Premorbid Cognitive Deficits in Young Relatives of Schizophrenia Patients

Neurocognitive deficits in schizophrenia (SZ) are thought to be stable trait markers that predate the illness and manifest in relatives of patients. Adolescence is the age of maximum vulnerability to the onset of SZ and may be an opportune “window” to observe neurocognitive impairments close to but prior to the onset of psychosis. We reviewed the extant studies assessing neurocognitive deficits in young relatives at high risk (HR) for SZ and their relation to brain structural alterations. We also provide some additional data pertaining to the relation of these deficits to psychopathology and brain structural alterations from the Pittsburgh Risk Evaluation Program (PREP). Cognitive deficits are noted in the HR population, which are more severe in first-degree relatives compared to second-degree relatives and primarily involve psychomotor speed, memory, attention, reasoning, and social-cognition. Reduced general intelligence is also noted, although its relationship to these specific domains is underexplored. Premorbid cognitive deficits may be related to brain structural and functional abnormalities, underlining the neurobiological basis of this illness. Cognitive impairments might predict later emergence of psychopathology in at-risk subjects and may be targets of early remediation and preventive strategies. Although evidence for neurocognitive deficits in young relatives abounds, further studies on their structural underpinnings and on their candidate status as endophenotypes are needed.


INTRODUCTION
Schizophrenia (SZ) was originally described over a century ago with the earlier name "dementia praecox," which literally means "cognitive decline with onset in youth." Cognitive impairment is highly prevalent in patients with SZ as determined by the majority of patients who show cognitive decrement relative to parental education (Keefe et al., 2005) or to their own estimate of premorbid intelligence measured by single word identifi cation (Kremen et al., 1995). Meta-analyses show that cognitive impairment distinguishes patients with SZ from healthy comparison subjects to a robust degree (i.e., an effect size of approximately one with approximately one standard deviation); these defi cits are apparent at the fi rst episode and roughly are equal to those observed in chronic cases (Mesholam-Gately et al., 2009). Average effect sizes for cognitive impairments in SZ are about twice as large as those obtained in structured magnetic resonance imaging studies (Heinrichs, 2005). Cognitive impairment is a stable, trait-related aspect of SZ, being present in the early phase of the illness and persisting during the long-term course (Rund, 1998). Cognitive impairment is a predictor of social and vocational outcome as evaluated longitudinally . Recent studies suggest that social cognition may have a particularly strong relation to functional outcome . Finally, cognitive impairment may differ to some extent between SZ and other psychiatric disorders (MacDonald et al., 2005). Cognitive adulthood. These alterations may also represent endophenotypes (i.e. markers intermediate between phenotypic manifestations of the disease and the genotype) (Gottesman and Gould, 2003).
In this paper, we review studies that have examined various cognitive domains including attention, learning and memory, general intelligence, social-cognition, speed of processing and executivefunction (Henry and Crawford, 2005) in unaffected young relatives presumed to be at high genetic risk. We did an extensive PubMed search using keywords "schizophrenia", "relatives" and "cognition." In particular, several high-risk studies conducted over the last three decades were reviewed ( Table 1). We also summarize fi ndings from our ongoing studies related to neurocognition, as well as provide some additional data on the nature of cognitive defi cits and their relation to neurobiological alterations as well as the dose of familial risk (fi rst vs. second-degree relatives) in young relatives at risk for SZ.

NATURE OF NEUROCOGNITIVE DEFICITS IN YOUNG RELATIVES AT RISK FOR SZ
There are prominent impairments in SZ in several domains of cognition, including psychomotor speed, memory, attention, reasoning, and social cognition ( Table 1). These may be easily remembered by the mnemonic SMART [Speed of processing, Memory, Attention, Reasoning and Tact (or social cognition)]. Studies have suggested some inter-correlation between cognitive performance on these domains, although there is no clear consensus regards the degree of shared variance across domains. Studies have shown both signifi cant (Dodrill, 1997) and non-signifi cant correlations between these domains  in patients with SZ. A common "general intelligence" factor, correlating with all domains may explain the lack of independence of cognitive performance on these domains. Evidence suggests correlation of IQ (an index of general intelligence) with performance across domains in SZ patients (Bell and Roper, 1998;Tremont et al., 1998;Horton, 1999;Jung et al., 2000;Kremen et al., 2008) and may represent this common general intelligence factor. Alternatively, the inter-dependence of specifi c cognitive domains could be due to similarities across the different neuropsychological tests used to assess different domains (Larrabee, 2000). Although IQ defi cits generally share variance with specifi c cognitive defi cits, defi cits in some domains such as speed of processing and verbal memory have been found to be independent of the IQ defi cits. It is therefore unclear if domain-specifi c defi cits can be fully accounted for by a super-ordinate factor like general intelligence.

