Dissociable networks for the expectancy and perception of emotional stimuli in the human brain

William James posited that comparable brain regions were implicated in the anticipation and perception of a stimulus; however, dissociable networks (at least in part) may also underlie these processes. Recent functional neuroimaging studies have addressed this issue by comparing brain systems associated with the expectancy and perception of visual, tactile, nociceptive, and reward stimuli. In the present fMRI study, we addressed this issue in the domain of pictorial emotional stimuli (IAPS). Our paradigm involved the experimental conditions emotional expectancy, neutral expectancy, emotional picture perception, and neutral picture perception. Specifically, the emotional expectancy cue was uncertain in that it did not provide additional information regarding the positive or negative valence of the subsequent picture. Neutral expectancy and neutral picture perception served as control conditions, allowing the identification of expectancy and perception effects specific for emotion processing. To avoid contamination of the perception conditions by the preceding expectancy periods, 50% of the pictorial stimuli were presented without preceding expectancy cues. We found that the emotional expectancy cue specifically produced activation in the supracallosal anterior cingulate, cingulate motor area, and parieto-occipital sulcus. These regions were not significantly activated by emotional picture perception which recruited a different neuronal network, including the amygdala, insula, medial and lateral prefrontal cortex, cerebellum, and occipitotemporal areas. This dissociation may reflect a distinction between anticipatory and perceptive components of emotional stimulus processing.     

Cerebral correlates of alerting, orienting and reorienting of visuospatial attention: an event-related fMRI study

The identification of brain systems contributing to different aspects of visuospatial attention is of both clinical and theoretical interest. Cued target detection tasks provide a simple means to dissociate attentional subcomponents, such as alerting, orienting or reorienting of attention. Event-related functional magnetic resonance imaging (fMRI) was used to study neural correlates of these distinct attentional processes. Volunteers were scanned while performing a centrally cued target detection task. Four different types of trials (no cue, neutral cue, valid cue and invalid cue trials) with targets appearing either in the right or left hemifield were randomly intermixed. Behaviourally, the data provided evidence for alerting, spatial orienting and reorienting of attention. Neurally, the alerting effect was seen in bilaterally increased extrastriatal blood oxygenation level-dependent (BOLD) activity in neutral as compared to no cue trials. Neural correlates of spatial orienting were seen in anterior cingulate cortex, which was more active during valid as compared to neutral cue trials. Neural correlates of reorienting of attention, that is, higher BOLD activity to invalid as compared to validly cued trials were evident in several brain regions including left and right intraparietal sulcus, right temporo-parietal junction and middle frontal gyrus bilaterally. The data suggest that frontal and parietal regions are specifically involved in reorienting rather than orienting attention to a spatial position. Alerting effects were seen in extrastriate regions which suggest that increased phasic alertness results in a top-down modulation of neural activity in visual processing areas.     

Left and right occipital cortices differ in their response to spatial cueing

We investigated cue and target-related laterality effects with event-related fMRI. Both left and right occipital areas responded maximally when both cue and target were presented in the contralateral visual hemifield (VF), and minimally when cue and target were presented in the ipsilateral VF. However, whereas signal increases in right ventromedial and lateral occipital cortex were intermediate in those trials in which the cue appeared in the VF contralateral to the target (invalid cue trials), signal strength in left occipital cortex was almost identical for valid and invalid cues, i.e., high for RVF cues, and low for LVF cues, independent of the VF of the target. These data support theories which postulate a greater ability of the right hemisphere for bilateral processing. However, these laterality effects were observed earlier in the visual pathway than previously thought, leading to the question whether the hemispheric differences observed in occipital cortex are generated in the activated areas or are the effect of reentrant processes from more anterior areas, potentially in parietal cortex.     

Functional connectivity of the frontoparietal network predicts cognitive modulation of pain

The experience of pain can be significantly influenced by expectancy (predictive cues). This ability to modulate pain has the potential to affect therapeutic analgesia substantially and constitutes a foundation for nonpharmacological pain relief. In this study, we investigated (1) brain regions involved in visual cue modulation of pain during anticipation of pain, pain administration, and pain rating; and (2) the association between pretest resting state functional connectivity and the magnitude of cue effects on pain ratings. We found that after cue conditioning, visual cues can significantly modulate subjective pain ratings. Functional magnetic resonance imaging results suggested that brain regions pertaining to the frontoparietal network (prefrontal and parietal cortex) and a pain/emotion modulatory region (rostral anterior cingulate cortex) are involved in cue modulation during both pain anticipation and administration stage. Most interestingly, we found that pretest resting state functional connectivity between the frontoparietal network (as identified by independent component analysis) and the rostral anterior cingulate cortex/medial prefrontal cortex was positively associated with cue effects on pain rating changes. We believe that these findings will shed new light on our understanding of variable cue/expectancy effects across individuals and how the intrinsic connectivity of the brain may influence expectancy-induced modulation of pain.     

