Are Anticorrelated Networks in the Brain Relevant to Schizophrenia?

Abstract

Most models of schizophrenia are based on basal ganglia-thalamocortical (BGTC) neuronal circuits or brain structures that project to them. Two new neuronal networks have been described which include many of the brain regions associated with BGTC neuronal circuits. These networks have been characterized with a new brain-imaging technique based on low-frequency fluctuations of the blood oxygen level–dependent (BOLD) signal. The new network associated with attention-demanding tasks is referred to as the task-related network and the network associated with stimulus-independent thought during the resting state is referred to as the default network. The 2 networks have been proposed to be negatively correlated or anticorrelated. This article critically reviews the rationale for these anticorrelated networks, the technique with which they are characterized, and preliminary findings in schizophrenia and other neuropsychiatric disorders. Regions associated with the default network overlap with regions important in motivation and are activated by memory retrieval, auditory hallucinations, and ketamine. Task-related networks are necessary for performance of neurocognitive tasks on which schizophrenic patients often perform poorly. It is concluded that anticorrelated networks can be viewed as complementary ways of understanding self-monitoring and task performance which extend present models of schizophrenia based on BGTC circuits. However, there are some limitations with regard the present understanding of brain structures involved in self-monitoring and the lack of asymmetry in the network which may mediate stimulus-independent thought. Further investigations of the default network assessed by low-frequency fluctuations in the BOLD signal seem warranted.

There is an emerging consensus that schizophrenia is a neurodevelopmental disorder caused by a number of genetic, hypoxic, metabolic, and/or viral factors.^1–4 However, all may affect one or more neuronal circuits. Brain structures implicated include the prefrontal cortex, anterior cingulate, thalamus, basal ganglia, superior temporal gyrus, hippocampus, cerebellum, and the neuronal circuits connecting these structures involving a number of neurotransmitters such as dopamine, glutamate, and γ-aminobutyric acid (GABA).^5–13 All these brain structures are part of or connect with basal ganglia-thalamocortical (BGTC) neuronal circuits.

Some new neuronal networks have recently been characterized in the human brain. These networks include regions associated with BGTC neuronal circuits. A network of brain regions active at rest, referred to as the default network, has been proposed. The network includes brain regions such as the posterior medial cortices, the posterior lateral cortices, and the ventral and dorsal medial prefrontal cortices.¹⁴ These regions are suggested to be associated with internally generated “stimulus-independent” thought as well as self-monitoring and salience monitoring.¹⁴ The default network is deactivated during attention-demanding tasks which are associated with a task-related network of activity in the frontal and parietal cortices. Thus, it has been suggested that the activity in these 2 functional networks are negatively correlated or anticorrelated.¹⁵

All regions associated with anticorrelated networks have been implicated in schizophrenia. Some of the functions of the default network may also be related to deficits in self-monitoring seen in schizophrenic patients.¹¹ Consequently, it seems reasonable to ask whether these networks might be relevant to the pathophysiology of schizophrenia? This article will describe the nature of anticorrelated networks and a new brain-imaging technique based on spontaneous low-frequency fluctuations of the blood oxygen level–dependent (BOLD) signal with which these networks have been defined. Preliminary evidence of anomalies in these networks in schizophrenia and implications for understanding the final common pathway of schizophrenia will be discussed.

Anticorrelated Networks

No region of the brain functions in isolation. During the performance of attention-demanding tasks, frontal and parietal cortical regions are usually activated in subjects, while posterior cingulate, medial prefrontal, and medial and lateral parietal regions are less active.^16–22 Correlations have been demonstrated between regions that are less active during attention-demanding tasks with conventional functional magnetic resonance imaging (MRI). Geicius et al²³ examined functional connectivity of various brain regions at rest. In this study, the posterior cingulate was strongly coupled with the ventral anterior cingulate and several other brain regions implicated in the default network. However, there are some limitations associated with quantification of the resting state in humans with the correlational analysis of conventional functional MRI. Conventional functional MRI is typically a subtraction technique used to assess differences, which are less dependent on absolute values, between 2 or more states or tasks within an individual which can then be compared between groups.

