Neurocognitive Functioning in Patients with Schizophrenia

Terry E. Goldberg and James M. Gold

Increasingly, neurocognitive paradigms have been used to study patients with schizophrenia. Such paradigms use experimental and clinical tests to better characterize the cognitive abnormalities in schizophrenia. This approach differs from earlier psychological research, in that neurocognitive tests have been validated, often in a principled manner, in brain-damaged populations or by functional neuroimaging studies in normal controls, such that inferences can be made about the cerebral regions or neural systems that subserve them. Use of these techniques has also indicated that some types of cognitive impairment are central and enduring features of schizophrenia and are reliably present. Moreover, impairments may account for a significant proportion of the social and vocational morbidity in schizophrenia. The remainder of this chapter marshalls evidence that bears on this view.

Cognitive abnormalities were noted by early investigators of schizophrenia. In the original clinical descriptions of schizophrenia made by Kraepelin (45), he commented "Mental efficiency is always diminished to a considerable degree. The patients are distracted, inattentive . . . , they `cannot keep the thought in mind.' " Some years later, Shakow (66) began a series of studies in which he examined abnormalities in patients' readiness to respond to different types of imperative stimuli. Hunt and Cofer (38) noted that schizophrenic patients' intellectual quotient was lower than that of normal controls. However, the increasing influence of psychodynamic theory tended to minimize the significance of cognitive deficits of schizophrenia. It was thought that the deficits displayed on formal psychological testing were secondary to impaired motivation, gross breakdowns in reality testing, or disordered thought processes. Early application of neuropsychological testing in the assessment of schizophrenia was therefore to "rule out" an organic basis for psychiatric symptomatology that was thought somehow to be instantiated in mind, but not brain. It was within this context that a series of studies using broad neuropsychological test batteries found that chronic schizophrenic patients could not be reliably discriminated from heterogeneous brain-damaged populations on the basis of discriminant scores (48). The possibility that these results indicated that schizophrenia was, in fact, a brain disease was considered, but rejected by many researchers, who continued to emphasize the role of motivation, institutionalization, medication, and psychiatric symptomatology in cognitive impairment.

This view changed rapidly with the advent of in vivo brain imaging techniques. First, it became evident that patients with schizophrenia had larger lateral cerebral ventricles than did controls on computed tomography (CT) scans (68). Second, functional brain imaging suggested that schizophrenic patients had decreased frontal lobe blood flow and/or metabolism. Moreover, one type of cognitive impairment, namely, poor performance on the Wisconsin Card Sorting Test (WCST) was directly linked to impaired activation of the prefrontal cortex in regional cerebral blood flow (rCBF) (77). These findings led to a reinterpretation of the original neuropsychological studies. There was growing realization that patients with schizophrenia performed in the range typically found in brain-damaged populations, because schizophrenia involved structural and functional abnormalities of the brain that were, in some sense, primary.

Despite the consistency of the neurocognitive findings and the relevance they have to the interpretation of old findings, the relationship between symptomatology and cognitive impairment has been a continuing subtext in the scientific literature on schizophrenia. The distraction of hallucinations, lack of cooperation, distortions in interpersonal relationships caused by paranoid delusions, and neuroleptic medication have been thought to have an adverse effect on cognitive functions. Further complicating interpretation of the significance of these results is the heterogeneity in the clinical presentation of schizophrenia. A number of different subtyping schemes have been devised to account for this and include distinctions between patients with primarily positive symptoms (hallucinations, delusions, thought disorder) and those with negative symptoms (flat affect, impoverished speech), patients with family histories of schizophrenia and those without, male patients and female patients, and paranoid and nonparanoid patients.

We examine these crucial conceptual issues in turn: Are the deficits restricted to various subgroups given the clinical variability of the disorder within and across patients or are they present in nearly every patient? Are the cognitive deficits in some way real or are they artifactual? What are the core neurocognitive impairments in schizophrenia? What is the course of cognitive impairment in schizophrenia? What is the specificity of cognitive impairment both in terms of profile and severity level to schizophrenia as compared to other neuropsychiatric disorders? We conclude this chapter by noting that neurocognitive impairments may have prognostic significance in schizophrenia because of the importance of such functions in orienting to relevant environmental information, remembering new information, propitiously retrieving old information, and working on-line with old and new information to make responses or decisions. In this account, cognitive impairments in schizophrenia should be considered a target symptom that requires amelioration.

The frequency of cognitive impairment in schizophrenia has often been examined by using binary cut-offs. In these studies, less than 40% of cases are generally considered to be abnormal (6). Variations of this approach have been used when various subgroups of schizophrenic patients are compared. In particular, the paranoid–nonparanoid distinction has been examined. When differences were found, they usually favored the paranoid group; that is, the paranoid group was less severely impaired. Classification schemes have also distinguished between patients with primarily positive symptoms and those with primarily negative symptoms. Patients with negative symptoms often have been considered to exhibit more cognitive impairment. Factor analytical studies of symptomatology have led to refinements in classification between positive and negative symptoms with the addition of a third factor involving disorganized behavior and thought. Liddle (47) found evidence for differences in profile and frequency of cognitive impairment between the three symptom-driven syndromes.