Predictability modulates the affective and sensory-discriminative neural processing of pain

Knowing what is going to happen next, that is, the capacity to predict upcoming events, modulates the extent to which aversive stimuli induce stress and anxiety. We explored this issue by manipulating the temporal predictability of aversive events by means of a visual cue, which was either correlated or uncorrelated with pain stimuli (electric shocks). Subjects reported lower levels of anxiety, negative valence and pain intensity when shocks were predictable. In addition to attenuate focus on danger, predictability allows for correct temporal estimation of, and selective attention to, the sensory input. With functional magnetic resonance imaging, we found that predictability was related to enhanced activity in relevant sensory-discriminative processing areas, such as the primary and secondary sensory cortex and posterior insula. In contrast, the unpredictable more aversive context was correlated to brain activity in the anterior insula and the orbitofrontal cortex, areas associated with affective pain processing. This context also prompted increased activity in the posterior parietal cortex and lateral prefrontal cortex that we attribute to enhanced alertness and sustained attention during unpredictability.     

Getting mixed messages: The impact of conflicting social signals on the brain's target emotional response

Amidst a barrage of sensory information in the environment, the impact that individual stimuli have on our behaviour is thought to depend on the outcome of competition that occurs within and between multiple brain regions. Although biased competition models of attention have been tested in visual cortices and to a lesser extent in auditory cortex, little is known about the nature of stimulus competition outside of sensory areas. Given the hypothesized role of multiple pathways (cortical and subcortical) and specialized brain regions for processing valence information, studies involving conflicting basic emotional stimuli provide a unique opportunity to examine whether the principles of biased competition apply outside of sensory cortex. We used fMRI to examine the neural representation and resolution of emotional conflict in a sample of healthy individuals. Participants made explicit judgments about the valence of happy or fearful target facial expressions in the context of emotionally congruent, neutral, or incongruent distracters. The results suggest that emotional conflict is reflected in a dissociable manner across distinct neural regions. Posterior areas of visual cortex showed enhanced responding to congruent relative to neutral or incongruent stimuli. Orbitofrontal cortex remained sensitive to positive affect in the context of conflicting emotional stimuli. In contrast, within the amygdala, activity associated with identifying positive target expressions declined with the introduction of neutral and incongruent expressions; however, activity associated with fearful target expressions was less susceptible to the influence of emotional context. Enhanced functional connectivity was observed between medial prefrontal cortex and the amygdala during incongruent trials; the degree of connectivity was correlated with reaction time costs incurred during incongruent trials. The results are interpreted with reference to current models of emotional attention and regulation.     

Neural correlates of unconditioned response diminution during Pavlovian conditioning

Pavlovian conditioning research has shown that unconditioned responses (UCR) to aversive unconditioned stimuli (UCS) are reduced when a UCS is predictable. This effect is known as UCR diminution. In the present study, we examined UCR diminution in the functional magnetic resonance imaging (fMRI) signal by varying the rate at which a neutral conditioned stimulus (CS) was paired with an aversive UCS. UCR diminution was observed within several brain regions associated with fear learning, including the amygdala, anterior cingulate, auditory cortex, and dorsolateral prefrontal cortex when a CS continuously relative to intermittently predicted the UCS. In addition, an inverse relationship between UCS expectancy and UCR magnitude was observed within the amygdala, anterior cingulate, and dorsolateral prefrontal cortex, such that as UCS expectancy increased the UCR decreased. These findings demonstrate UCR diminution within the fMRI signal, and suggest that UCS expectancies modulate UCR magnitude.     

