The nature of the general cognitive impairment in schizophrenia

Psychometricians have long recognised that the performance of healthy populations on batteries of diverse neuropsychological tests varies at three hierarchical levels. These are: ability on specific tests; ability on groups of tests assessing similar cognitive constructs (such as working memory, perceptual organisation); and general ability. ^{Reference Deary6} General ability or g is thought to contribute to the performance of all cognitive tests and approximately 50% of the variance in performance across tests can be accounted for by differences at this level. ^{Reference Deary6} Statistical accounts of neuropsychological performance further suggest that there are two highly inter-correlated components of g: ‘crystalline intelligence’, derived from knowledge-based tests that rely on education and past experience, and ‘fluid intelligence’, calculated from tests of reasoning and rapid information processing. ^{Reference Deary6} The general ability of individuals is commonly measured by instruments such as the Wechsler Adult Intelligence Scale (WAIS) ^{Reference Weschsler7} that assess a wide range of abilities from which a composite score (IQ) is derived as well as the crystalline and fluid components denoted as verbal IQ and performance IQ respectively.

Recent research has investigated whether there is more to the concept of general ability than a statistical construct. In other words, whether g is an entity characterised by a set of fundamental cognitive processes with a neural basis. In this respect, two things are noteworthy about neuropsychological test performance in schizophrenia. First, when compared with healthy populations, the difference in performance is mostly at the level of general ability. ^{Reference Dickinson, Goldberg, Gold, Elvevag and Weinberger8} Second, performance is less differentiated than normal in that more of the within-group variation is accounted for by general than specific ability. ^{Reference Dickinson, Goldberg, Gold, Elvevag and Weinberger8} This pattern of effects suggests that people with schizophrenia have a more limited general cognitive resource than average and that this constrains the performance of a wide range of specific cognitive functions. Arguably, if a specific neurobiological counterpart of this general resource can be identified, an important clue to the aetiology of schizophrenia will be revealed.

One lead concerns information-processing speed, which has been shown to explain up to 25% of the variance in the general ability of healthy populations. ^{Reference Deary6} This is relevant because a meta-analysis of the performance of schizophrenia and control groups on a large number of different neuropsychological tests found that the greatest effect size for group difference was not for IQ but for a test of processing speed. ^{Reference Dickinson, Ramsey and Gold9} In support of the importance of processing speed, several other studies have found that even when patients with schizophrenia and healthy controls are precisely matched for WAIS IQ, there is a mismatch in performance across WAIS subtests so that in schizophrenia, the processing speed domain is disproportionately impaired. ^{Reference Leeson, Barnes, Harrison, Matheson, Harrison and Mutsatsa10} This suggests that reduced processing speed may be one of the fundamental causes of the general cognitive impairment in schizophrenia. Regarding neurobiology, speed of processing is highly dependent on the integrity of myelinated axons. This can be measured in the brain with magnetic resonance diffusion tensor imaging. Using this methodology, a recent study of healthy people found that WAIS IQ correlated with an index of white matter integrity across 12 major white matter tracts. ^{Reference Penke, Maniega, Bastin, Valdes Hernandez, Murray and Royle11} Importantly, structural equation modelling showed that this effect was fully mediated by information-processing speed measured separately with non-WAIS tasks. As there is now compelling evidence that white matter development is abnormal in schizophrenia, these findings provide a testable hypothesis for linking generalised cognitive impairment in schizophrenia to a neurodevelopment risk factor.

Other research into the neurobiological underpinnings of g concerns executive function. This consists of a set of cognitive processes that serve to control behaviour in the face of changing environmental contingencies and includes working memory, attention switching and response inhibition. The cognitive requirements of tests of fluid intelligence overlap considerably with those of executive function and functional imaging studies have linked both fluid intelligence and executive function with a common neural substrate. Thus, performance accuracy on a working memory task at a point requiring maximal executive control was shown to depend on fluid intelligence ability on the one hand and the degree of engagement of specific areas in frontal and parietal cortex on the other. ^{Reference Gray, Chabris and Braver12} Human lesion studies have also determined that fluid intelligence is supported by brain regions corresponding to those identified in functional imaging studies, namely posterolateral frontal, dorsomedial frontal and midparietal cortex. In one important study the degree of volume loss within, but not outside, these regions was critically related to the loss of fluid intelligence. ^{Reference Gray, Chabris and Braver12} The implication of these and other studies is that there is a frontoparietal neural network that provides the processing requirements for the efficient execution of disparate cognitive tasks that enable the environmentally adaptive behaviour encapsulated in the concept of fluid intelligence. ^{Reference Gray, Chabris and Braver12–Reference Jung and Haier14}

This hypothesis provides a unitary framework for reconciling the diverse cognitive findings in schizophrenia. That is, the dysfunction of a specific set of cortical areas and their white matter connections may account for a substantial amount of the cognitive impairment seen in this disorder. Illustrating this point is the finding of a meta-analysis of functional imaging studies that examined frontal lobe function in patients with schizophrenia and healthy controls. The studies that were evaluated used a variety of different cognitive tests but a common set of frontal cortical areas was activated each time. These conformed to nodes of the same neural network implicated in the support of intelligence. Importantly, although the same areas were activated in both groups, this was atypical in patients suggesting dysfunctional processing within this network. ^{Reference Minzenberg, Laird, Thelen, Carter and Glahn15}

The brain abnormalities of schizophrenia admittedly extend well beyond this frontoparietal network and involve other cortical areas, subcortical structures and their white matter connections. Dysfunction of these different neural networks is also likely to account for other important facets of the disorder, such as impaired emotional and social cognition, and they ultimately may prove equally or more deleterious to everyday function than intellectual impairment. However, in terms of elucidating the neurobiological substrate of the disorder, the generalised cognitive impairment may be more tractable than these other problematic areas. This is because the cognitive processes mediated by the frontoparietal network lend themselves more to comparative and computational neuroscience. For example, Duncan has shown that the frontoparietal network is activated by diverse cognitive task requirements in healthy people. On the basis of this, he proposes that it serves as a ‘multiple demand system’ important for the ‘programming of goal directed behaviour’ subserving fluid intelligence. ^{Reference Duncan16} Using analogies with artificial intelligence and data from studies of non-human primate single-cell recordings and human functional neuroimaging, he provides evidence for how this system can exert cognitive control. This includes the fractionation of encountered tasks into essential subroutines, the switching of attention from one to another in a sequential manner and the extraction of important information from each step to pass onto to the next for successful completion of the programme.

Implications

An understanding of the core functions of this frontoparietal network, its neurobiology and how it is influenced during development may lead to important new insights into the causative abnormal mechanisms of schizophrenia. Such research may also elucidate whether the plasticity of this system is modifiable during adult life, for example by pharmacological or psychological means. If so, this will illuminate the search for much needed new methods of remediation.

Funding

E.M.J. is supported by the UCLH/UCL Biomedical Research Centre.