However, using a paradigm that involved monozygotic twins discordant for schizophrenia, it is possible to arrive at different conclusions (28). In this paradigm, each affected individual is yoked to an ideal control, namely, his or her unaffected cotwin. Thus age, sex, and genetic factors are completely controlled, and educational opportunity, socioeconomic status, and family emotional climate are well controlled. Goldberg et al. found that the affected twins performed consistently worse than their unaffected cotwins on many of the tests from a large neuropsychological battery. Crucially, even when affected twins were ostensibly performing within the normal range, their cotwins performed at still higher levels. Thus, cognitive impairment was apparent throughout the whole range of the distribution and was not restricted to a subgroup of outliers. Hit ratios (the proportion of affected twins that performed below unaffected cotwins) for tests involving memory (Wechsler Memory Scale), attention and vigilance, verbal fluency, lexical access and response speed (Stroop), psychomotor speed and scanning (trailmaking), intelligence quotient or IQ [Wechsler Adult Intelligence Scale-Revised (WAIS-R)], and set shifting and abstraction (WCST) were very high, that is, greater than 85%. These differences occurred irrespective of patients' diagnosis (paranoid, undifferentiated, schizoaffective) or symptom profile. Moreover, these data suggest that it is important to note not only patients' current level of functioning, but the magnitude of decline relative to their potential. The finding is important because it suggests that while the level of a cognitive score may be an admixture of insult and endowment, the magnitude of decline may reflect more accurately the severity of insult.

In summary, a variety of findings suggest that a simple model in which patients with schizophrenia vary along a severity dimension on cognitive impairment cannot easily be rejected and, for parsimony's sake, can even be favored. Moreover, recent neurobiological studies support a view of homogeneity in schizophrenia. Daniel et al. (11) could not discern a bimodal distribution of ventricular brain ratios using mixture distribution analysis of over 1,000 published data points. Furthermore, Suddath et al. (71) using magnetic resonance imaging (MRI) data obtained from monozygotic twins discordant for schizophrenia, demonstrated consistent enlargement in various components in the ventricular system and diminution of the left anterior hippocampus. Similarly, Berman et al. (72) found that prefrontal cerebral blood flow measured while twins were being administered the WCST was lower in the affected than unaffected twin in all pairs discordant for schizophrenia. Thus, one of the more interesting features of schizophrenia may be that beyond the phenomenological level, which is exceptionally heterogeneous, patients may display unexpected homogeneity at the neuropsychological, neuroanatomical, and neurophysiological levels of analysis.

The clinical efficacy of a variety of antipsychotic medications in the reduction of positive psychotic symptoms in many patients with schizophrenia is beyond dispute. However, the cognitive effects of antipsychotic medications remain more poorly understood. Indeed, most of the literature examining this issue comes from the 1950s and 1960s when these medications were first widely utilized and the paucity of positive findings using cognitive outcome measures may well have led to their abandonment as a means of assessing the impact of medications whose clinical, symptomatic benefits were frequently obvious and dramatic. This abandonment was particularly unfortunate given the recent prominence of cognitive and neuropsychological research in the illness. Not unexpectedly, this older literature has a number of serious methodological limitations. Despite these limitations, several reviews of literature over the past decade have come to similar conclusions on the limited cognitive impact of chronic treatment with antipsychotic agents (36, 52).

Clearly, the most important question for neuropsychological research in schizophrenia is whether the impairments observed in patients might not be, in fact, the results of treatment. There are three major lines of evidence that argue strongly against this view. First, there is clear evidence of cognitive impairment noted in the literature from the preneuroleptic era. For example, Rapaport et al. (59) noted impairments of judgment, concentration, and planning ability and anticipation, and described memory and concept formation deficits. The work of Shakow and colleagues (66), documenting impairments in motor speed and reaction time, preceded the introduction of neuroleptic medications, as did the observation that the general intellectual function of patients appeared to have declined from premorbid levels. Thus, the basic cognitive "profile" of schizophrenic patients has been observed for nearly half a century, long before the introduction of antipsychotic medications. Recently, Saykin et al. (61) reported striking cognitive impairments in a sample of unmedicated patients. Thus, unmedicated patients meeting current diagnostic criteria for schizophrenia as well as older samples (which may have been heterogeneous) demonstrate deficits. Second, as noted (see below) many of the impairments observed in patients treated with neuroleptics have also been observed, albeit typically to a lesser degree, in unmedicated first-degree relatives of patients, in the premorbid histories of ill patients, and in populations of children at genetic risk for the illness. Thus, it appears that at least some deficits are related to vulnerability or risk for the illness rather than the treated full clinical syndrome. Third, and most directly addressing the issue, is the fact that there is simply very little evidence that chronic neuroleptic treatment impairs neuropsychological performance in ill patients. The data base supporting this conclusion involves multiple studies examining variables ranging from intelligence test performance, the Halstead Reitan Battery, reaction time, and graphomotor speed. Recent studies are generally in agreement with this older literature. For example, Seidman et al. (64) found that a marked reduction in neuroleptic dosage in patients who did not relapse had virtually no effect on performances in a wide variety of neurocognitive tests.

Several reviewers (see above) have pointed out two potentially important exceptions to this overall conclusion. First, timed measures of motor dexterity, such as the Purdue Pegboard, may be particularly vulnerable to the impact of neuroleptic dopamine blockade, a finding which is expected given the prominent role of dopaminergic neural transmission in the motor system. The extent to which this is a persistent deficit or a more temporary effect of treatment initiation remains unclear. Indeed, what is most remarkable is that simpler speed measures such as finger tapping, reaction time, and complex graphomotor tasks such as digit symbol coding, rarely demonstrate drug-related performance decrements.

A second area in which drug-related performance decrements have been reported involves memory functions. These appear to be sensitive to the impact of anticholinergic activity inherent in neuroleptics or in the medications used as adjunctive treatment of extrapyramidal side effects, such as rigidity and tremor, a negative effect fully expectable given the effect of anticholinergics in normal subjects (70). Although this adverse effect has been demonstrated in a number of study designs, its size generally appears to be small, and there is little reason to believe that the recent work documenting memory impairment in schizophrenic patients is the result of negative treatment effects: such effects were noted prior to neuroleptics and in unmedicated groups.