Another approach to brain connectivity at rest is the examination of low-frequency (<0.1 Hz) spontaneous fluctuations in the BOLD signal over time.²⁴ Slow spontaneous fluctuations in the BOLD signal in a specific “seed” region, deactivated during attention-demanding tasks, can be correlated with other regions in the brain. Two independent groups have confirmed a resting network of slow spontaneous BOLD signal activity using this approach.^15,25 This default network includes the anterior and posterior cingulate, the medial prefrontal cortex, and the inferior temporal and cerebellar regions. Anticorrelated with the default network is an attention-demanding, task-related network including the dorsolateral prefrontal cortex, supplemental motor area, inferior parietal lobule, and middle temporal region. These regions are illustrated in figures 1 and 2.

Fig. 1.

Intrinsically defined anticorrelated processing networks in the brain. Positive nodes are significantly correlated with seed regions involved in focused attention and working memory (task-positive seeds) and significantly anticorrelated with seed regions routinely deactivated during attention-demanding cognitive tasks (task-negative seeds). Negative nodes are significantly correlated with task-negative seed regions and significantly anticorrelated with task-positive seed regions. (Left) Lateral and medial views of left hemisphere. (Center) Dorsal view. (Right) Lateral and medial views of right hemisphere. Reprinted with permission from Fox et al¹⁵ Copyright 2005, National Academy of Sciences, USA.

Fig. 2.

A simplified diagram showing the principal regions associated with the default network in gray and the task-related network (unshaded) from a top-down, transverse perspective. DLPC, dorsolateral prefrontal cortex; SMA, supplementary motor area, frontal eye field; MTC, middle temporal region, insula; IPL, inferior parietal lobule; LPC, lateral parietal cortex; CB, cerebellar tonsils; PC, posterior cingulate, precuneous, retropsenial cortex; ITR, inferior temporal region; MPFC, medial prefrontal cortex.

Physiological Underpinnings of Low-Frequency Spontaneous BOLD Signal Fluctuations

Biswal et al²⁴ were the first to demonstrate correlations in low-frequency (<0.1 Hz) spontaneous fluctuations in BOLD signal in the somatosensory system. Other investigators have observed coherent slow-activity networks in the visual, auditory, and language-processing systems.^26–28 It has been suggested that lower frequency fluctuations may be related to electrophysiological fluctuations in the gamma band implicated in “binding.”^29,30 However, neural synchrony likely occurs across multiple frequency bands and brain regions involved in a particular perception or thought.³¹

Traditionally resting wakefulness has been associated with alpha activity which is replaced with beta and gamma activity during orienting and cognitive tasks. One would expect that a similar default network could be defined with this frequency band. An attempt to plot alpha activity during rest showed a positive correlation with the anterior cingulate and occipital cortex and a negative correlation with frontoparietal regions during rest which corresponded to some degree with functional MRI signal changes.³² However, a follow-up investigation failed to map alpha changes onto the default network as clearly.³³

Are Anticorrelated Networks Human Phenomena?

Stimulus-independent thought is not often attributed to rodents. However, a rodent has to be aware of its internal state and the likelihood of obtaining food. It is also likely that the anterior cingulate participates in a primitive vocalization network.³⁴ It is not known whether anticorrelated networks are seen only in humans. It seems likely that there are some commonalities with nonhuman primates. For example, posterior medial regions including parietal and posterior cingulate regions seem to be associated with monitoring the environment for signs of danger,¹⁴ while the dorsolateral prefrontal cortex is critical for performance of delayed response tasks in primates.³⁵ It is much more difficult to establish whether the default regions are associated with other types of mental activity in primates but low-frequency fluctuations on functional MRI have been observed in the visual cortex in monkeys during various behavioral states.³⁶

One of the interesting features of the default network is the connection between the anterior and posterior cingulate and inferior temporal regions. If there is something different about language processing in humans during stimulus-independent thought, one would expect a connection with regions in the inferior temporal cortex that are interconnected with long-term memory structures like the hippocampus. Nonhuman primates seem to process short-term representations of auditory information but do not seem to utilize brain structures associated with long-term memories of auditory information.³⁷

