely to occur when control is weak. It also translates into a testable prediction: If ACC activation reflects conflict detection, then, on the basis of the model, ACC activation during incongruent trials in the Stroop task should vary inversely with the strength of control, defined as the effort to attend exclusively to color. We recently tested this prediction in a functional neuroimaging study (Carter et al., 2000). Here, the strength of top-down control was influenced indirectly by manipulating trial-type frequency. As shown by a number of behavioral studies (e.g., Lindsay & Jacoby, 1994; Logan & Zbrodoff, 1979), participants display a smaller Stroop interference effect if incongruent trials are frequent than if they are rare. In our terms, frequent incongruent trials lead to a high-control state (a tight focus on the color-naming task as opposed to the word-reading task). The Carter et al. (2000) study exploited this phenomenon to test for the predicted relationship between control state and ACC activation. Participants performed the Stroop task while undergoing fMRI. Trial-type frequency was varied across blocks; in one half of the blocks, incongruent trials occurred frequently, in the other half, relatively infrequently. Behavioral results confirmed the expected effect of trial-type frequency on control state. Participants were faster on incompatible trials when these were frequent than when they were rare. Eventrelated scan acquisition allowed evaluation of the time course of ACC activation on individual trials. As predicted, peak activation on incongruent trials differed as a function of trial-type frequency, with greater activity occurring during blocks where incompatible trials were rare. Simulation IB: Stem Completion As in response override tasks, we have attributed ACC engagement in underdetermined responding tasks to the engagement of a conflict monitoring function. Using the approach taken in Simulation 1A, this proposal was tested against a relevant model of information processing, in this case a model of the stem completion task. Method. Stem completion can be simulated using the interactive activation (IA) model of word reading introduced by McClelland and Rumelhart (1981; Rumelhart & McClelland, 1982), illustrated in Figure 2 (left). The model consists of three interconnected sets of processing units. External input is applied to a layer encoding featural elements—vertical, horizontal, and diagonal line segments—from which individual letters are constructed. Activation feeds forward from this feature layer to a layer of units representing individual letters. This layer connects to a third set of units, each standing for an individual four-letter word. Between layers, compatible units (e.g., the unit for the letter A in the first slot and the word unit for ALSO) are connected by excitatory weights, and incompatible ones by inhibitory weights. There are also symmetrical inhibitory connections between each pair of units in the word layer. Stem completion can be simulated in the IA model by presenting letters in the first two positions, leaving the third and fourth slots blank. Given such input, the model completes it, settling into a final state dominated by a word unit (and corresponding letter units) representing a word beginning with the two letters presented, similar to what would have resulted had all four letters of the word been present. As in Simulation 1A, a conflict monitoring element was added to the underlying model. Here, the conflict monitoring unit takes its input from the units in the word layer and assumes an activation equal to the current level of energy in that layer. In order to account for the finding of ACC activation in association with stem completion, the activity of the conflict monitoring unit was evaluated during simulations of both stem completion (partial input) and word reading (full input). A total of 20 words were chosen at random from the corpus represented in the word-unit layer of the model. In the word-reading condition, each word was presented in full to the feature layer. In the stem-completion condition, only the first two letters of each word were presented, with the last two slots receiving no input. As in Simulation 1A, energy was measured at regular intervals throughout each trial. Results. Results are shown in Figure 2 (right). Whereas presenting a full word led to only a fleeting rise in energy, stem presentation led to much greater and sustained levels of energy. As in Simulation 1A, these results can be understood as deriving from the different degrees of crosstalk involved in the two task conditions. Although the final outcome in the word layer is similar for full words and word stems, the paths by which the network reaches its final representation entail quite different amounts of crosstalk. For full words, the process is fairly straightforward. The input for each letter activates its corresponding feature units and letter unit. The selected letter units together strongly activate one word unit. Although subgroups of letters might also weakly activate a small set of other words (e.g., the FIS in FISH might activate FIST), the support for the fully specified word is stronger, and this word unit quickly dominates the word layer. The small increase in energy associated with word reading corresponds to the minor conflict among words partially matching the input. For word stems, processing unfolds differently. Initially, the input activates one letter unit in each of the first two letter positions. These two units together activate a wide range of word units (FI_ will activate FISH, FIND, FINE, FIRE, etc.). These word units compete through inhibitory interconnections, also sending activation to the letter units associated with them in the third and fourth positions of the letter layer. Although this conflict is ultimately resolved in 5 The effects of varying task-unit input were first explored in an earlier version of the model by Cohen et al. (1990). Usher and Cohen (2000) have replicated and extended these findings in the context of the Cohen and Huston (1994) model. Simulation 2B in the present article also contains relevant findings