Neuroleptics do appear to offer certain limited cognitive benefits. It is clear that neuroleptics reduce formal thought disorder. Insofar as thought disorder interferes with performance (19), particularly on tasks requiring open-ended and coherent verbal responses, such as certain measures of verbal conceptualization and proverb interpretation, improvement has been noted following neuroleptic treatment. Perhaps the clearest evidence of neuroleptic-related performance enhancement has come on several measures of attention. The most consistent results have been observed on visual and auditory versions of the Continuous Performance Test (CPT), which generally span several minutes and require the subject to selectively respond to a specified target, such as a letter or sequential combination of letters (70). It should be noted that there is evidence that CPT deficits are also observed in high-risk populations and in subjects in states of relative symptomatic remission, suggesting that there are both stable trait- and clinical state-related dimensions of impairment (55). In addition to this evidence for improvements in sustained attention, there are several reports of improvement in measures of distractibility: patients are less stimulus bound while on active treatment (56). Mixed, but largely positive findings, have also emerged from studies using span of apprehension type tasks to assess selective attention during time-dependent visual information processing (1).

Although positive cognitive effects have typically coincided with symptomatic improvements, these may not be related in a simple causal fashion. Serper et al. (65) recently reported that patients on neuroleptics were able to benefit from extended practice on a dual task paradigm, whereas unmedicated patients at the same level of symptom severity were not. This suggests a direct pharmacological cognitive effect of antipsychotic treatment, independent of symptomatic status. It is intriguing that these enhancements appear to be somewhat selective, as many other measures of attention such as reaction time or digit symbol do not appear overly sensitive to neuroleptic effects.

Admittedly, the literature addressing the cognitive effects of neuroleptics has many shortcomings and several investigators have recently again raised the issue of negative effects on the basis of finding a correlation of impaired performance and drug dose or by comparing patients on and off neuroleptics when treatment status was not randomly assigned. Although the results of such studies are of interest, they remain difficult to interpret as the issue of clinical state, compliance, and dose are inevitably confounded. Furthermore, there may be large individual differences in the relationship between oral neuroleptic dose and actual central nervous system bioavailability. Thus, the possibility of specific adverse effects remains an open issue, which will require a new generation of studies to answer. Indeed, the comparison of the newer atypical antipsychotics with current standard treatments will offer another opportunity to examine the relationship of symptomatic status and cognitive functions. Initial results suggest that atypical neuroleptics do not offer cognitive benefits concomitant with more effective treatment of psychotic symptoms, and in this way very much resemble the lack of cognitive benefits associated with conventional treatments (see below).

The prototypical atypical neuroleptic clozapine has been used to dissociate symptoms from cognition (27). The design of the study was straightforward: Patients were first neuropsychologically examined while receiving a variety of typical neuroleptic medications. Symptoms were rated on the Brief Psychiatric Rating Scale (BPRS). Patients then received clozapine for an average of 15 months and were reassessed. Psychiatric symptoms improved significantly; symptom scores declined by 35% as had been shown previously (41). However, no cognitive measure improved. Thus, no significant changes were observed in WAIS-R; trailmaking; the WCST; the Category Test; tests of visual processing, namely, facial recognition and judgment line orientation; and measures of verbal memory from the Wechsler Memory Scale. Visual memory for designs declined. Moreover, the subgroup of patients that improved the most psychiatrically did not differ from the subgroup that improved least on tests of cognitive function (as measured by change in scores). These results suggested that symptoms and cognitive function in this sample of patients with schizophrenia were dissociable, in that despite remarkable improvement in symptom status due to clozapine, no change in cognitive function was noted. Thus, it appeared that the data were consistent with the conclusion that neuropsychological deficits might be fundamental manifestations of the illness, and that the cognitive and symptomatic effects of neuroleptic treatment are independent.

A second study found that clozapine improved tasks requiring speeded motor output including verbal fluency, mazes, and digit symbol coding, but not executive set shifting on the WCST (35). These results may have been attributable to reductions in extrapyramidal toxicity associated with clozapine. Moreover, performances generally remained in the impaired range and had tenuous relations to changes in symptomatology.

Another approach that is useful in assessing the strength of the association between psychiatric symptoms and cognitive function in schizophrenia involves correlational statistics. Faustman et al. (14) found no relation between Luria Nebraska Battery neurocognitive performance and symptoms in unmedicated patients. These results are consistent with those found in a study that directly compared the relationship between symptoms and cognition in patients with schizophrenia and mood disorder (25). In unipolar depressed and bipolar groups, redundancy analysis (derived from canonical correlations) indicated that symptom factors of the BPRS accounted for 15% to 30% of the variance in cognition. However, in the schizophrenic group, analysis suggested that symptoms accounted for less than 5% of variance in cognition.

Based on an extensive review of original and variegated studies, Gold et al. (19) concluded that overall symptom severity had little impact on the cognitive level of neuropsychological tests or indices. Moreover, the association of positive symptoms and cognition has been found to be limited. This is surprising, as on the face of it such symptoms might be thought to interfere with information processing. The relation between cognitive failures and negative symptoms is somewhat stronger. However, it is often the case that negative symptoms improve with treatment, whereas cognition does not. Thus, neurocognitive testing may be a more reliable and more precise way of measuring deficiencies in volition, planning, and productivity that have a chronic, adverse effect on a patient's adjustment.

There are two sharply contrasting views of the course of cognitive function in schizophrenia. One view suggests that cognitive deficits become progressively worse throughout the long duration of the illness. After an insidious onset, patients intellectual functions become weaker and social skills become coarser (53). A second view suggests that cognitive deficits, once they arise, are relatively stable. It is thus consistent with the notion of a static encephalopathy. There is much longitudinal, cross-sectional, and correlational evidence that supports the latter view.