There are some interspecies differences in intracingulate connections which may suggest that there are some unique aspects of human and primate brain. Rodents have a connection between anterior area 24 and posterior area 29, whereas primates have a connection between anterior area 24 and posterior area 23 which in turn has more complicated input from posterior areas 29 and 30.³⁸ Posterior retrospenial cingulate regions receive input from the subiculum and project to the anterior thalamus providing an alternative route between the hippocampus and the thalamus. Both the anterior thalamic nuclei and the retrospenial cortex are thought to play a pivotal role in long-term memory. The anterior cingulate and posterior cingulate cortices activate during auditory-verbal memory function tasks in humans.³⁹ There is also some evidence that lesions to the posterior cingulate lead to episodic memory loss in humans but not primates, although some studies are open to other interpretations.⁴⁰

Is the Default Network Really Associated With Stimulus-Independent Thought?

The default network has been demonstrated by a number of independent investigators using both seed regions and statistical parametric mapping of low-frequency activity which does not depend on prior assumptions about which regions might be involved.^15,23,25,41 Considering these different approaches, it is remarkable that the medial prefrontal, posterior cingulate, and temporal lobe structures show consistent bilateral connectivity during rest. However, some investigations have found that the network extends to other structures such as the thalamus and insula,²⁵ visual processing regions,⁴¹ and angular and parahippocampal gyrus on the left side.²³

One of the questions which may be asked is whether these regions are connected as a consequence of some low-frequency artifacts such as autonomically driven fluctuations in cerebral blood flow? These fluctuations can be modeled and do not seem to account for the default network.⁴¹ A further question, which has been asked, is whether there is anything special about the resting state? ⁴² The default network could simply be the brain's response to sitting in an MRI machine which is no different from any other task. Fox et al¹⁵ suggest that this is not likely the case as findings are reproducible across different states: eye fixation, eyes close, and eyes open and even during anesthesia. Thus, there is reason to believe that this network is not an artifact or nonspecific response to scanning but does the network reflect stimulus-independent thought?

Gusnard and Raichle¹⁴ argued that the medial prefrontal cortex is associated with self-generated thoughts, intended speech, and emotions. There is now some direct evidence that mind wandering is associated with activity in the default network.⁴³ Further support comes from lesion studies. Damage to the medial prefrontal cortex and anterior cingulate in humans is associated with the description of an “empty mind” with “nothing to say.”⁴⁴ Brain structures involved in distinguishing self from nonself fairly reliably activate the medial prefrontal cortex and posterior cingulate cortex on one or both sides.^45–49 However, it is also likely that other brain structures such as the right parietal cortex, insula, and cerebellar parietal network are important in self-representation.^50–52 Some of these structures are part of the task-related network.

Fransson²⁵ pointed out that the medial prefrontal cortex together with the posterior cingulate cortex, angular gyrus, and temporal lobe structures sustain a memory-retrieval system for autobiographical events.^53,54 This suggestion is consistent with a role for these structures in stimulus-independent thought. It is also consistent with the suggestion that the default mode may be essential for manipulation of past events in order to consider a course of action.⁵⁵ However, part of this planning must involve language. The medial prefrontal cortex and cingulate cortices have not typically been associated with language, but there is no doubt that the anterior cingulate is closely connected with speech areas and may regulate their activity during silence.^56,57 Consequently, there is reason to believe that this network is involved with stimulus-independent thought, but there are likely other associated networks.

Relationship to Attention

Networks which may be closely related to both the task-related and default networks are those involved in attention. Behavioral, neuroimaging, lesion, and electrophysiological studies have suggested 2 separate frontoparietal systems: the dorsal system consisting of the intraparietal sulcus and frontal eye field in each hemisphere and a ventral attentional system which is right lateralized and comprised of the right temporal-parietal junction and the right ventral prefrontal cortex.¹⁶ The ventral system likely responds to salient targets and is able to over-ride the dorsal system. These systems have recently been mapped by Fox et al⁵⁸ with the connectivity of low-frequency oscillations at rest similar to the methods used to map the task-related and default networks. Many of the regions in the dorsal and ventral attentional network overlap the task-related networks in prefrontal, premotor, and parietal regions raising the possibility that spontaneous activity may be necessary for prediction of regions which may be used together in the future.⁵⁸

Are the Networks Anticorrelated?