SPEED OF PROCESSING
This domain measures cognitive effi ciency and involves the ability to automatically and fl uently perform relatively easy or repetitive cognitive tasks. Shakow (1963) originally described this defi cit in SZ studying reaction time slowing. Speed of processing has been posited as a predictor of global functioning, autonomy, self care and hence of illness outcome and quality of living (Sánchez et al., 2009). Reaction time, an indicator of speed of processing, is increased in relatives of patients (Birkett et al., 2007). Young relatives at risk for SZ have reduced processing speed even after controlling for IQ as shown by the Edinburgh High-Risk Study (EHRS) (p = 0.044) (Byrne et al., 2003;Cunningham Owens and Johnstone, 2006;O'Connor et al., 2009), as well as our studies which will be described later. These defi cits might be state-independent given that psychotic symptoms do not alter the severity of speed of processing defi cits in patients (O'Connor et al., 2009). Evidence suggests that processing speed may depend on testing conditions. In a study with varying cognitive processing loads, while the fastest reaction times (that happen during low-cognitive load tasks) were not increased in relatives (Birkett et al., 2007), mean reaction time was slower, suggesting slower reaction times during high-cognitive-load tasks. Slowed performance on the various psychomotor measures has been shown to be independent of medication (Morrens et al., 2007). Speed of processing has also been found to predict negative symptoms and impaired functional outcomes (Niendam et al., 2006;Morrens et al., 2007).
A genetic-load effect is noted with performance of relatives being intermediate between that of healthy controls and patients on this domain (Birkett et al., 2007;Gur et al., 2007b;Bertisch et al., 2009). Reaction time has been proposed as a putative endophenotype of the illness . However, a study showed that both patients with and without reaction time defi cits on the Continuous Performance Test (CPT) have relatives showing these defi cits (Birkett et al., 2007). Thus, the candidacy of reaction time defi cits as an endophenotype, as well as their role as premorbid vulnerability indicators deserve further consideration.

MEMORY
Working memory (WM) involves holding information online for brief periods of time, and typically involves processes like information manipulation, maintenance and monitoring in verbal, visual and spatial domains (Kellogg et al., 2007). Maintenance involves retaining information in a sequential manner, manipulation deals with rearrangement of the information sequence while monitoring checks and updates the contents of WM to determine the next step in a sequential task. WM in all subsystems is impaired in fi rst episode patients (Zanello et al., 2009) and unaffected fi rst degree relatives (Conklin et al., 2005;Saperstein et al., 2006;Horan et al., 2008). Relatives of SZ patients perform poorly on spatial WM (Awh et al., 1998;Saperstein et al., 2006) and spatial memory capacity (O'Connor et al., 2009). Several studies report impairments in verbal, spatial and object WM domains with a graded pattern of impairment; defi cits in patients > relatives > controls are observed for verbal WM (Niendam et al., 2003;Conklin et al., 2005). Defi cits in WM appear to correlate with negative symptoms (Chkonia and Tsverava, 2007). Impaired WM has been proposed as a putative endophenotype for SZ (Niendam et al., 2003).
Verbal declarative or long-term memory is signifi cantly reduced in patients (Chkonia and Tsverava, 2007), is associated with earlier disease onset, is related to social functioning and negative symptoms (Niendam et al., 2006), and is proposed to be a predictor of later SZ in high-risk individuals (Niemi et al., 2003;Groom et al., 2007) (Table 1). However, a study reported no verbal-memory deficits in high-risk offspring after controlling for education (Chkonia and Tsverava, 2007). In the EHRS, defi cits in Rey's auditory verbal learning test predicted later SZ but defi cits in Rivermead Behavioral Memory Test did not (Byrne et al., 1999(Byrne et al., , 2003Cosway et al., 2000;Johnstone et al., 2002Johnstone et al., , 2005Whyte et al., 2006;Whalley et al., 2007). The New York High-Risk Project (NYHRP) reported that verbal  (Seidman et al., 2006) showed that verbal memory impairment may have promise as a premorbid predictive marker in those at genetic risk for the illness, but further investigation is needed into confounding mediator factors such as affective symptomatology, education and environmental factors in these defi cits. Another multi-site study (The Consortium on the Genetics of Schizophrenia) has proposed verbal WM defi cits to be putative inherited endophenotypes of SZ (Greenwood et al., 2007;Horan et al., 2008). In general, verbal memory has been shown to be one of the most robust defi cits in studies of relatives. Visual memory has been studied less frequently than verbal memory in patients, and impairments in the visual domain among family members appear to be somewhat less severe than in the verbal domain (Snitz et al., 2006). A study reported verbal recall defi cits over short and long delays in both patients and relatives of patients but visual recall defi cits only in patients (Heinrichs and Zakzanis, 1998;Whyte et al., 2005;Delawalla et al., 2006). Visual recall defi cits have been thought to be state dependent while verbal memory defi cits may be heritable stable trait markers (Skelley et al., 2008). Visio-spatial memory defi cits in relatives correlate with their proximity to probands (genetic loading) (Robles et al., 2008). Also, visual recall defi cits in delayed recognition tasks have been observed in high-risk relatives (Byrne et al., 1999).