A number of different research designs have been utilized to study cognitive performance in schizophrenia prior to overt onset of the disorder. In the high-risk approach, consistent deficits have been found on various measures of vigilance, including the degraded version of the CPT and selective attention in tachistoscopically presented span of apprehension task in children (10, 55). Comparison of unaffected and affected members of monozygotic twin pairs discordant for schizophrenia and comparisons of the unaffected group and normal monozygotic twins have also yielded information on premorbid neurocognitive function. Given that prior to the onset of illness in the affected twin, members of these discordant pairs were similar cognitively (31) on measures of putative premorbid IQ and school grades, the functioning of the unaffected twin may be an accurate representation of the premorbid performance of the affected twin. Observations of subtle attenuations (i.e., performance was at the low end of average) on some aspects of visual and verbal memory, psychomotor speed and scanning, speed of access to the lexicon on automatized naming, and abstraction and set shifting were found in the unaffected group in comparisons with the normal twins and suggest a premorbid "forme fruste" profile of dysfunction.

Once the clinical manifestations of the illness become overt, there is evidence to suggest that a sharp decline in cognitive ability occurs. In longitudinal studies spanning the premorbid and morbid periods, Schwartzman and Douglass (63) found that schizophrenic patients who were tested on an army intelligence examination displayed a significant decrement (of nearly 0.5 standard deviations) in performance, whereas controls improved their score. (The patients were similar to controls in the premorbid period.)

Recent studies of patients during their first episode of schizophrenia substantiate the view that marked cognitive abnormalities are present at the very onset of the illness. Goldberg et al. (26) found that adolescents with diagnoses of schizophreniform disorder or schizophrenia demonstrated patterns in IQ similar to those of patients with chronic schizophrenia (i.e., lower performance IQ than verbal IQ), but dissimilar to psychiatric controls with adjustment and conduct disorders (who had lower verbal IQ than performance IQ.) Both groups had similar premorbid abilities as measured by a reading test. Hoff et al. (37) found that the neuropsychological profile of first-episode schizophrenic patients was remarkably similar to that of patients with chronic schizophrenia. Both groups performed poorly on a wide range of tests that included those assessing memory, executive functioning, and attentional abilities. Bilder et al. (4) examined a large group of first-break patients and found neuropsychological deficits in language, motor, attention, executive, and memory ability. Deficits were as frequent, though not as severe, as those found in a group of chronic patients.

During the chronic phase of the illness, a number of longitudinal studies have suggested that there is no progressive deterioration. For example, Klonoff et al. (43) studied a large group of patients with chronic schizophrenia over an 8-year period and observed that no deterioration was evident in Wechsler IQ.

A number of cross-sectional studies also examined decline during the chronic phase of the illness. Davidson et al. (12) found that large cohorts of patients lost approximately one to two points on the Mini Mental State Examination per decade of life. This rate is greater than that which occurs in normal groups. It is, however, much slower than that found in patients with a progressive cortical dementia such as Alzheimer's disease. Goldstein and Zubin (33) found no differences on the complex cognitive tasks of the Halstead Reitan Battery between large samples of younger and older chronic schizophrenic patients. Hyde et al. (39) also used a cross-sectional approach, in which successive cohorts of schizophrenic patients were assessed. Although a cross-sectional paradigm has obvious shortcomings, including increased variability, as comparisons are made between groups rather than within subjects and decreased power, the study design also had strengths in that it allowed comparison over an extremely wide range of duration of illness (patients ranged in age from 18 to 70 years) and it minimized attrition. In addition, each cohort was matched on a measure of premorbid intellectual ability, and patients with confounding neurological or systemic diseases were excluded. No significant age cohort differences were noted on tests known to be sensitive to progressive dementias: the Mini Mental State Examination, Dementia Rating Scale, verbal list learning, and semantic fluency. A cohort effect was found for the Boston Naming Test, but this appeared to be due to age rather than duration of illness. Thus, over five decades of illness, there was no indication of relentless cognitive decline.

This view of the natural history of schizophrenia is consistent with a neurodevelopmental perspective (76) in that a prenatal lesion remains silent for years before manifesting itself in overt symptomatology and cognitive impairment. There is much evidence that the markers of such a lesion are static. Thus, serial CT scan studies have indicated that there is no progression in enlargement of the ventricular system, nor are gliotic changes indicative of atrophy (26). Contrary to recent interpretations, Kraepelin (45) held to this account, stating, "As a rule, if no essential improvement intervenes in at most two of three years after the appearance of the more striking morbid phenomena, a state of weak mindedness will be developed which usually changes slowly and insignificantly."

Early description of the clinical phenomenology of schizophrenia emphasized the impairment of volitional attention. This clinical observation has been amply supported by many years of experimental study using a wide variety of tasks. Recent neuropsychological research has suggested that attention is not a unitary construct and is likely to involve several different component operations mediated by multiple brain areas (54). However, even these more specific frameworks do not appear to provide a useful heuristic for understanding the impairment in schizophrenia as available data would suggest that most, if not all, attentional processes (such as alerting, sustaining attention, rapid encoding, shifting) are impaired to some degree. This conclusion is similar to that voiced earlier in a review of work inspired by information-processing models, which were the focus of intense research interest a decade ago. The promise of these models was to isolate a particular stage of processing that was particularly impaired in schizophrenia. However, impairment limited to a specific stage could not be identified, leading to the notion that schizophrenia involves a general limitation of attentional processing resources, a limitation likely to effect a wide range of tasks. Although these results may be disappointing from the point of view of anatomical specificity, there is little doubt that impairments of attention are among the most frequently observed cognitive deficits in schizophrenia, and further developments in the understanding of normal attentional processes may allow for a reinterpretation of this rich experimental literature.

