Matching emotional facial expressions produced task-related neural activity that is typical of facial affect processing in a group of MA-dependent participants in early abstinence, as well as a group of healthy comparison participants. Both groups exhibited activation in ventral temporal cortex, linked to face detection and recognition (Haxby et al., 1996
; Kanwisher et al., 1997
), amygdala, linked to emotional responsivity to threatening facial expressions (Adolphs et al., 1994
; Morris et al., 1996
), dorsomedial PFC, linked to the formation of mental representations about the internal states of others (Frith and Frith, 2001
; Ochsner et al., 2004
), and dorsolateral PFC and parietal cortex, which participate in cognitive and executive processes necessary for successful task performance (Kane and Engle, 2002
Despite these similarities in activation patterns among groups, the two groups differed in activation of a number of cortical regions linked to affective and social cognitive processing. These differences did not stem from mood disorders in the MA-dependent sample, as volunteers with current psychiatric diagnoses were excluded from the study, and MA-dependent participants had recovered from abstinence-related depression and fatigue. Instead, activation differences are likely to reflect differences in processing of socio-emotional information.
The ventrolateral prefrontal cortex (VLPFC), which showed more task-related activity in the healthy than in the MA-dependent participants, has consistently been implicated in affect processing (Phan et al., 2002
), in particular in relation to emotional facial expressions (Nakamura et al., 1999
; Iidaka et al., 2001
; Kesler-West et al., 2001
; Keightley et al., 2003
; Phillips et al., 2004
; Ishai et al., 2005
). VLPFC has been suggested to represent an endpoint for the distributed networks involved in emotion recognition, where the information contained in emotional faces can be integrated for further processing (Sprengelmeyer et al., 1998
), or to participate in the semantic processing of this information (Hornak et al., 1996
). The ability to effectively process or interpret socio-emotional information could therefore be compromised in MA-dependent individuals, owing to a functional deficit in this region.
Alternatively, VLPFC involvement has been associated with regulation of emotional -- particularly aggressive (Blair, 2005
) -- impulses, with right VLPFC (RVLPFC) thought to exert top-down inhibitory control over limbic structures (Hariri et al., 2000
; Eisenberger et al., 2003
). RVLPFC projects to limbic regions of the brain (Carmichael and Price, 1995
; Cavada et al., 2000
; Vogt and Pandya, 1987
), inhibiting amygdala reactivity during intentional (Ochsner et al., 2004
) and unintentional (Lieberman et al., 2007
) regulation of affect, as well as the anterior cingulate cortex during placebo analgesia (Lieberman et al., 2004
) and regulation of social threat (Eisenberger et al., 2003
). In addition, RVLPFC has been implicated in motor response inhibition (Ma et al., 2003
; Eagle et al., 2004
; Aron et al., 2004
), and regulation of impulsively aggressive behaviors (Bufkin and Luttrell, 2005
; Raine et al., 1998
; Pietrini et al., 2000
). Notably, chronic stimulant administration to animals, and stimulant abuse by humans, are associated with both deficits in inhibitory function (Jentsch et al., 2002
; Goldstein & Volkow, 2002
; Salo et al., 2005
; Monterosso et al., 2005
) and structural deficits in RVLPFC (Thompson et al., 2004
). Although the current study was not designed to detect an inhibitory contribution of RVLPFC, future studies could determine whether the difference in task-related RVLPFC activation found here reflects a regulatory deficit over limbic structures (amygdala and/or dACC).
Of particular interest is the potential regulation of dACC, which was more active in the MA-dependent than healthy participants in the present study. Dorsal ACC plays a role in pain and distress (Rainville et al., 1997
), including social distress (Eisenberger et al., 2003
). In a study of social rejection (Eisenberger et al., 2003
), healthy individuals who felt distressed by an exclusionary experience exhibited dACC activation; however, individuals who activated RVLPFC more during the social exclusion showed less activation in dACC and reported less social distress, suggesting that RVLPFC regulated the distress of socio-emotional threat. A later study (Eisenberger et al., 2007
) found relationships between dACC reactivity during social exclusion, self-reports of social sensitivity and trait aggression, and genetic factors previously linked to aggressive behavior, further suggesting that dACC activation reflects social hyper-reactivity that can lead to aggressive behavior. The present findings parallel these results, as the MA-dependent participants showed higher dACC activation than the healthy participants, and behavioral measures hinted at higher interpersonal sensitivity. In addition, those individuals in the MA-dependent group who gave particularly high self-reports of hostility and interpersonal sensitivity showed the highest dACC activation, and those who had high interpersonal sensitivity scores tended to have the lowest RVLPFC activity. In light of these preliminary self-report findings, higher dACC and lower RVLPFC activity in the MA-dependent participants could point to a disruption in the system implicated in socio-emotional regulation (Eisenberger et al., 2003
), indicating a potential mechanism underlying the mood disturbances and socially maladaptive behaviors often exhibited by MA-dependent individuals.