One of the problems with the idea of anticorrelated networks is that tasks often require activation of regions involved in both the default and task-related networks for completion. For example, Morcom and Fletcher⁴² point out that episodic memory tasks actually lead to an increase rather than decrease in activity in the medial prefrontal cortex and precuneus. There does not seem to be a good explanation for how this could occur if the regions are anticorrelated. These regions are likely very complex, and it is an oversimplification to suggest that a region such as the medial prefrontal cortex has a single function or activates as a block. This could provide a partial explanation but it appears that the default region is not localized but involves a fairly large proportion of these structures. Another partial explanation may be that the default network is not completely suppressed but rather attenuated during the performance of some tasks.^59,60

How Are the Networks Anticorrelated?

How the task-related network suppresses the default network is not known. It has been suggested that the prefrontal regions may suppress the default network^23,61 or alternatively another region such as the thalamus may influence cortical-cortical interactions.^23,62 It is of interest that a unique neuronal cell type has been described in humans and higher primates—the spindle or Von Economo neuron.⁶³ The distribution of this neuron type curiously superimposes onto both the default regions of the anterior cingulate and medial prefrontal cortex and the regions involved in memory retrieval. This type of neuron is much more numerous in humans, particularly on the right side. Von Economo neurons can also be found in the insula and the claustrum which surrounds the basal ganglia (J. Allman, personal communication). Von Economo neurons in the claustrum could influence wide areas of the heteromodal cortex. It is possible that these neurons may have a role with interactions between the 2 networks.

Preliminary Investigations of the Default Network in Schizophrenia

Liang et al⁶⁴ examined 15 chronic, medicated patients and 15 healthy volunteers with conventional functional MRI in the resting state. This study did not examine spontaneous low-frequency fluctuations in the BOLD signal but did compare correlation coefficients of each pair of 116 brain regions between groups. The global distribution of abnormal functional connectivities revealed generally decreased connectivity in schizophrenic patients involving all 9 brain regions studied. However, the insula, temporal lobe, corpus stiatum, and prefrontal lobe deficits were more striking. Increased functional connectivities were mainly related to the cerebellum in patients. Many of the regions in which decreased connectivity was found are related to the default network, but this study suggests that connectivity deficits are much more widespread.

More compelling evidence of the involvement of the default network in schizophrenia comes from Garrity et al⁶⁵ who examined 21 mostly medicated, chronic schizophrenic patients and 22 healthy volunteers. This study was not done in the resting state. Conventional functional magnetic resonance data was obtained during an auditory oddball task. Independent component analysis was used to identify the default network during the oddball task which was deemed to be not highly demanding of attention. Differences in the spatial aspects of the default network were compared between groups. Temporal aspects of low-frequency oscillations in the BOLD signal were also examined globally in the default network although not in the same way as previous studies in normal subjects which had looked at the connectivity of these signals.^15,24 Significant spatial differences between groups were found in the frontal, anterior cingulate and parahippocampal regions, all regions associated with the default network. Patients also showed significantly higher frequency fluctuations in the temporal evolution of the default mode. Positive symptom scores were positively correlated with increased deactivation in the medial frontal gyrus, precuneus, and left middle temporal gyrus. The authors hypothesized that the default network may be under- or overmodulated by regions including the anterior and posterior cingulate.

A further study, submitted without knowledge of the Garrity et al⁶⁵ study, examined slow fluctuations of the BOLD signal in 17 chronic, medicated schizophrenic patients and 17 healthy volunteers.⁶⁶ Methodology was based on the Fransson²⁵ study of healthy volunteers which correlated low-frequency fluctuations in the BOLD signal between the posterior cingulate and other brain regions. Schizophrenic patients had striking deficits in regions associated with the default network including the posterior cingulate, lateral parietal, medial prefrontal, and cerebellar regions. A positive symptom (Scale for the Assessment of Positive Symptoms Total Score)–dependent correlation was found between the posterior cingulate and several regions including the auditory cortex in patients.