Several experimental paradigms have played a central role in this research tradition. The study of various reaction-time paradigms has produced reliable evidence that patients are typically slow and variable in initiating volitional responses. Such slowing may also be a characteristic of perceptual processes as patients often require longer stimulus exposures in backward masking paradigms before they are able to identify various target stimuli (7).

The CPT has played a central role in the study of sustained attention. Typically patients make an excessive number of errors, both failing to respond to targets and responding incorrectly to nontarget stimuli. Although the CPT is typically regarded as a test of the ability to sustain attention, it appears that patients make errors throughout the test, suggesting that the deficit is one in selective attention and maintaining a readiness to respond.

Such a readiness deficit has also been demonstrated in the reaction time literature where patients fail to benefit normally from a regular predictable series of preparatory intervals between warning cues and imperative stimuli (66). This type of inability to benefit from warning cues has been elaborated in studies of the startle reflex and electrophysiological studies of the P50, an evoked potential that is elicited when pairs of stimuli are presented and the initial stimulus serves to prepare the subject for the forthcoming stimulus resulting in a lessened response to the second stimulus (7). Thus, across paradigms, it appears that patients often fail to benefit from a variety of cues to facilitate either overt responses or perceptual processes.

Studies with span-of-apprehension type tasks demonstrate that patients are unable to quickly process large amounts of data. In this task, subjects view very brief presentations of up to 12 letters and have to identify the presence of a specific target. Patients can perform such discriminations fairly well with small arrays. However, their performance declines rapidly with large arrays. This type of paradigm suggests a failure in the amount of selective information that can be processed (1). In addition, several studies have demonstrated that patients are unusually sensitive to the presence of distraction, with their already diminished processing capabilities deteriorating even further when the additional demand is made to ignore irrelevant stimuli (56).

The literature documenting attentional impairment is voluminous, and we have only discussed some of the more prominent themes. As noted above, the cognitive-anatomical interpretation of these results remains highly ambiguous. On a more practical level, the relationship between attentional deficits and the compromise of more complex cognitive processes in schizophrenia remains unclear. Certainly, other patient populations with attentional disorders such as children with Attention Deficit Hyperactivity Disorder, do not typically present with severe generalized deficits. Furthermore, it appears that treatment with antipsychotic medications often improves (but fails to fully normalize) attentional functions in patients with schizophrenia, but has little benefit for more complex cognitive functions (see above). Thus, impairment of attention does not necessarily predict impairment in other cognitive domains, and change in attentional performance may not be predictive of broader cognitive changes.

Memory functions have been frequently assayed in schizophrenia. Recall of verbal paired associates, stories, recurring digits, and geometric designs has been reported to be deficient (61). These deficits are often prominent even against a background of general intellectual impairment. For example, Gold et al. (22) found that schizophrenic patients frequently displayed large discrepancies between IQ and the Wechsler Memory Scale-Revised General Memory Index, with the latter test being lower, and that these deficits were not attributable to attentional dysfunction.

Various stages in the processing of declarative memory (material encoded in distinct, event-related, spatiotemporal contexts) have been implicated. A number of investigators have suggested that schizophrenic patients use inefficient encoding strategies and thus do not take advantage of semantic regularities (21). Inefficient retrieval strategies or poor effortful recall have also been noted. When memory for materials after delays of 20 min or more is assessed, schizophrenic patients demonstrate only mildly accelerated rates of forgetting (22, 30), in contradistinction to amnesic patients, whose rate of forgetting is markedly accelerated. The rate of learning for item-based lists of words has also been studied. Schizophrenic patients display the capacity to learn lists, but do so at rates slower than that of normal subjects. Reduced memory for the initial portions of verbal lists (the primacy effect) has also been found (51). These three pieces of information (slowed rate of learning, mildly increased rate of forgetting, and lack of a primacy effect) may indicate that patients have difficulty consolidating material in a long-term store.

Recognition paradigms, which involve less elaborate search strategies (subjects simply state whether a target is more familiar or has been presented more recently than a foil) and thus may measure consolidation, have elicited conflicting results. Patients sometimes display relatively better, although not necessarily normal, performance on these tasks compared to recall memory. However, it is possible that these results are an artifact of differences in the difficulty between recall and recognition tests (8). Moreover, in some studies, recall and recognition have been highly correlated (21).

Another type of memory system that is theoretically dissociable from the explicit or declarative memory system is the implicit system. In implicit paradigms, subjects are not called upon to consciously recall material, but demonstrate item-specific learning when required to perform certain tasks in which stimuli have in some way been encountered previously. In a systematic study, Schwartz et al. (62) found that schizophrenic patients exhibited normal implicit memory on tasks involving the identification of perceptually degraded material and on semantic category production tasks, whereas explicit cued recall was impaired.

Motor skill or procedural learning has also been tested. In paradigms that involved pursuing a rotating target with a stylus, schizophrenic patients appear to be able to increase their ability to track as well as controls (30).

Perhaps the widest discrepancy between clinical observations and formal assessment occurs in the language domain. While the conversation of schizophrenia is often marked by lack of pronominal referents, illogicality, and derailment, patients may perform unexpectedly well on tests of language. Rausch et al. (60) found that schizophrenic patients performed similarly to normal controls and significantly better than aphasic patients on tests of the ability to apply linguistic rules. Core verbal tests from the WAIS-R that include expressive vocabulary, knowledge of commonplace information, abstraction of similarities, and expression of comprehension of social situations may be near normal (20).