The remaining regions showing more task-related activity in the healthy participants than the MA-dependent participants – temporoparietal junction (TPJ), superior temporal sulcus (STS), and temporal pole – are often implicated in social cognition, particularly in understanding the motivations and emotional states of others (“theory of mind”) (Ochsner et al., 2004
; Gallagher and Frith, 2003
; Fletcher et al., 1995
; Brunet et al., 2000
). Theory of mind is thought to rely heavily on dorsomedial PFC (Frith and Frith, 2001
), which was in fact active in both groups. However, it also involves mentally representing the internal states of others, and constructing a coherent model about their beliefs, which has been shown to depend on the integrity of the TPJ (Saxe et al., 2004
; Vollm et al., 2006
; Saxe and Kanwisher, 2003
; Samson et al., 2004
; Saxe and Wechsler, 2005
). Our finding that MA-dependent participants activate these regions less than healthy participants suggests a functional impairment that could interrupt or impede the construction of such mental-state representations. The facial expression matching task is not commonly thought to engage theory of mind processing; however, these activation differences between groups could point to a spontaneous engagement of some social cognitive processes by the healthy participants that are impaired in the MA-dependent participants (although comparable dorsomedial PFC activations point to some theory of mind processing in both groups). This impairment in some aspects of social cognitive processing could, in turn, lead MA-dependent individuals to respond inappropriately to what others are expressing.
Contrary to expectation, we did not find reliable between-groups differences in amygdala activation during facial affect matching, despite previous findings that MA-dependent participants have abnormal glucose metabolism in this region (London et al., 2004
). Although amygdala activation tended to be lower in the MA-dependent participants than the healthy participants during Emotion Match trials, this difference was not significant. It is possible that adequate amygdala responsivity to threatening facial expressions is preserved in MA-dependent individuals, consistent with the finding that some, but not all, functions of the amygdala are impaired in reactive aggression with psychopathy (Blair, 2005
), and that it is perhaps amygdala downregulation, rather than activation, that is impacted in MA-dependent individuals. Contributing factors may also have included poor signal-to-noise ratio, as signal dropout can be a problem in acquisition of amygdala fMRI images, small sample size, or habituation to repeated and/or prolonged presentation of faces within each block (Wright et al., 2001
; Fischer et al., 2003
In sum, although the capacity to activate the amygdala in response to negative facial expressions appears equivalent in individuals who abuse MA and individuals who do not, differences exist in a set of cortical regions necessary for experiencing social threat (dACC), integrating and/or regulating emotional information (RVLPFC), and more general social cognitive functions (temporal regions and TPJ).
Behavioral performance differences between Emotion Match and Shape Match trials suggest that the two types of trials differed in difficulty, so that active regions in the Emotion Match > Shape Match contrast could reflect higher attentional or processing demands, independent of the effects of emotion. However, the results are highly consistent with the imaging literature focusing on facial affect. In addition, the contrasts of interest were between the two groups of participants, and as no effect of group, or group X trial type interaction were found in the behavioral data, potential differences in difficulty between the two trial types are expected to subtract out.
An alternative interpretation of the group differences in task-related brain activity could be simple strategy differences in performing the task. However, equivalent task accuracy, along with biochemical differences between groups in relevant regions, point to functional deficits in the MA-dependent group, rather than a voluntary strategy difference. An additional alternative interpretation is based on the fact that, as with most fMRI data, imaging data were analyzed using subtractive logic (statistical maps for Shape Match trials were subtracted from statistical maps for Emotion Match trials). Thus, observed differences are open to two interpretations, as group differences attributed to processing of faces could have in fact been due to differences in processing of shapes. It is, however, not likely to be the case, as, for example, there is no evidence to suggest that matching shapes would induce more limbic activity than processing emotional information, for which the limbic system is specialized. Similarly, greater activation in one group than the other could also reflect less de-activation. Moreover, interpretation of functionality of a region on the basis of BOLD signal change is inferential, so that a number of task-related (or unrelated) functions could have been supported by activation of a given region. For example, dACC is often thought of as important in cognitive control, and an emotion regulation region, and the link to socio-emotional threat inferred through previous findings (e.g., Eisenberger et al., 2003
). Future research will need to address direct links between regions and their function in socio-emotional regulation by testing specific hypotheses in specific regions. Another potential source of group differences in brain activity is the psychiatric history of the MA-dependent sample, as some mood disorders have been associated with altered brain structure and function (e. g., Leppanen et al., 2004
). While none of the participants had a current psychiatric diagnosis, three of the MA-dependent participants had histories of depressive episodes, and one had a history of anxiety disorder. These histories did not appear to impact the results of the present sample, however, as removal of the participants from the whole-brain random effects analysis did not qualitatively change the results. Finally, the relatively small number of Emotion Match trials, relatively limited sample size of 12 per group, and relatively liberal threshold of p < .005 suggest an increased possibility for Type I and Type II errors to occur. While use of a blocked design and inclusion of a cluster-size threshold of 10 contiguous voxels (which helps protect against false positives (Forman et al., 1995
) and has been used in previous studies (Eisenberger et al., 2003
; Lieberman et al., 2004
)) can partly alleviate these concerns, the limitations should be borne in mind when considering the present results.
Despite these caveats, the present study identifies functional differences between MA-dependent and healthy comparison participants in brain regions linked to emotion integration and/or cognitive control, social threat perception, and social cognition, and is the first to do so in the context of responsivity to facial expressions. Such deficits may lead to inappropriate socio-emotional behaviors associated with MA use, potentially contributing to states of stress and increased risk of relapse.