Implications for Pathophysiology

While the latter 2 studies are limited by the examination of chronic, medicated patients, they do offer some preliminary evidence of anomalies of the default network in schizophrenic patients. It is interesting that both found deficits in the anterior portions of the default network. In the Garrity et al⁶⁵ study, increased deactivation of medial frontal and left middle temporal structures was found to be in association with positive symptoms. Using a different seed region approach, Bluhm et al⁶⁶ found positive symptom–dependent correlations between the posterior cingulate and auditory cortex. These findings may be related. If the default network in schizophrenic patients is associated with decreased communication between the medial prefrontal and posterior cingulate regions, it is possible that self-monitoring systems may become split leading to the perception that auditory thoughts are externally produced.

Specificity

It is not entirely clear whether anomalies in the default network are specific to schizophrenia. Autistic patients have recently been shown to fail to deactivate the medial prefrontal cortex, rostral anterior cingulate, posterior cingulate, and precuneous during cognitive tasks.⁶⁷ Mostly chronic, medicated depressed patients, with and without psychosis, demonstrated significantly greater resting-state subgenual cingulate and thalamic functional connectivity with the default-mode network.⁶⁸ Independendent component analysis was used to isolate the default network in patients with mild Alzheimer disease and healthy volunteers.⁶⁹ Compared with the volunteers, Alzheimer patients showed decreased resting-state activity in the posterior cingulate and hippocampus, suggesting abnormal connectivity between these structures. Thus, anomalies in the default network may not be specific to schizophrenia. However, it is of note that anomalies also seem to involve structures which would be expected to be involved in these disorders such as the subgenual prefrontal cortex in depression and the hippocampus in Alzheimer disease.

Are Anomalies in Anticorrelated Networks Consistent With What We Know About Schizophrenia?

All models of the final common pathway of schizophrenia include structures that are part of BGTC neuronal circuits or the structures which project to them including those in the temporal lobe and cerebellum.^{8–10,13,70–76} In a model proposed by O'Donnell and Grace,⁷⁰ input from the hippocampus and possibly the amygdala to the nucleus accumbens lead to faulty regulation of BGTC circuits, while in a model proposed by Lieberman and associates⁷¹ faulty regulation is related to dopamine sensitization unregulated by cortical inhibition. These models have some support from in vivo positron emission studies showing abnormal presynaptic dopamine activity and regulation in the ventral striatum in schizophrenic patients.¹ Both task-related and default networks are closely related to BGTC circuits.

Relationship of BGTC Circuits to Anticorrelated Networks

Prefrontal BGTC neuronal circuits, which are likely more task-related, originate in the dorsolateral and lateral orbitofrontal regions projecting via glutamatergic neurons to dorsal basal ganglia where they project directly and indirectly to the thalamus and then back to the prefrontal cortex. A “limbic” BGTC neuronal circuit, which is likely closely related to the default network, begins in the medial orbital prefrontal cortex and anterior cingulate and projects to the nucleus accumbens in the ventral striatum which in turn projects to the thalamus and back to the prefrontal cortex and anterior cingulate.⁷⁷ The limbic circuit receives input from the amygdala and hippocampus at the level of the nucleus accumbens. These inputs regulate dopamine projections to the nucleus accumbens which, in turn, regulates cortical activity.⁷⁸

BGTC neuronal circuits are reputed to be important in learning new behaviors and integrating affect into behavior in both humans and nonhuman primates.^77,79 However, very little is known about how the prefrontal circuits interact with the limbic circuit. Anticorrelated networks provide a potential bridge between these networks. Abnormalities in dopamine regulation of the limbic BGTC neuronal circuit could lead to dysfunctional of the anterior cingulate and medial prefrontal cortex, structures that overlap with the default network, possibly leading to errors in self-monitoring. Errors in self-monitoring could result from a disruption of the default network and/or an inability of these regions to anticorrelate with the task-related network. Structures involved in the default network and BGTC neuronal circuits have close connections with the thalamus suggesting a possible point of convergence. It is interesting that this structure has been proposed to have a central role in consciousness.⁸⁰