Neural substrates of smoking cue reactivity: a meta-analysis of fMRI studies

Reactivity to smoking-related cues may be an important factor that precipitates relapse in smokers who are trying to quit. The neurobiology of smoking cue reactivity has been investigated in several fMRI studies. We combined the results of these studies using activation likelihood estimation, a meta-analytic technique for fMRI data. Results of the meta-analysis indicated that smoking cues reliably evoke larger fMRI responses than neutral cues in the extended visual system, precuneus, posterior cingulate gyrus, anterior cingulate gyrus, dorsal and medial prefrontal cortex, insula, and dorsal striatum. Subtraction meta-analyses revealed that parts of the extended visual system and dorsal prefrontal cortex are more reliably responsive to smoking cues in deprived smokers than in non-deprived smokers, and that short-duration cues presented in event-related designs produce larger responses in the extended visual system than long-duration cues presented in blocked designs. The areas that were found to be responsive to smoking cues agree with theories of the neurobiology of cue reactivity, with two exceptions. First, there was a reliable cue reactivity effect in the precuneus, which is not typically considered a brain region important to addiction. Second, we found no significant effect in the nucleus accumbens, an area that plays a critical role in addiction, but this effect may have been due to technical difficulties associated with measuring fMRI data in that region. The results of this meta-analysis suggest that the extended visual system should receive more attention in future studies of smoking cue reactivity.     

Deficient modulation of pain by a positive emotional context in fibromyalgia patients

This study aimed to investigate the modulating effects of emotional context on pain perception in 16 patients with fibromyalgia syndrome (FMS) and 16 healthy control (HC) subjects. An infrared laser was used to apply individually adapted painful stimuli to the dorsum of the left hand. The emotional background of the painful stimuli was modulated by concurrent presentations of negative, neutral, and positive picture stimuli selected from the International Affective Picture System. As control conditions, painful stimuli and the pictures were also presented by themselves. During each of the 5 laser-picture trials, subjects received 10 painful stimuli and were asked to rate the average intensity and unpleasantness of the experienced pain. Functional magnetic resonance images were obtained, using a T2^∗ sensitive echo planar sequence. HC subjects showed a linear increase in pain intensity and unpleasantness ratings when painful stimuli were presented during positive, neutral, and negative pictures. In contrast, FMS patients showed a quadratic trend for pain intensity ratings indicating a lack of pain reduction by the positive pictures. In addition, the FMS patients showed less activation in secondary somatosensory cortex, insula, orbitofrontal cortex, and anterior cingulate cortex during the positive picture pain trials. Our results suggest that fibromyalgia patients are less efficient in modulating pain by positive affect and may benefit less from appetitive events than healthy control subjects.     

The anticipation of pain at a specific location of the body prioritizes tactile stimuli at that location

This study investigated whether one becomes more quickly aware of innocuous somatosensory signals at locations of the body where pain is anticipated. Undergraduate students (N = 20) indicated which of 2 stimuli that were administered to each hand using a range of stimulus onset asynchronies (SOAs), was presented first. Participants were instructed that the color of a cue (1 of 2 colors) signaled the possible occurrence of pain on 1 hand (threat trials). The other color of the cue signaled that no pain would follow (control trials). Results showed that during threat trials tactile stimuli on the hand where pain was expected, were perceived earlier in time than stimuli on the “neutral” hand. These findings demonstrate that the anticipation of pain at a particular location of the body resulted in the prioritization in time of somatosensory sensations at that location, indicating biased attention towards the threatened body part. The value of this study for investigating hypervigilance for somatosensory signals in clinical populations such as patients with chronic lower back pain is discussed.     

Evidence for anterior cingulate cortex involvement in monitoring preparatory attentional set

An important cognitive function underlying unified, voluntary behavior is attentional control. Two frontal regions, anterior cingulate cortex (ACC) and dorsolateral prefrontal cortex (DLPFC), appear to be particularly involved in attentional control and monitoring. In this study, we investigated whether ACC is involved in monitoring the preparatory allocation of attention during task switching, or whether ACC is active only when subjects are processing target stimuli and selecting a response, via a cued-attention design. Event-related BOLD fMRI activity was examined using a cue-target paradigm in which subjects performed task switches that selectively required reallocation of attention when tasks changed. There were three cue conditions: informative switch, informative repeat, and neutral. There were four target conditions: informatively cued switch, informatively cued repeat, neutrally cued switch, and neutrally cued repeat. Significant ACC activity was observed following both informative switch and informative repeat cues, but not after neutral cues. No significant ACC activity was observed following any of the target conditions. Significant DLPFC activity was observed following all three cue conditions and following neutrally cued switch targets. Overall, our results suggest that ACC is involved in monitoring the preparatory allocation of attention for conflict at the level of activation of competing attentional sets. The results also support the role of DLPFC in holding cognitive goals in working memory and allocating attention to the appropriate processing systems to meet those goals.     

