However, the most recent contribution to this field comes from structural brain imaging in patients with schizophrenia or psychosis generally which have sought specific neurological correlates of poor insight. Most (although not all; [11]) MRI structural imaging studies of such patients have found significant relationships between lack of insight (variously defined) and a range of structural deficits (summarized in [12]). More recent sophisticated structural imaging studies have revealed relationships with specific brain regions or with fronto-temporal white matter [13]. The majority have found evidence for a relationship between poorer insight and either volumetric reduction or thinning of various cortical midline regions, most towards the anterior (frontal lobe: [14] medial-orbital prefrontal cortex PFC: [15]; medial PFC: [16,17]; medial-superior PFC: [18]; anterior cingulate: [15,19]; paracentral lobule: [16]) but also the posterior parts (posterior cingulate: [12,19]; precuneus: [18,20]; [16]). Two studies found poorer insight to correlate with increased volume in anterior midline regions [21,22]. Finally, a variety of other regions of reduced volume have also been implicated in the studies reviewed above including dorsolateral prefrontal cortex (DLPFC [23]), insula [24] and temporo-parietal regions [20].

The extent to which clinical insight is related to more general self-reflective and self-evaluative¹ processes – part of metacognition [25] - is beginning to be addressed. This is important, not only because having an accurate representation of one’s traits, abilities and attitudes is essential to evaluating one’s own behaviour and hence adjusting it to social circumstances [26], but also because it provides a plausible normative framework within which to understand lack of insight in psychiatry. Such a framework would get round the problem of how to examine processes and models relevant to acceptance of mental disorder in a range of individuals.

The functional neuroanatomy of self-evaluation in healthy subjects has begun to be mapped [27]. In the most commonly used experimental paradigm, subjects are presented with a trait adjective and are asked whether it applies to them as opposed to another person (a friend, relative or famous personality). The results have been the subject of conceptual reviews and meta-analyses which demonstrate that a core set of regions – cortical midline structures (CMS) - are consistently engaged in tasks in which the self is the object of contemplation [28-31]; and this applies to mental states as well as personality characteristics [32]. The CMS comprises medial pre-frontal cortex (MPFC), posterior cingulate cortex (PCC) and anterior cingulate cortex (ACC). Self-evaluation can be broken down into several component processing steps such as directing attentional focus to oneself, followed by holding information in mind (working memory) in order to carry out a comparison with stored representations (episodic/autobiographical memory), all of which lead to a judgement or appraisal (executive functions). Hence, several brain regions commonly associated with component processes (eg DLPFC, medial temporal lobe; inferior parietal lobe, etc.) would be expected to play a role in such tasks [29,33]. Moreover there is clearly overlap between processes and networks which enable self and other evaluation, the precise extent of which is currently debated [26,31,33,34]. It has been claimed that that the greater the social distance between the self and the other, the more likely self activation regions will appear distinct ([35,36]; but see [37]).

Few functional imaging studies using fMRI have examined this issue in schizophrenia. Perhaps the first [38] showed a correlation between improvement in clinical insight scores with activation of medial PFC during an empathy task. However, Murphy et al. [39], were first to use a task that required participants to make decisions about the self-relevance of positive personality traits, and did not find significant differences in activation between schizophrenia patients and healthy controls. By contrast, Holt and colleagues [40] used a similar task (but with negative as well as positive personality traits) and found that patients displayed lower activation of the ventromedial PFC but higher activation of the median and posterior cingulate during self-evaluation compared to healthy controls; however, these authors did not examine the activity of brain regions outside of the cingulate. Modinos et al. [41] studied theoretically psychosis prone students with fMRI and found increased activation in CMS (plus insula) with a valenced self-reflection task compared to those less psychosis prone. Finally, work in a different diagnostic group, those with traumatic brain injury [42] showed increased activation in posterior and anterior CMS compared to controls, with activation correlated with insight into cognitive deficits.