NMDA Hypofunction

Another influential model of schizophrenia has been the N-methyl-D-aspartate (NMDA) hypofunction hypothesis arising in part from the effects of phencyclidine administration in rodents. These animals demonstrated progressive neurodegenerative changes throughout the limbic system beginning in the posterior cingulate.⁸¹ Repeated exposure to NMDA antagonists causes subcortical dopaminergic hyperresponsivity in rats⁸² similar to ventral hippocampal lesioned rats.⁸ The model is particularly appealing because a deficit in GABAergic neurons could result in a pattern of limbic overactivity similar to the effects of NMDA antagonists because NMDA potentiates activity in GABAergic neurons.⁸¹ Both the posterior and anterior cingulate are part of the default network, but it is curious that ketamine, which has been shown to cause both positive-like and negative-like symptoms in healthy volunteers and schizophrenic patients, causes an increase in anterior rather than posterior cingulate activity, suggesting that humans might be wired differently.⁸³

Neurocognitive Findings

While it is beyond the scope of this review to discuss neurocognitive findings in schizophrenia, it is of note that schizophrenic patients have trouble with almost all cognitive tasks involving all the regions of the task-related network which to some extent correspond to prefrontal activation deficits in brain-imaging studies.¹ There is also a rich electrophysiological and neuropsychological literature suggesting problems with attention and an inability to suppress irrelevant material which may involve default network regions.¹ However, considering the heterogeneity and ubiquity of findings in schizophrenia, this may not be a very compelling argument. There are few regions in the brain that are not part of the task-activated and default networks, and almost all these have been found to function abnormally in some way in schizophrenia.

Auditory Hallucinations

A key feature of schizophrenia is auditory hallucinations. While the default network can be linked to stimulus-independent thought and self-representation, the connection with language-processing warrants further comment. The anterior cingulate and medial prefrontal cortices are closely connected with the superior temporal gyrus.⁵⁶ Recent in vivo evidence suggests a functional connection between these structures. A functional MRI study has shown that, in healthy subjects, the auditory cortex spontaneously activates during silence and that this activation is modulated by the anterior cingulate cortex.⁵⁷ Such activity likely reflects internal speech in the absence of external stimuli, a phenomenon possibly related to the experience of auditory hallucinations. Patients experiencing hallucinations in real time have demonstrated activation in the anterior cingulate and other regions associated with the default network including the orbitofrontal and temporolimbic regions.^84–86 The relationship between the default network and ratings of positive symptoms found in preliminary studies provides more direct evidence of the involvement of the default network in schizophrenia.^65,66

Negative Symptoms

The anterior cingulate and medial prefrontal cortex are involved in affective processing.^34,87 In fact, damage to this structure in humans leads to symptoms that look a lot like the amotivation and flat affect seen in patients with schizophrenia.⁴⁴ However, negative symptoms have also been linked to regions associated with task-activated networks such as the prefrontal cortex, thalamus, and parietal regions.¹ Nevertheless, the fact that default regions can be linked to both negative and positive symptoms is potentially important as schizophrenia is characterized by both, and it is likely that both have a related substrate in the brain. Otherwise, patients would present with only auditory hallucinations or only negative symptoms which is not the case.

Forward Model Mechanisms

Many have suggested that there may be abnormalities in the brain mechanisms which predict the outcome of self-initiated actions.^13,88 These predictions involve interactions between the frontal cortex and posterior sensory processing regions of the brain. Many of these same regions have been implicated in the task-related and attentional networks. Raichle and Gusnard⁵⁵ propose that intrinsic activity sets the stage for expression of motivated behavior. However, there is considerable debate as to the nature of prediction abnormalities in schizophrenia. Some suggest a problem with the “who” system, while another suggests that there may be an exaggerated sense of agency.^89,90 Nevertheless, both suggestions could involve components of task-related and default networks.

What Does Not Fit With What We Know About Schizophrenia?

While there are a number of appealing aspects to thinking about schizophrenia as being related to anticorrelated networks, there are also some findings which do not easily fit into this approach to brain connectivity and function.