Separating the effects of alcohol and expectancy on brain activation: an fMRI working memory study

The aim of this study was to use BOLD fMRI to evaluate the effect of alcohol intoxication on neuronal activation, when controlling for expectancy. Behavioural studies have shown that both alcohol intoxication and expectancy affect cognition, mood and behaviour. However, previous neuroimaging studies have not separated the effects of alcohol intoxication from the possible confounding effects of expectancy. Forty-five healthy male participants participated in the study. A balanced placebo design with four groups was used together with a working memory paradigm. Half of the participants consumed a soft-drink before the MR scanning session (half of them were correctly informed about the content of their drink, and half were incorrectly informed that they consumed an alcoholic beverage), and the other half consumed an alcoholic beverage before the MR scanning session (half of them were correctly informed about the content of their drink, and half were incorrectly informed that they consumed a soft-drink). A blood alcohol concentration (BAC) of 0.08% was used as the alcohol intoxication level. The most conspicuous result was that alcohol intoxication decreased neuronal activation especially in the dorsal anterior cingulate cortex (dACC) and in prefrontal areas, while expectancy increased neuronal activation in the same areas. This study shows that alcohol intoxication and expectancy have opposite effects on neuronal activation. The present results could have implications for pharmacological imaging studies.     

Two electrophysiological stages of spatial orienting towards fearful faces: early temporo-parietal activation preceding gain control in extrastriate visual cortex

Visuo-spatial attention tends to be prioritized towards emotionally negative stimuli such as fearful faces, as opposed to neutral or positive stimuli. Using a covert orienting task, we previously showed that a lateral occipital P1 component, with extrastriate neural sources, was selectively enhanced to lateralized visual targets replacing a fearful face (fear-valid trial) than the same targets replacing a neutral face (fear-invalid trial), providing evidence for exogenous spatial orienting of attention towards threat cues. Here, we describe a new analysis of these data, using topographic evoked potentials mapping methods combined with a distributed source localization technique. We show that an early field topography (40-80 ms post-target onset) with a centro-parietal negativity and a left posterior parietal source distinguished fear-valid from fear-invalid trials, whereas a distinct activity with anterior cingulate sources was selectively evoked during fear-invalid trials. At the same latency, or later, no difference in field topography was found for valid compared to invalid trials with happy faces. The early parietal map preceded a modulation in amplitude of the field strength (approximately 130 ms), corresponding to the enhanced lateral occipital P1 during valid trials in the fear condition. Furthermore, this early topography at 40-80 ms was positively correlated with the subsequent amplitude modulation of P1 at 130-160 ms in the fear condition, suggesting a possible functional coupling between these two successive events. These data have important implications for models of spatial attention and interactions with emotion. They suggest two successive stages of neural activity during exogenous orienting of attention towards visual targets following fearful faces, including an early posterior parietal negativity, followed by gain control mechanisms enhancing visual responses in extrastriate occipital cortex.     

Investigating dose-dependent effects of placebo analgesia: a psychophysiological approach

Investigating dose-dependent effects of placebo analgesia (PA) in laboratory subjects undergoing pain testing, we evaluated 2 hypotheses: (1) greater expectancy for relief produces greater PA, and (2) cued expectancy for relief triggered by a predictive cue leads to more enhanced analgesia than does passive expectancy (no predictive cue). We used conditioning procedures in which 84 subjects experienced reduced stimulation intensity following the application of purported analgesic creams to the 2 experimental fingers, while the control finger received the same levels of stimulation as in the baseline block. The dose of placebos was manipulated by creating 2 levels of expectations for relief. The form of expectation (cued vs uncued) was also manipulated by a predictive cue specifying the next finger to be stimulated. Subjective reports and psychophysiological responses served as critical indicators for evaluating impacts of the placebo manipulation on subsequent pain processing. The dose-dependent PA was unambiguously demonstrated by the predicted ordering of the 3 fingers (ie, manipulated expectation levels) in terms of both response sensitivity and average response magnitude, in mixed-effects analysis of 3 outcome indicators (evoked potential, skin conductance response, pain report). Greater expectation for relief led to both (1) greater reductions in the average dependent variable slope (response sensitivity) as a function of stimulus intensity, and (2) greater reductions in average response magnitude. Unexpectedly, uncued expectation led to a slightly larger PA than did cued expectation. The study provided clear evidence that PA can occur in a “dose”-dependent manner, mediated by the levels of expectancy for pain relief.     