We predicted that the schizophrenia group would show reduced activation during self-evaluation in regions within the CMS. This was supported with a single cluster of voxels within the main medial frontal region which showed a significant interaction between diagnostic group and self-evaluation when responses during all the trait types were combined, thus giving the most robust data (see Figure ​Figure2).2). A second region of significant interaction was also noted in the right posterior cingulate, a region well established in the self-appraisal literature [29,30,56] after adjustment for reaction time. The PCC appears to be important in experiential self reflection [56] and was also highlighted as a region of reduced grey matter density in relation to the symptom relabeling dimension of insight [12]. The medial frontal locus is somewhat more superior (z=32) than the main focus derived from meta-analyses (z=6; [31]) although consistent with the self-appraisal imaging literature [29,30,54]. The middle-posterior cingulate was shown by Holt et al. [40] to be more activated in schizophrenia patients than controls in a similar self-reflection paradigm. They went on to show abnormal connectivity between this region and more anterior parts of the cingulate, which was also ‘overactive’ in our patients during self evaluation of specific traits. Exploratory analyses were carried out to look at cerebral responses to self-evaluation of individual traits and correlation with insight, which was also part of the hypothesis under test. Interactions indicated that the patients were particularly hypoactive during self-evaluation in the right medial frontal gyrus (positive traits), left superior frontal and superior temporal gyrii (negative traits) and right inferior parietal lobule (BA40; mental illness traits). There was no obvious pattern to indicate that the valence of traits was decisive in determining the locus of activation. Lower activity in all these areas except the right medial frontal gyrus tended to show weak-to-moderate correlations with lower SAI-E insight in the patients (data not shown). Interactions indicated that the patients were particularly hyperactive during self-evaluation in the right anterior cingulate and bilateral precuneus (positive traits), right middle frontal gyrus (BA10; mental illness traits), and left inferior parietal lobule (BA40; physical illness traits). Higher activity in all these areas showed a correlation with lower SAI-E insight in the patients and this was significant in the right middle frontal gyrus for mental illness traits and the left precuneus for positive traits (Figures ​(Figures44 and ​and5).5). Such activity was also related to higher endorsement of positive traits in both healthy controls and patients which is consistent with the observation that grandiosity and lack of insight frequently go together [44,57,58].

Activation in the non-dominant parietal lobe is of interest since damage to this region is a frequent finding in neurological patients with anosognosia [59] and has been linked with disorders of awareness in schizophrenia [60]. While hypoactivation (in relation to mental illness traits) was noted, so too was bilateral inferior parietal (plus somatosensory cortex and ACC) hyper-activation in relation to physical illness terms which could plausibly be interpreted as correlating with ‘somatic insight’. While relative over- and under-activation should be interpreted cautiously it may be tempting to consider lack of insight as having both deficit (failure to engage in self-evaluation) and excess (active ‘denial’) elements. Correlating fMRI activation and the clinician rated insight score is a partial validation of the method used to elicit self-appraisal of mental disorder although we were not in a position to validate or confirm the participants’ accuracy of self-appraisal of personality traits.

However, we suggest that the main finding of reduced dorso-medial PFC activation in patients versus controls is best explained by a failure to carry out effective self-evaluation and not merely a reluctance to engage in the activity. This correlated with the symptom awareness component of the insight scale particularly although other components weakly related. Our own work (with these patients [54,61]) and that of others using subjective reports of self-reflection does not paint a picture of lack of effort to understand the self in schizophrenia but rather a failure to reach satisfactory conclusions or at least conclusions which accord with the consensus judgements of others [61-63]. In acquired brain injury it appears that CMS activity is increased during attempts to self-reflect (compared to controls) and that the greater activation, the greater the insight [42]. An analogous pattern is seen in healthy students who are theoretically psychosis prone [40], as if to say these states place greater demands on self-evaluation systems. We propose that with established schizophrenia the ability to self-evaluate effectively, diminishes [64] and this is mirrored in reduced brain activation in key structures underpinning the function. The possibility that this may recover with recovery of the illness forms an intriguing hypothesis (see [38]) and would be consistent with insight having both state and trait components [65]. The precise nature and content of ‘generic’ self-evaluation is presumably modulated by other brain systems, as evidenced by patterns of increased and decreased activation on fMRI in posterior cerebral regions (PCC, precuneus, occipital cortex) and other more lateral frontal, parietal and temporal regions. Inferior parietal, precuneus and right middle frontal regions providing mental-illness-specific ‘fine tuning’.

The functional abnormalities discussed above were specific to self-evaluation. However the medial PFC is consistently implicated in both structural and functional abnormalities in schizophrenia per se with multimodal imaging techniques including diffusion tensor imaging, voxel based morphometry and fMRI [66]. This background plus the specific structural MRI findings in relation to insight [12,14,18] speak to the ‘trait’ aspects of poor insight and its associated cognitive deficits [9,67,68]. The patients in the current study had slightly lower IQ than controls although they showed well preserved functioning on working memory and set shifting, so cognitive impairment is unlikely to explain the results to a large extent. Nevertheless, the partial structural brain basis for poor insight in schizophrenia may explain why correlations with insight scores were not strong in relation to the locus of diminished medial PFC activation in the current fMRI study, since variation in BOLD signal is presumably limited by the neural substrate.