Self-monitoring

The anterior cingulate and medial prefrontal cortices are activated in conflict monitoring and “theory of mind” tasks^91,92 which is consistent with a self-monitoring function of these regions in the default network. However, the involvement of these brain structures during these tasks has recently been questioned. Patients who have lesions of the anterior cingulate and medial prefrontal cortex can not only carry out theory of mind tasks very well but also perform tasks requiring high levels of attention and error detection such as the Stroop.^93–95 Consequently, it is not clear what these regions actually do but most patients with lesions to the anterior cingulate, medial prefrontal cortex, and the premotor areas have difficulties with affect, both with regard to motivation and assessing probability of positive outcomes.⁹⁴

Some time ago, Frith⁹⁶ suggested that defective self-monitoring in the prefrontal cortex may lead to hallucinations, but there has been little work on the physiological interactions between the anterior cingulate and dorsolateral prefrontal cortex. There is some evidence of decreased physiological interactions during word fluency tasks between the dorsolateral prefrontal cortex and anterior cingulate cortex in schizophrenic patients compared with healthy volunteers.⁹⁷ However, another word fluency investigation in never-treated schizophrenic patients did not find abnormal connectivity between the right anterior cingulate and dorsolateral prefrontal cortex but did find deficit connectivity between the right anterior cingulate and left inferior temporal region.⁹⁸

Inner speech in patients prone to hallucinations has not been found to be associated with abnormal dorsolateral prefrontal or anterior cingulate activity.⁹⁹ However, this study examined activation during the silent repetition of the word “rest” which differs from paradigms used to assess the default network. There is also little evidence of impairment of the phonological loop in patients who hallucinate.¹⁰⁰ These findings are not necessarily unexpected because patients are able to perceive and respond to speech normally most of the time. It is only when they are hallucinating that there is a problem. Unfortunately, none of the studies of patients experiencing auditory hallucinations at the time of the scan have done correlation analyses between brain regions which might be involved in either the task-related or default networks.

Asymmetry

Activation in the default and task-related networks tends to be bilateral. Many would argue that the final common pathway of schizophrenia has to be asymmetrical. Gruzelier⁷³ proposed that symptoms reflecting relatively transient excitation were associated with right hemisphere dysfunction, whereas more persistent withdrawal symptoms were associated with underactivity of the left hemisphere. Crow¹⁰¹ extended this idea to speech processing, suggesting that acoustically coded information is retained briefly in the dominant hemisphere so that a visuospatial sketch pad can be activated in the nondominant hemisphere allowing executive function to determine the direction of thought. These models fail to account for the role of the anterior cingulate and medial prefrontal cortex in schizophrenic symptomatology. However, it is likely that the final common pathway of schizophrenia has some lateralized aspects.

The default network is bilateral in normal subjects. Preliminary data in schizophrenic subjects suggest that there are bilateral deficits in the default network.^65,66 However, correlations with positive symptoms are lateralized. There are some interesting asymmetries with regard to the ventral attentional network, suggesting that it may merit further consideration in the pathophysiology of schizophrenia. In fact, it is almost the mirror image of the language-processing regions in the left hemisphere which include Broca and Wernicke areas.⁵⁸ The nature of the default network in schizophrenia and its relationship to attentional networks has not as yet been examined.

Conclusions

Much more is now known about cortical networks involved in self-monitoring and auditory processing in the human brain. The key question is how are the structures which mediate auditory processing, self-monitoring, and motivation linked? A reasonable case can be made that the default network links at least some of these functions. There is some overlap between default and task-related networks and BGTC neuronal circuits implicated in the pathophysiology of schizophrenia by most models. They are not necessarily exclusive and the inclusion of wider networks such as the default network may make it possible to translate dopaminergic anomalies in these circuits into the experiential aspects of schizophrenia. There are also some limitations. The structures in the brain related to self-monitoring are by no means clear, and default networks do not easily lend themselves to asymmetrical processes.

There are some promising data directly implicating anomalies in default networks in schizophrenia.^65,66 However, this work should be considered preliminary. Connectivity deficits appear to be fairly widespread in schizophrenia. Deficits in the connectivity of structures which are associated with the default network may just be part of a wider connectivity deficit. Nevertheless, it is interesting that there appears to be a relationship between the default network and the auditory cortex and positive symptoms. Similar correlations with the default network are seen in depression with structures that are implicated with affective regulation such as the subgenual prefrontal cortex. This suggests that low-frequency fluctuations of the BOLD signal may be tapping something which is not seen with either positron emission tomography or conventional functional MRI. Whether the default network, as defined with this technique, reflects a fundamental functional network in the brain related to a number of neuropsychiatric disorders has yet to be determined. Nevertheless, further studies in more carefully defined, untreated patients seem to be warranted.