CONFLICT MONITORING 633 Conflict monitoring 0.07 I I'l l I 1 I 1 * Cycto Figure 2. Left panel: Illustration of the IA model. From "An Interactive Activation Model of Context Effects in Letter Perception: Part I. An Account of Basic Findings," by J. L. McClelland and D. E. Rumelhart, 1981, Psychological Review, 88, Figure 3, p. 380. Copyright 1981 by the American Psychological Association. Eq. = equation. Right panel: Energy as measured in the word layer of the IA model during simulation of word reading and stem completion. *Energy during simulation of stem completion with the weight of inhibitory connections in the letter layer set to match those in the word layer, in order to prevent two-way ties between words. Energy was recorded every five cycles of processing. favor of one word that completes the input pattern,6 the degree of crosstalk sustained in the response selection process is much greater than for word reading, giving rise to the greater amount of energy observed. Discussion. Underdetermined responding tasks make up an important subset of the tasks that have been shown to activate the ACC. The results of the present simulation demonstrate how the engagement of the ACC in this setting can be understood in terms of a conflict monitoring function. This result stems from the same factor that produced the results of Simulation IA. On the basis of the models we have adopted, conflict in both types of task is the result of crosstalk between processing pathways. Thus, despite the superficial differences between response override and underdetermined responding tasks, the ACC activation associated with both can be understood as a response to precisely the same type of system state. As in Simulation 1 A, consideration of the factors that affect the degree of conflict in the underlying model leads to testable predictions. One example involves the fact that, in the IA model, the degree of crosstalk associated with stem completion depends on the specific stem tested. Energy varies with the degree to which words other than the eventual winner are excited by the stem. Stems that activate one completion much more strongly than any other will be associated with the least conflict, because the preferentially activated word unit quickly suppresses its competitors. Considered in the context of the conflict monitoring hypothesis, this leads to the prediction that stem completion should engage the ACC more strongly when the stem presented is associated with several completions than when the stem is associated with one strongly preferred response. A finding related to this prediction has been reported in the context of another underdetermined responding task, verb generation. Thompson-Schill et al. (1997) recorded the frequency with which specific responses were elicited in this task by a set of nouns. For each noun, they divided the frequency of the most frequent response by that of the second most frequent, using this response strength ratio as an index of the degree to which each noun was associated with a single predominant response. Using fMRI, the group compared brain activation during completion of stems with high and low response strength ratios. Consistent with the conflict monitoring hypothesis, greater ACC activation was observed for low-response-ratio nouns. This finding was recently replicated by Barch et al. (2000) in a study that also tested further predictions based directly on the 6 In simulations of stem completion using the model's original parameters, the settling process sometimes resulted in a two-way tie between word units. This is reflected in a plateau in the average energy trajectory, as shown in Figure 2. We tested whether the occurrence of these ties might be responsible for the higher levels of energy during stem completion by introducing reciprocal inhibitory weights between each pair letter units, similar to those in the word layer (and equal in strength). This had the effect of eliminating deadlocks, but without otherwise affecting the differences between processing in the word-reading and stem-completion conditions. The resulting energy trajectory, shown as a dashed line in Figure 2 (right), remained significantly greater in amplitude than the baseline condition, confirming that the differences in energy between the two conditions were not due to the incidental occurrence of two-way ties, but instead to the transient competition among processing pathways triggered by presentation of the word-stem input

点击进入文档下载页（PDF格式）

共29页，试读已结束，阅读完整版请下载

点击下载（PDF格式）

浏览记录