ATTENTION
Attention involves the appropriate allocation of processing resources to relevant stimuli, and includes sub-processes like sustained attention and selective attention. A frequently used test to assess attention-performance is the CPT. Several CPT versions vary with regards to modality (auditory or visual), type of stimulus (letters, numbers, colors, or geometric forms), and nature of the task (Miranda et al., 2008). Attentional abnormalities have been well documented in SZ; attentional defi cits are associated with negative and disorganized symptoms and persist despite treatment. Impaired sustained attention indexed by perceptual sensitivity (d′) in the CPT task strongly discriminates high-risk relatives from healthy controls (Erlenmeyer-Kimling et al., 2000); attention defi cits are consistent, temporally stable, and independent of environmental factors or onset of psychotic symptoms (Freedman et al., 1998;Cornblatt et al., 1999;Erlenmeyer-Kimling et al., 2000). Attention defi cits predicted more than half (58%) of the high-risk offspring who developed SZ in their future (Erlenmeyer-Kimling et al., 2000). Measures of attention deviance predicted social outcomes while poor neurobehavioral functioning predicted future SZ spectrum disorders (Marcus et al., 1987;Erlenmeyer-Kimling et al., 2000). Attention defi cits have been observed in unaffected relatives in the prodromal as well as premorbid phases and have been considered as "endophenotypes" for later emergence of SZ (Cornblatt and Malhotra, 2001), using poor attentional performance as a marker of vulnerability to SZ could provide a valuable measure of genetic risk.
Unaffected relatives of patients show defi cits in emotion recognition (Kee et al., 2004;Bediou et al., 2007;Eack and Mermon, 2009), and theory of mind tasks (Anselmetti et al., 2009). A study on siblings of SZ patients (Leppanen et al., 2008) demonstrated signifi cant performance defi cits in the recognition of facial anger. Recently, one study  found that individuals clinically at high risk (HR) for developing SZ (i.e., those with prodromal symptoms) performed as poorly as fi rst episode patients on an emotion identifi cation task. Theory of mind defi cits also have been shown to be compromised in relatives and together with emotion perception may predict functioning in the community (Irani et al., 2006;Marjoram et al., 2006;Pijnenborg et al., 2009). As reviewed in Phillips and Seidman (2008), emotion perception deficits in relatives are consistently present, as well as social anhedonia and negative affect. Some studies have found high-risk offspring to have poor social competence (Goodman, 1987;Marcus et al., 1987;Dworkin et al., 1993).

VERBAL FLUENCY
Language related cognitive defi cits, verbal memory Riley et al., 2000), verbal fl uency Riley et al., 2000), semantic memory (Lorente-Rovira et al., 2007), comprehension (Condray et al., 2002), and receptive language (Condray et al., 2002) are found to be defi cient in patients with SZ and are also present in at-risk children (Keefe et al., 1994;Chen et al., 2000;Weiser et al., 2007). Category verbal fl uency indexes semantic memory, lexical access, and executive function while letter fl uency may index psychomotor speed (Benton and Hamscher, 1978). Although verbal fl uency is shown to be altered in relatives of SZ patients (Bhojraj et al., 2009), few studies have assessed young relatives (Broome et al., 2009). A recent meta-analysis revealed a large effect size (0.68) in category fl uency (Snitz et al., 2006). Verbal fl uency may be signifi cantly correlated with intelligence (Gilvarry et al., 2001); another study reported defi cits in verbal fl uency and executive function among relatives of SZ patients (Keefe et al., 1994). The possibility of verbal fl uency defi cits in young relatives was assessed by the Pittsburgh High-Risk Study (see below) which found signifi cant defi cits at the baseline assessment.