Fear-relevant outcomes modulate the neural correlates of probabilistic classification learning

Although much work has implicated the contributions of frontostriatal and medial temporal lobe (MTL) systems during probabilistic classification learning, the impact of emotion on these learning circuits is unknown. We used a modified version of the weather prediction task in which two participant groups were scanned with identical neutral cue cards probabilistically linked to either emotional (snake/spider) or neutral (mushroom/flower) outcomes. Owing to the differences in visual information shown as outcomes, analyses were restricted to the cue phase of the trials. Learning rates did not differ between the two groups, although the Emotional group was more likely to use complex strategies and to respond more slowly during initial learning. The Emotional group had reduced frontostriatal and MTL activation relative to the Neutral group, especially for participants who scored higher on snake/spider phobia questionnaires. Accurate performance was more tied to medial prefrontal activity in the Emotional group early in training, and to MTL activity in the Neutral group later in training. Trial-by-trial fluctuations in functional connectivity between the caudate and MTL were also reduced in the Emotional group compared to the Neutral group. Across groups, reaction time indexed a switch in learning systems, with faster trials mediated by the caudate and slower trials mediated by the MTL and frontal lobe. The extent to which the caudate was activated early in training predicted later performance improvements. These results reveal insights into how emotional outcomes modulate procedural learning systems, and the dynamics of MTL-striatal engagement across training trials.     

Impact of experimentally-induced expectancy on the analgesic efficacy of tramadol in chronic pain patients: a 2 x 2 factorial, randomized, placebo-controlled, double-blind trial

Variations in treatment effects between drug trials are usually attributed to different patient characteristics, variations in outcome assessment, and random error. We have previously hypothesized that part of the variation in treatment effects between drug trials might be caused by differences in nonspecific factors. In a randomized clinical trial, we aimed to investigate whether experimentally induced expectancy can modify the analgesic effect of tramadol relative to placebo in chronic pain patients. In a 2 x 2 factorial, randomized, placebo-controlled, double-blind trial, chronic pain patients attending a chronic pain outpatient clinic were randomized to receive a single oral dose of 50 mg tramadol or placebo, and they were further randomized to receive positive or neutral information, verbally expressed by the physician, regarding the expected analgesic effect of the drug. Pain intensity was measured using a 10 centimeter visual analogue scale at baseline, and 0.5, 1, 2, 4, 6, and 8 hours after baseline. The one-hour pain intensity difference, calculated as the sum of pain intensity differences between baseline and 0.5 and 1 hour, was taken as main outcome measure. The one-hour sum of pain intensity differences of 28 patients treated after positive expectation and randomized to tramadol was 1.4 cm, while in 27 patients randomized to placebo, it was 0.8 cm. This corresponds with an analgesic effect of tramadol relative to placebo of 0.6 cm (95% confidence interval [CI], -0.5 cm to 1.8 cm). The 28 patients in the neutral expectancy group who were randomized to tramadol reported a 1.4 cm decrease on the sum of pain intensity differences, while 28 patients in the placebo group reported a 0.9 cm decrease. This corresponds with an analgesic effect of tramadol relative to placebo of 0.5 cm (95% CI, -0.9 cm to 1.8 cm). The 0.1 cm difference (0.6 cm - 0.5 cm) in analgesic effect between positive and neutral expectancy group was not statistically significant (95% CI, -0.7 cm to 1.0 cm). This trial did not discern a significant difference in the analgesic effect of tramadol between a positive and neutral expectancy group. This means that the phenomenon either does not exist, or we had an inappropriate model to demonstrate it. Regardless, this study demonstrates the type of quality trial that should be done to find out which non-specific factors, such as information regarding the expected effect, can modify treatment effects.     

Neural systems for orienting attention to the location of threat signals: an event-related fMRI study