The basis for patients' disordered use of language might theoretically lie in abnormalities in the semantic organization. Semantic priming studies (50) have indicated that patients with schizophrenia demonstrate greater facilitation of reaction time to a semantically primed target word than do normal subject (i.e., a response to cat is faster if preceded by dog rather than stone). Furthermore, Gourovitch et al. (34) also found that semantic fluency performances were more impaired than phonological fluency. This also implicates the organization of the semantic system. However, the relationship between these types of semantic system abnormalities and disordered speech has not been empirically tested.

The performance of patients with schizophrenia on various tests of visual processing has often been found to be intact. Two distinct cognitive systems are thought to play a role in this type of processing. An object location system determines the orientation of objects in space and the relationship between one object and another in space, or where something is; an object recognition system determines an object's identity based on a critical set of perceptual attributes, or what something is. Tests of location, involving spatial analysis including block design and object assembly from the WAIS-R test and judgment of line orientation, generally elicit normal performance in patients with schizophrenia (28, 44). Tests of the object recognition system, which includes facial perceptual matching have elicited somewhat more equivocal data as group differences or effect sizes are often larger than those on object location tests, but are not often significant. These findings imply that some posterior cortical zones may be relatively uncompromised in schizophrenia.

Patients with schizophrenia often seem unable to maintain some form of volitional control over information processing. They appear to have difficulty formulating plans, initiating them, and recovering from errors once a plan has begun (i.e., using feedback efficiently). Moreover, patients sometimes have problems when their behavior is interrupted: They appear to forget what they had been formerly doing even after short delays.

Such behavior can be elicited by formal tests. Several investigators demonstrated that schizophrenic patients have deficits on the WCST of set shifting, response to feedback, and abstraction (15). Patients seem to have both difficulty abstracting concepts and make perseverate responses to incorrect responses. Shallice et al. (67) stressed the consistency of frontal deficits in their detailed analyses of single cases, as most patients in their series displayed difficulties generating rules to the WCST or solving Tower of Hanoi type puzzles.

Patients with schizophrenia have difficulty on the Brown-Petersen test in which words have to be remembered over short delays during which overt rehearsal is prevented (i.e., behavior is interrupted). Interference conditions that imposed even moderate processing loads caused a sharp decline in Brown Petersen performance in schizophrenic patients, suggesting that coordination or allocation of resources is deficient or that resource capacity itself is limited (17). Interestingly, Brown-Petersen performance has been correlated to a test of executive function as well as short-term memory and anterograde memory. As such, it might be a test that demands working with information in various memory stores. In an elegant study that also emphasized the role of delay in information control in schizophrenia, Cohen and Servan-Schreiber (9) found that patients had difficulty understanding sentences in which they (the patients) had to retain information over short periods to resolve ambiguity (by comprehending semantic context). The authors went on to note that patients may have difficulty maintaining or using the internal representation of context to control action.

One construct that attempts to capture this type of processing involves notions of working memory. Because working memory theoretically involves the simultaneous storage and processing of information, it is presumed that a central executive component partitions or allocates limited resources during complex, novel, or dual tasks and deploys additional cognitive resources to aid in maintenance of material in some short-term store or computational workspace.

There is evidence that the Wisconsin Card Sort might involve the working memory system. For instance, Sullivan et al. (72) found that WCST perseveration was strongly associated with other tests that are thought to require working memory including self-ordered pointing (in which a subject monitors his or her own series of responses). A second line of evidence comes from attempts to teach patients with chronic schizophrenia how to do the WCST. In one such study, Goldberg et al. (32) provided incremental instructions about categorization and set shifting, followed by intensive card-by-card teaching in an effort to ensure that patients maintained an optimal test taking set. Goldberg et al. found that incremental information did not aid patients (the performance was poor), but, when card-by-card teaching was withdrawn and normal testing procedures were resumed, patients' performance returned to baseline. These results suggested that psychological factors were not the primary explanation of the patients' difficulties, but rather failures were manifestations of unique task-specific computational systems. In particular, it appeared that patients could not maintain information in some working memory store to control responses. A number of other studies have also demonstrated that it is exceedingly difficult to normalize patients performance on the WCST, although not necessarily to improve it (32).

Another study also argues for a deficit in working memory's visual scratchpad. Using an ocular motor-delayed response paradigm that was developed by Goldman-Rakic and colleagues for use in primates, Park and Holzman (57) found that patients with schizophrenia had grave difficulties maintaining information for location over a brief delay in which they had to perform an interference task. Although interpretation of results in this particular study was complicated because patients also had difficulty on a control task that did not involve a delay, the corpus of the work suggests that schizophrenic patients have difficulty maintaining and transforming information over short delays in the service of a response.

Abnormal voluntary and involuntary movements have long been recognized as common in schizophrenic patients. With the introduction of neuroleptic medications and the apparent link between such medications and the development of tardive dyskinesia, motor abnormalities have not been widely studied (with the exception of eye movements), perhaps out of concern that such symptoms are treatment-induced rather than inherent features of the illness. However, this being said, there is a large, consistent literature suggesting abnormal motor function in schizophrenia. Patients with schizophrenia tend to be slow in initiating movements as demonstrated in prolonged reaction time. Repetitive movements are carried out slowly (as in finger tapping) and this slowing is often exaggerated by the complexity of the required motor act. The abnormality is not simply one of the psychomotor retardation. When patients perform a simple motor act, such as button pressing, their physical effort has been characterized as abnormally discontinuous and irregular (75). There is direct electrophysiological evidence that the voluntary preparation processing typically seen in normal subjects prior to the initiation of a motor act is delayed in onset in patients (69).