GENERAL INTELLIGENCE
Intelligence defi cits in relatives at risk for SZ are equivocal with studies both showing (Mednick and Schulsinger, 1968;Rieder et al., 1977;Dworkin et al., 1993;Byrne et al., 1999;Goldstein et al., 2000) and not showing signifi cant IQ defi cits (Sameroff et al., 1987). IQ defi cits tend to progress with time as evidenced by some studies (Worland et al., 1982) while others did not fi nd such a pattern (Goodman, 1987;Dworkin et al., 1993). Some studies with HR offspring bearing IQ defi cits predicted adult SZ (Cosway et al., 2000) while others could not (Dworkin et al., 1993). A study reported low social status and severity of maternal illness to be strong predictors of low IQ in offspring of patients (Sameroff et al., 1993). Worland et al. (1982) reported a time by parental diagnosis interaction on verbal IQ among HR offspring, children of mothers with SZ showed more defi cits than children of fathers with SZ during a 16-year follow-up, and also children of SZ parents had the lowest stability on IQ scores. The question of whether the liability to SZ is mainly related to a generalized intellectual defect or whether there exists unique cognitive domains with selectively more prominent impairments remains unclear (Woodberry et al., 2008).
Premorbid cognitive defi cits may map onto observed structural defi cits in brain regions mediating corresponding cognitions. Relations between cognitive defi cits and brain structural alterations in high-risk relatives have not been systematically examined. If such relations are established, cognitive and brain structural defi cits, both considered to be endophenotypes of SZ, might be more parsimoniously explained by the "extended endophenotype" concept (Kippenhan et al., 2005).

NEUROCOGNITIVE DEFICITS IN YOUNG HR RELATIVES: FINDINGS FROM THE PITTSBURGH HIGH-RISK STUDY
In an ongoing longitudinal study, the Pittsburgh Risk Evaluation Program (PREP), we assess young (10-25 years) fi rst-and seconddegree relatives of SZ probands and healthy controls. The participants were identifi ed at the Western Psychiatric Institute and Clinic (WPIC), Pittsburgh or related clinical sites. Young HR relatives were recruited by fi rst approaching patients with SZ with eligible relatives in our outpatient clinical services and via advertisements in community locations. Participants were included if they had a fi rst or second degree relative with SZ or schizoaffective disorder, had an IQ ≥ 80, did not have any lifetime evidence of psychotic disorders, antipsychotic medication exposure, history of substance use, and neurological or medical condition. Age and gender matched healthy controls were recruited from the same community neighborhoods as HR subjects. The study design, demographic, and clinical characteristics of these subjects have been described elsewhere . We report herein summary observations in key neurocognitive domains and their neuroimaging correlates.
Previously published fi ndings from the PREP study involve defi cits in memory, attention, verbal fl uency, executive function, social cognition, and general intelligence. High-risk offspring performed poorer compared to controls on spatial WM, sustained attention, category verbal fl uency (Eack et al., 2008), executive function (Keshavan et al., , 2005Eack et al., 2008), and general intelligence (Eack et al., 2008). Social cognition defi cits in facial emotion recognition were also noted (Eack and Mermon, 2009). Relatives were found to over-attribute negative valence to neutral faces and took longer to identify neutral faces. These defi cits were independent of other neurocognitive dysfunction and correlated with positive symptoms and general psychopathology scores . Compared to healthy controls, relatives of SZ patients were more prone to develop attention defi cit hyperkinetic disorder  and schizotypal personality traits. Using a multivariate psychobiological prediction model comprised of neuroimaging, neurocognitive, and psychosis proneness measures, these variables together predicted 71% chance to develop psychopathology, in contrast to individuals not identifi ed to develop psychopathology by the model who only had a 17% chance of developing psychopathology (Eack et al., 2008). In this review, we provide additional data on (a) neurocognitive fi ndings and their familial dose effects and (b) brain structural correlates of neurocognitive defi cits in young relatives at risk for SZ.

