Although previous studies have found associations between symptoms and functional imaging correlates in patients with established illness, the current study involved a large number of un-medicated high risk subjects experiencing psychotic symptoms and as such avoids potential contamination of findings by anti-psychotic medication and chronicity of disease. We performed these analyses by examining the effects of the presence of these symptoms on the neural responses to cognitive tasks. The tasks were not designed specifically to elicit symptoms, but were chosen to involve networks and cognitive processes known to be abnormal in patients with schizophrenia. The results indicated that symptom scores correlated with brain activation in a number of regions broadly consistent with the extant literature. Since in the analysis conducted the three symptom dimensions were modelled simultaneously within one design matrix, intersubject variability common to two or more symptom dimensions is not reflected in the correlation values. The values obtained therefore represent correlations specific to each symptom dimension which is not accounted for by either of the other two. However, due to the approach taken, we would like to state that since all high risk subjects were entered into the analysis, the correlation values may to some extent reflect the presence of symptoms as well as severity. As hypothesised there were significant correlations with the hallucination score and activation in the left lateral temporal cortex for the sentence completion task. For the encoding/retrieval task there were significant correlations with the suspiciousness/persecution score and activation in the bilateral MTL. Across both tasks there were significant correlations between delusions and suspiciousness/persecution scores and activation in the cerebellum, with different patterns of laterality.
The finding of an association between the hallucinations measure and activation in lateral temporal cortex is consistent with other reports linking this region to auditory hallucinations [9–11, 14], and with cognitive neuropsychological theories of auditory verbal hallucinations regarding the misattribution of inner speech [1, 34]. Although regions most typically associated with hallucinations are mid to posterior sections of the superior temporal gyrus, other studies have indeed reported associations with activation in the middle temporal gyrus [10, 14, 16]. The cognitive model most consistent with our task and results is that auditory hallucinations represent sub-vocal verbalisation which is not associated with the usual inhibition of auditory cortex that signifies self-production .
The memory task revealed strong negative correlations between the suspiciousness/persecution score and activation in bilateral MTL structures. This association was seen for correct rejection versus experimental baseline and not correct recognition. There is previous evidence however to indicate MTL regions are preferentially involved in correctly identifying new words (correct rejection) compared to correctly identifying old words (correct recognition) . Previous literature has also implicated MTL regions in the formation of positive psychotic symptoms [3–5, 36], and those that have examined delusional beliefs specifically also report decreased activation with increasing symptom severity . There is also substantial evidence for structural abnormalities in this region in patients with established schizophrenia and high risk individuals [37–39]. The MTL, including the amygdala and hippocampus, are known to be involved in emotional, mnemonic and social processing [40–42] and dysregulation of these processes may contribute to the development of persecutory beliefs. For example, persecutory beliefs have been hypothesized to arise from a disruption in the processes mediating the formation and maintenance of normal social beliefs, and patients with persecutory beliefs have been shown to attend excessively to threat-related stimuli and to preferentially recall such information . Impaired processing of threat-related stimuli has also been previously linked to dysfunction of the amygdala in paranoid patients with schizophrenia .
Although not part of our original hypothesis, we also found correlations between both suspiciousness/persecution and delusion scores and activation of the cerebellum across both of the tasks. Abnormal cerebellar function has been previously reported in the literature on the established illness  and in those at high genetic risk . Further, it has been reported that patients with posterior lobe lesions have deficits which are not restricted to motor function, but also present affective disturbances, disinhibition, and psychotic symptoms [46, 47]. The cerebellum has also been implicated in the 'cognitive dysmetria' model of schizophrenia where dysfunction in cortico-cerebellar-thalamic-cortical circuits are proposed to result in the abnormal synchrony or co-ordination of mental processing, ultimately resulting in the clinical features of the illness . Although there are few imaging studies of patients with persecutory delusions, neuropsychological studies have repeatedly demonstrated that patients commonly attribute negative events to external sources . Neurobiological theories have also directly implicated the cerebellum regarding deficits of 'internal monitoring' which are suggested, for example, to underlie the formation of delusions of alien control where the patient attributes their own actions to an external agent . The current findings are therefore consistent with a growing literature implicating a role for the cerebellum in the formation of delusions.
The nature of the correlations in the cerebellum are however complex. The correlations are primarily right sided with regards to the delusions score, and primarily left sided with the suspiciousness/persecution score. It has been suggested that different types of delusions may differ pathophysiologically and there is an important role for emotion in the formation of paranoid/suspicious delusions via their effects on memory . Following the model of brain function where emotion is predominantly associated with right sided cerebral regions, crossed cerebellar connections could mean that left sided cerebellar regions would be more involved in emotional/affective components of delusions, consistent with the present findings. Furthermore, since emotion is considered to be involved via its effect on memory processes, our MTL associations with the severity of suspiciousness/persecution also fits with this theory. We however recognise the speculative nature of these findings, particularly with regard to the paucity of other studies examining delusional beliefs. Further, in the group we have examined many had both hallucinations and delusions (including persecutory and non-persecutory) and it is therefore difficult to disambiguate individual components. That said, however, the fact that this was a study examining correlations with symptom scores reduces the chances of confounding effects by other symptoms. Studies directly manipulating experimental affective components in subjects with isolated symptoms may help clarify these results.