Unsurprisingly, more complex forms of motor performance that involve continuous adjustment to feedback are also impaired in patients. Thus patients have been reported to have remarkable difficulties monitoring their own actions and often fail to correct errors (49). It has been suggested that this failure to monitor action execution represents a fundamental deficit in schizophrenia and may be one of the mechanisms implicated in the genesis of psychotic symptoms (18). The motor abnormalities extend beyond slowing and lack of normal refinement of action sequences; there are reports of gross perseverative behavior in patients with schizophrenia when performing graphomotor tasks (3). Thus, many of the themes that are well developed in the literature on more complex cognitive functions in schizophrenia can be traced to the purely motor domain. Indeed, this may suggest that many of the abnormalities in schizophrenia are primarily located at the stage of response preparation, execution, and monitoring, abnormalities that are evident in any modality and at any level of processing complexity where responses are not fully specified by the imperative stimulus involved in task presentation.

The motor abnormality most thoroughly documented in schizophrenia has involved the oculomotor system. More than 80 years ago, Difendorf and Dodge reported impaired eye tracking in schizophrenic patients. This observation has been replicated numerous times in recent years and has been comprehensively reviewed by Levy et al. (46). When engaged in pursuit of an object, patients with schizophrenia demonstrate several abnormalities including impaired gain. That is, they fail to match the velocity of the target. They also demonstrate the intrusion of saccadic movements into what should be smooth tracking. Abnormal tracking has been observed in both chronically ill and first-episode samples. It is a matter of dispute whether there is any effect of illness chronicity or severity on tracking performance. Neuroleptic medications do not appear to have a significant effect on eye tracking, although other psychotropic medications, such as lithium, may impair tracking. Several independent groups have reported an increased frequency of abnormal tracking in the first-degree relatives of patients, suggesting that eye movement abnormalities may be a marker of vulnerability to schizophrenia. Poor smooth pursuit, and poor performance in an antisaccade task has been associated with poor performance on neuropsychological tests of frontal cortical function by several investigators, suggesting that the eye movement deficits of patients may be another manifestation of frontal pathology (46).

Mapping cognitive performance onto neuroanatomical measures has met with mixed results. It is an important endeavor because positive associations suggest that cognitive impairments do not arise de novo, but may have a specific or circumscribed neurobiological substrate. Inconsistent findings may have been due in part to difficulties in ascertaining precisely to what degree each measure deviates from its genetically and environmentally determined potential level. Goldberg et al. (29) attempted to surmount this problem by examining difference scores derived from the unaffected member and affected member of a twin pair, reasoning that such scores should represent the degree of pathological involvement irrespective of actual level. In correlating intrapair difference scores of anatomical structures measured from MRI with cognitive abilities (after partialling IQ), they found strong associations between the left hippocampus and a parameter of verbal memory.

Despite the theoretical power of this technique, there was a general paucity of significant correlations. In contrast, Goldberg et al. found a wider array of significant association between some cognitive performances and a neurophysiological measure, prefrontal rCBF, assayed during administration of the Wisconsin Card Sorting Test. In particular, correlations between perseveration and a prefrontal rCBF were robust (>0.60). These results suggest that (a) physiological measures may be better measures of the functional integrity of a neural system than anatomical volumes and (b) the restriction of range inherent in both cognitive and anatomical measures compared to a neurodegenerative disease like Alzheimer's, may limit positive findings. These results, although limited, do support other research implicating medial temporal and prefrontal regions as important in the symptomatic expression and cognitive failures of schizophrenia.

Delineating the characteristic neurocognitive profile and global level in schizophrenia has implications for (a) weighing the fundamental validity of accumulated neuropsychological findings, (b) inferring neuroanatomical and neurophysiological dysfunction, (c) identifying measures to be used in high-risk or linkage studies, and (d) providing targets for rehabilitative efforts. Recent evidence suggests that relative neuropsychological specificity may be present in schizophrenia. This has not always been viewed as the case, as cluster analytical techniques delineated groups of psychiatric patients with unique profiles of cognitive impairments that were independent of diagnosis (74). However, these studies did not always use DSM-III diagnoses and did not include key neurocognitive measures that tapped memory and set shifting.

Two obvious groups of patients to contrast with schizophrenic patients are those with unipolar and bipolar mood disorders. Differences in neuropsychological performance between patients with schizophrenia and patients with mood disorder might account for some of the differences in outcome between these disorders. Schizophrenia is characterized by a chronic course with much associated morbidity, whereas mood disorder is thought to be more episodic and less disabling. If cognitive impairment is frequent in schizophrenia and accounts for some proportion of social and vocational disability, it would be expected at the very least that patients with schizophrenia should have more severe deficits on neuropsychological tests. In a recent study, Goldberg et al. (25) found that patients with schizophrenia performed significantly and consistently below unipolar and bipolar affective patients on test of attention, psychomotor speed, verbal and visual memory, and problem solving and abstraction. Also, IQ was lower in the schizophrenic group and appeared to have deteriorated from normal premorbid levels. Such a decline was not apparent in the mood-disordered groups. When IQ was controlled, differences in problem solving on the WCST and visual memory for design remained. These results were consistent with outcome data and also suggested prominent executive and mnemonic deficits in the schizophrenic group against a background of more generalized impairment.

Various neurological groups have been considered to be "mock-ups" of schizophrenia. Temporal lobe epilepsy, for instance, is thought to have neurodevelopmental origins, and left temporal lobe epilepsy is thought to be associated with an increased incidence of psychosis. When directly compared to patients with left temporal lobe epilepsy and right temporal lobe epilepsy, Gold et al. (20) found that patients with schizophrenia exhibited superior semantic memory relative to the left temporal group and more severe attentional and problem-solving deficits relative to both the right and left epilepsy groups. Memory performance, especially in the verbal domain and including immediate and delayed recall appeared to fall between the two epilepsy groups. Thus a unilateral (left) temporal lobe model of schizophrenia does not appear to be consonant with the neurocognitive data.