Attention may reflexively shift towards the location of perceived threats, but it is still unclear how these spatial biases recruit the distributed fronto-parietal cortical networks involved in other aspects of selective attention. We used event-related fMRI to determine how brain responses to a neutral visual target are influenced by the emotional expression of faces appearing at the same location during a covert orienting task. On each trial, two faces were briefly presented, one in each upper visual field (one neutral and one emotional, fearful or happy), followed by a unilateral target (a small horizontal or vertical bar) replacing one of the faces. Participants had to discriminate the target orientation, shown on the same (valid) or opposite (invalid) side as the emotional face. Trials with faces but no subsequent target (cue-only trials) were included to disentangle activation due to emotional cues from their effects on target detection. We found increased responses in bilateral temporo-parietal areas and right occipito-parietal cortex for fearful faces relative to happy faces, unrelated to the subsequent target and cueing validity. More critically, we found a selective modulation of intraparietal and orbitofrontal cortex for targets following an invalid fearful face, as well as an increased visual response in right lateral occipital cortex for targets following a valid fearful face. No such effects were observed with happy faces. These results demonstrate that fearful faces can act as exogenous cues by increasing sensory processing in extrastriate cortex for a subsequent target presented at the same location, but also produce a cost in disengaging towards another location by altering the response of IPS to invalidly cued targets. Neural mechanisms responsible for orienting attention towards emotional vs. non-emotional stimuli are thus partly shared in parietal and visual areas, but also partly distinct.     

Dissociable brain mechanisms for processing social exclusion and rule violation

Social exclusion inherently involves an element of expectancy violation, in that we expect other people to follow the unwritten rule to include us in social interactions. In this functional magnetic resonance imaging (fMRI) study, we employed a unique modification of an interactive virtual ball-tossing game called "Cyberball" (Williams et al., 2000) and a novel paradigm called "Cybershape," in which rules are broken in the absence of social exclusion, to dissociate brain regions that process social exclusion from rule violations more generally. Our Cyberball game employed an alternating block design and removed evoked responses to events when the participant was throwing the ball in inclusion to make this condition comparable to exclusion, where participants did not throw. With these modifications, we replicated prior findings of ventral anterior cingulate cortex (vACC), insula, and posterior cingulate cortex activity evoked by social exclusion relative to inclusion. We also identified exclusion-evoked activity in the hippocampi, left ventrolateral prefrontal cortex, and left middle temporal gyrus. Comparing social exclusion and rule violation revealed a functional dissociation in the active neural systems as well as differential functional connectivity with vACC. Some overlap was observed in regions differentially modulated by social exclusion and rule violation, including the vACC and lateral parietal cortex. These overlapping brain regions showed different activation during social exclusion compared to rule violation, each relative to fair play. Comparing activation patterns to social exclusion and rule violation allowed for the dissociation of brain regions involved in the experience of exclusion versus expectancy violation.     

Memory retrieval under the control of the prefrontal cortex

Memory retrieval is a process wherein a distributed neural network reactivates the brain's representation of past experiences. Sensory long-term memory is represented among a population of neurones in the modality-specific posterior association cortex. The coded representation of memory can be retrieved by interactions of hierarchically different cortical areas along bottom-up and top-down anatomical connections. We examined the function of the prefrontal cortex in memory retrieval by two different approaches. Firstly, a meta-analysis of brain imaging studies revealed that the prefrontal cortex is reliably activated by memory retrieval in humans. Secondly, in order to determine the causal relationship between the prefrontal activations and memory retrieval, we designed a new experimental paradigm using posterior-split-brain monkeys. Following section of the splenium of the corpus callosum and the anterior commissure, visual stimulus-stimulus association learning within one hemisphere did not transfer to the other. Nevertheless, when a visual cue was presented to one hemisphere, the prefrontal cortex could instruct the contralateral hemisphere to retrieve the correct stimulus specified by the cue. These findings suggest that the prefrontal cortex can regulate the recall of long-term memory in the absence of bottom-up sensory inputs.     

Differential effects of distraction during working memory on delay-period activity in the prefrontal cortex and the visual association cortex

Maintaining relevant information for later use is a critical aspect of working memory (WM). The lateral prefrontal cortex (PFC) and posterior sensory cortical areas appear to be important in supporting maintenance. However, the relative and unique contributions of these areas remain unclear. We have designed a WM paradigm with distraction to probe the contents of maintenance representations in these regions. During delayed recognition trials of faces, selective interference was evident behaviorally with face distraction leading to significantly worse performance than with scene distraction. Event-related fMRI of the human brain showed that maintenance activity in the lateral PFC, but not in visual association cortex (VAC), was selectively disrupted by face distraction. Additionally, the functional connectivity between the lateral PFC and the VAC was perturbed during these trials. We propose a hierarchical and distributed model of active maintenance in which the lateral PFC codes for abstracted mnemonic information, while sensory areas represent specific features of the memoranda. Furthermore, persistent coactivation between the PFC and sensory areas may be a mechanism by which information is actively maintained.