NEUROCOGNITIVE DEFICITS AND GENETIC DOSE EFFECTS
Neurocognitive scores (measured in parentheses) were collected from a neuropsychological battery including IQ (Wechsler Abbreviated Scale of Intelligence; Wechsler, 1999); WM (Cogtest Spatial Working Memory Test; distance median after a 12-s delay; Cogtest, 2009); executive functioning (Wisconsin Card Sorting Test perseverative error scores; Heaton et al., 1993); attention (Continuous Performance Test, IP version visual d prime; Cornblatt et al., 1988); and verbal fl uency (Benton and Hamscher total correct from the category/letter fl uency task; Benton and Hamscher, 1978). Table 2 denotes defi cits seen in fi rst-and second-degree relatives (HR) compared to HC controlling for age at baseline assessment of the PREP study. Signifi cant defi cits were noted in HR in IQ (p < 0.000). Higher order cognitive domains like executive function and spatial-WM  were not as prominently affected in HR as were simpler domains such as psychomotor speed, sustained attention, and verbal fl uency. Defi cits in both attention and spatial WM were attenuated and those in verbal fl uency lost signifi cance after controlling for psychomotor speed, suggesting that higher order cognitive defi cits may be mediated by defi cits in hierarchically more basic cognitive processes such as speed of processing (Nuechterlein et al., 2004). We assessed familial-loading effects by comparing groups of fi rstdegree relatives (n = 122), second-degree relatives (n = 23) and healthy controls (n = 109) using ANCOVA models. Familial-loading effects were seen at p < 0.05 for psychomotor speed (F = 5.89, p = 0.043), executive-function (F = 4.56, p = 0.05) and verbal-fl uency (F = 3.91 p = 0.078) with fi rst-degree relatives performing poorer than second-degree relatives on all domains except WM. Figure 1 shows that fi rst-degree relatives have the more prominent defi cits, while second-degree relatives have impairment intermediate to that of fi rst-degree relatives and healthy controls in all domains except WM. No moderating effects of gender on the main effect of study group (HR vs. HC) were noted.
As the exact relation of IQ defi cits with domain specifi c defi cits is unclear, we conducted parallel analyses controlling, as well as not controlling for IQ. Defi cits in sustained attention (F = 5.1, p = 0.025), speed of processing (F = 5.2, p = 0.023), and verbal fl uency (6.2, p = 0.011) in relatives survived controlling for IQ. All neurocognitive scores were signifi cantly correlated with IQ (r ranging from 0.30 to 0.43). Studies in patients have shown most neurocognitive defi cits, except for psychomotor speed and verbal memory, to be mediated by a latent "cognitive ability factor" (Weickert et al., 2000;Dickinson et al., 2008). This agrees with fi ndings of attention and verbal fl uency defi cits but not psychomotor speed defi cits losing signifi cance after controlling for IQ in the PREP study. A latent cognitive ability factor as underpinning all neurocognitive defi cits is debatable as the latent factor was revealed using a correlation method in a cross-sectional design (Dodrill, 1997(Dodrill, , 1999Bell and Roper, 1998;Tremont et al., 1998;Horton, 1999;Jung and Haier, 2007;Dickinson et al., 2008). Longitudinal studies have shown cognitive defi cits to precede generalized cognitive defi cits like IQ (Weickert et al., 2000). Controlling for IQ when assessing cognitive defi cits may be unnecessarily conservative, especially given the equivocal evidence about the role of IQ in cognitive defi cits (Dickinson et al., 2008). IQ defi cit may be an inherent, natural property of subjects at genetic risk instead of a confound and hence controlling it may have the effect of throwing the baby out with the bathwater (Miller and Chapman, 2001). Also, correlations between a dependent variable and a putative confound argue against controlling for that confound as it may obscure real group differences of the dependent variables (Miller and Chapman, 2001).
The spatial-working-memory defi cits noted at the 12-s delay were absent for a 2-s delay. This supports previous evidence suggesting a task diffi culty by group interaction when comparing SZ  (Callicott et al., 1998(Callicott et al., , 2000Tan et al., 2006). Longitudinal neurocognitive assessments are needed to explore further temporal decline in attention, verbal fl uency, and psychomotor to detect a possible emergence of executive function and spatial WM defi cits.