Subcortical dementias also share certain clinical features, including bradyphrenia and motor abnormalities, with schizophrenia. A group with subcortical dementia comprised of patients with Huntington's disease was compared to patients with schizophrenia (23). In this study, the groups were matched on WCST perseveration. Both groups also had equivalent full scale IQs. Patients with schizophrenia exhibited better perceptual organization on the WAIS-R than did patients with Huntington's, while concomitantly exhibiting slightly better verbal and poorer attentional abilities. A double dissociation at the neurophysiological level was also observed. Patients with schizophrenia exhibited reduced prefrontal and augmented parietal flow, whereas patients with Huntington's disease exhibited the reverse of this pattern upon rCBF testing with WCST cognitive activation.

There is increasing evidence that neuropsychological impairment is associated with deficits in social and vocational functioning and with long-term outcome. For instance, Goldberg et al. (31) found that intrapair differences between neuropsychological test scores of the unaffected and affected monozygotic twins discordant for schizophrenia correlated with intrapair differences in social and vocational functioning on the Global Assessment Scale. In particular, IQ, memory for stories, fluency, and WCST perseveration were significantly associated with global level of functioning. These results suggests that patients' deficits in learning new information, purposefully recalling old information, and generating novel plans or hypotheses may have an impact on their capacity to efficiently perform a job, take part in social transactions, and make decisions. Moreover, using intrapair differences in twins concordant for schizophrenia, Goldberg et al. (31) observed that nearly all of the variance on the Global Assessment Scale could be accounted for by differences in the performance of four neuropsychological variables. In this design, the experience of illness, institutionalization, medication, psychotic symptomatology, and, of course, genome are shared.

Although in one sense the design "stacks the deck" because of its artificiality, it does illustrate the often underappreciated importance of neurocognition in predicting level of functioning. The aforementioned study of schizophrenic patients' response to clozapine (27) also is relevant. The investigators found that patients' living arrangement and job status did not change significantly with clozapine treatment, although patients' symptoms were much reduced. Because neuropsychological impairments remained substantial and constant, they may have been rate limiting in the rehabilitation on this group of patients. In another study long-term inpatients could be discriminated from patients who lived outside the hospital on the basis of memory, attention, and motor tests (58). This is not to say that symptoms do not have an impact on social and vocational outcome; they do, at least in the short term. What is important to note is that cognitive impairment may also contribute in a unique manner to outcome as well.

Based on the preceding arguments that cognitive impairment is an enduring and frequent aspect of schizophrenia, it would appear reasonable to target these deficits for treatment. This being said, few pharmacological studies have set out to directly improve cognition in schizophrenia (as opposed to improving it by ameliorating symptoms). In the first study, Fields et al. (16) used clonidine, an a2-adrenergic agonist, to increase memory performance in schizophrenia. Clonidine had been previously shown to improve memory in amnesics. After approximately 3 weeks of treatment with the drug, patients, who were otherwise unmedicated, scored higher on memory quotients and delayed recall scores than they had while receiving placebo. This change occurred independently of changes in severity of psychosis. However, the exact mechanisms responsible for the improvement were obscure, as clonidine may act as a direct postsynaptic agonist or exert presynaptic inhibition. In another study, Goldberg et al. (24) used dextroamphetamine as a cognitive enhancer. The rationale was as follows. Cognitive failure and psychiatric symptoms might be the result of an imbalance between dopamine systems, the latter the result of dopamine overactivity in limbic regions, the former the result of dopamine underactivity in prefrontal cortical regions. With this in mind and capitalizing on differences in distribution of dopamine type I and type II receptors, patients were maintained on chronic haloperidol treatment to blockade dopamine type II receptors in subcortical regions and administered an oral dose of amphetamine (at 0.25 mg/kg) to stimulate cortical type I receptors. On cognitive measures, patients demonstrated significant improvement in psychomotor speed and a near-significant improvement on concept formation on the WCST. The latter finding was consistent with changes in regional cerebral blood flow in dorsolateral prefrontal cortex as detected by SPECT (12). In fact, increases in prefrontal flow were correlated with improvements in the Card Sort.

The complexity of cognitive enhancement in schizophrenia was recently demonstrated by Bilder et al. (4) who found that verbal fluency was reduced by neuroleptic treatment relative to placebo, but that the impact of methylphenidate infusion differed as a function of treatment status. Performance was enhanced in patients on neuroleptic and decreased in neuroleptic-free patients to whom it was administered. Bilder et al. suggested that there is an optimal level of dopamine tone for cognitive function; both too little and too much dopaminergic activity have deleterious effects.

In the lay imagination, schizophrenic patients experience problems in living because they are divided against themselves, out of touch with reality, and disorganized. The view of scientists, once not altogether different, has begun to change. Not only have the symptoms been defined and codified, but the neurobiological underpinnings of the disorder have begun to be described. Emerging also is a view in which cognitive impairments may be a relatively central feature of the disorder. Why central? First, they seem enduring in that they are dissociable from psychiatric symptomatology. Second, they are very frequent: most patients exhibit cognitive impairments to a greater or lesser degree. Third, cognitive impairments also may have a relatively well-delineated profile in which executive, memory, and attentional deficits are prominent and are associated with social and vocational disability. This account carries with it implications for treatment in that cognitive impairments should also be considered target symptoms in the way that hallucinations, delusions, and anergia are. That is, drugs or a combination of drugs in conjunction with various cognitive rehabilitational techniques should be marshalled against the impairments, because certain cognitive functions are necessary for even rudimentary successes in modern life.

published 2000