BRAIN STRUCTURAL CORRELATES OF NEUROCOGNITIVE DEFICITS
The Pittsburgh High-Risk Study also involved a structural brainimaging component. Relatives were categorized into low cognitive scoring and high cognitive scoring groups based on verbal fl uency, attention, psychomotor speed, and executive-function scores using K-means cluster analysis. This method is an iterative procedure, which clusters cases into two groups. The iterations seek to minimize within cluster variance and maximize variability between clusters in an ANOVA-like fashion. Brain regions involved in these cognitions and implicated in SZ were compared across low and high scoring clusters of relatives. As seen in Figure 2, the low scoring subset of relatives (n = 59) had lower volumes in critical brain regions compared to the high scoring subset (n = 35), with the exception of the middle frontal gyrus. Relatives of patients show cognitive defi cits that cooccur with alterations of regions mediating these compromised cognitions. This association may tentatively suggest structural alterations to underpin cognitive defi cits seen in relatives. Brain regional abnormalities with their "downstream" attendant cognitive defi cits may together be considered as "extended endophenotypes", a parsimonious conceptualization of SZ .
In summary, fi ndings from the PREP study are consistent with previous reports of cognitive defi cits in relatives of SZ patients and suggest that these defi cits may be related to neuroanatomical defi cits of corresponding brain regions. The existence of distinct subgroups of low and high cognitive scoring subjects within the sample of relatives is a critical fi nding from the PREP study. The clustering of structural alterations within the low-scoring subgroup tentatively suggests a neuroanatomically and cognitively compromised "hypervulnerable" subset within relatives with a familial diathesis for SZ. The risk of SZ and SZ spectrum disorders in genetically liable relatives of patients is 11-15% and about 40%  respectively. This further suggests a heterogeneous risk-profi le of the genetically vulnerable population for future psychotic illness and the occurrence of "hypervulnerable" subgroups . The latent genetic heterogeneity in SZ explains the existence of these subgroups rather than a uniform vulnerability for SZ within genetically predisposed populations Eack et al., 2008).

CONCLUSIONS
In summary, cognitive defi cits are a core feature of the premorbid vulnerability to SZ. Impairments are seen in several cognitive domains in unaffected relatives of patients including attention, WM, verbal memory, visual memory, executive function, speed of information processing, social cognition, and general intelligence. In general, the abnormalities appear more severe in fi rstdegree relatives, and are associated with more prominent brain structural alterations. These observations are of clinical as well as pathophysiological signifi cance.
An important question of clinical relevance is whether premorbid cognitive deficits can predict the emergence of later SZ in non-symptomatic at-risk subjects. As reviewed, the NYHRP and EHRS studies suggest that deficits in memory, attention, and social cognition in young relatives of SZ patients may predict later psychosis. Attention deficits in young relatives of SZ patients frequently have features of attention deficit disorder . This often leads to the clinical practice of treating such individuals with stimulant medications, which may have the undesirable effect of triggering psychosis in these vulnerable individuals. It is important to distinguish attentional impairments that are the precursors of a serious illness such as SZ and treat them with the disease appropriate interventions. Thus, children and adolescents newly presenting with attentional impairments should not, as often happens, be automatically diagnosed as having attention deficit disorders, but should be assessed to rule out early features of SZ (such as prodromal symptoms and schizotypy) or bipolar disorder (mood dysregulation). Inquiring for family histories of major psychiatric disorders is also important. Investigating premorbid neurocognitive deficits is also of importance for early intervention. FIGURE 2 | The low scoring subset of relatives had volumetric defi cits in critical brain regions compared to the high scoring subset. Regional gray-matter-volumes (right and left combined) for each group were z-transformed to the control mean. Group-means of the z-scores are plotted on the y-axis in the low-scoring and the high-scoring groups (see text for description of approach to this classifi cation). HG, Heschl's gyrus; SMG, supramarginal gyrus; MFG, middle frontal gyrus; IFG, inferior frontal gyrus; Hippo, hippocampus; CN, caudate nucleus.
Further research is needed to evaluate the effi cacy of cognitive remediation approaches, shown to benefi t early phases of SZ (Eack et al., 2007), in at-risk individuals with cognitive defi cits. Pharmacological interventions, including low dose atypical antipsychotics, have also been piloted in cognitively impaired relatives at risk for SZ .
Cognitive defi cits, being core impairments in the premorbid phase of SZ, offer the best way to defi ne the neurobiology of the vulnerability to this illness. As reviewed in this paper, cognitive defi cits are robust, highly prevalent, stable, easily quantifi able, correlate with defi ned biological alterations in the illness, and are present in both those with the illness and those at risk. These features qualify cognitive impairments as endo-(or intermediate) phenotypes, which are beginning to pave the way to identifi cation of the susceptibility gene(s) (Gur et al., 2007c).

ACKNOWLEDGMENTS
This work was supported by NIMH grants MH 64023, MH 01180 and a NARSAD Established Investigator award (MSK). We thank Drs. Konasale Prasad and R. P. Rajarethinam who helped with research design issues; Diana Mermon who did the assessments; and Jean Miewald who helped with the data management and analyses.