Schizophrenia Research 93 (2007) 325 – 333 www.elsevier.com/locate/schres
Factor analytic support for social cognition as a separable cognitive domain in schizophrenia Daniel N. Allen a,⁎, Gregory P. Strauss a , Brad Donohue a , Daniel P. van Kammen b a
Department of Psychology, University of Nevada, Las Vegas, 4505 Maryland Parkway, Las Vegas, Nevada, 89154-5030, USA b ACADIA Pharmaceuticals Inc., San Diego, CA 92121, USA Received 18 August 2006; received in revised form 10 February 2007; accepted 16 February 2007 Available online 10 May 2007
Abstract Social cognition has received increasing attention in schizophrenia due to its theoretical relevance to core features of the disorder as well as the marked deficits in social functioning exhibited by these patients. However, there remains a need to develop and validate measures of social cognitive abilities and to demonstrate that they are constructs that are separable from non-social neurocognitive processes. In the current study, the Wechsler Adult Intelligence Scale-Revised (WAIS-R) was administered to 169 males with schizophrenia, and test results were subjected to confirmatory factor analysis (CFA) to determine if those WAIS-R subtests containing social content would form a distinct Social Cognition (SC) factor. CFA was used to evaluate various models that hypothesized an SC factor, and for comparison purposes the same models were evaluated in the WAIS-R standardization sample. Results confirmed the presence of a four-factor model that included an SC factor, as well as the more commonly reported Verbal Comprehension, Perceptual Organization, and Working Memory factors. The SC factor consisted of the Picture Arrangement and Picture Completion subtests, and demonstrated small but significant correlations with disorganization and negative symptoms, as well as with an index of social functioning. Results provide support for the validity of the SC factor as a measure of social cognition in schizophrenia, and demonstrate that at least some aspects of social cognition represent separable cognitive domains in schizophrenia. © 2007 Published by Elsevier B.V. Keywords: Schizophrenia; Social cognition; Confirmatory factor analysis; Intelligence; Latent structure; Symptoms; Function
1. Introduction Schizophrenia is characterized by severe impairment of neurocognitive function with differential impairment of attention, abstraction and problem solving, and learning and memory (Heinrichs and Zakzanis, 1998; Fioravanti et al., 2005; Seaton et al., 2001). More recently, evidence indicates that individuals with schizophrenia demonstrate impairments in social cognition. Social cognition broadly ⁎ Corresponding author. Tel./fax: +1 702 895 1379/+1 702 895 0195. E-mail address:
[email protected] (D.N. Allen). 0920-9964/$ - see front matter © 2007 Published by Elsevier B.V. doi:10.1016/j.schres.2007.02.008
refers to the diverse set of cognitive abilities that are required for social interaction (Fiske and Taylor, 1991; Ostrum, 1984) including the perception, interpretation, and generation of responses necessary to determine “the intentions, dispositions, and behaviors of others” (Green et al., 2005, p. 882; Penn et al., 1997). Some have linked social cognition to core symptoms of schizophrenia (Frith, 1992) while others have demonstrated that it is associated with poorer functional outcomes (Green et al., 2005; Penn et al., 1996). Thus, understanding abnormalities in social cognition is imperative in schizophrenia, and research to date has generally focused on emotional processing (i.e.,
326
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333
aspects relevant to perceiving and generating emotion), social intelligence (i.e., abilities relevant to inferring the intentions and beliefs of others), social perception (i.e., ability to judge roles and rules relevant to social contexts), social knowledge (i.e., awareness of roles, rules, and goals that are relevant to social situations), and attributions (i.e., how one explains the causes of positive and negative outcomes involving social behavior) (see Green et al., 2005, pp. 882–883). Because of its theoretical and practical significance for schizophrenia, social cognition was included in the National Institute of Mental Health — Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognitive battery as one of seven separable neurocognitive domains. Although the MATRICS Neurocognition Committee initially did not identify social cognition as a separable cognitive domain because of a lack of factor analytic support and the absence of well-validated measures in schizophrenia (Nuechterlein et al., 2004), it was included because of the aforementioned theoretical and functional considerations. The test that was selected to represent this domain differs from other measures in the MATRICS battery in a number of important respects (e.g., not extensively validated in schizophrenia). Thus, there remains a clear need to develop new measures to assess social cognition, and to determine whether or not its various subcomponents represent constructs that are separable from more basic (non-social) neurocognitive tests, with factor analysis particularly well-suited to address this issue. In establishing the factorial validity of social cognitive constructs from nonsocial cognitive processes, one approach is that of considering already established and well-validated measures whose items contain social content. The Wechsler Intelligence Scales may prove useful in this regard, not only because they are wellvalidated and commonly used in evaluation of schizophrenia, but also because their subtests assess a variety of cognitive abilities, some which are deficient in schizophrenia (Aylward et al., 1984; Goldberg et al., 1993) and others that contain social content. Indeed, for many years, Comprehension and Picture Arrangement have been used as measures of social reasoning (Rapaport et al., 1968; Schafer, 1948) and used to investigate social cognition in schizophrenia (Lipsitz et al., 1993; Shean et al., 2005). However, it remains unclear whether a combination of these subtests would form a separate social cognition construct. Early exploratory factor analysis demonstrated the WAIS was composed of three factors, namely Verbal Comprehension (VC), Perceptual Organization (PO), and memory or alternatively Working Memory (WM) (Cohen, 1957). These findings have since been replicated in many
investigations of both clinical and non-clinical populations including schizophrenia (Allen et al., 1998; Dickinson et al., 2002). However, despite extensive factor analytic investigation and the apparent social content of some WAIS subtests, to our knowledge there has not been an attempt to identify a WAIS social cognition factor in schizophrenia using confirmatory factor analysis (CFA), although such a factor has been recently identified in autism (Goldstein et al., 2006) and the WAIS-III standardization sample (Allen and Park, 2007). CFA is preferable to exploratory factor analysis, particularly when there are expectations regarding the factor structure of the data, because it allows for testing of specific hypothesis regarding the adequacy of competing factor solutions. The current study investigates the possibility of a social cognition factor in a sample of 169 males with schizophrenia utilizing the 11 subtests from the Wechsler Adult Intelligence Scale—Revised (WAIS-R). The study represents a follow-up to a CFA investigation of the WAIS-R that identified the standard three-factor model in this sample (Allen et al., 1998), with the current goal being that of determining the validity of a fourth Social Cognition factor. It was hypothesized that a Social Cognition factor would be identified in schizophrenia, and CFA was used to test this hypothesis. Additionally, based on theoretical considerations (Frith, 1992) and prior investigations (Shean et al., 2005) it was further hypothesized that the Social Cognition factor would be differentially sensitive to neurocognitive dysfunction in schizophrenia, and have significant associations with negative and disorganization symptoms, as well as with an index of social functioning. 2. Methods Participants included 169 males with schizophrenia whose mean age was 36.2 years (SD = 7.9) and had an average of 12.3 years (SD = 1.9) of education. Participants were inpatients on a schizophrenia treatment and research unit at a Veterans Hospital, and were participating in a clinical research protocol. Diagnosis of schizophrenia was established using structured clinical interviews (SCID or SADS), psychological testing, and all available information from medical and educational records. Thorough physical, neurological, and psychiatric evaluations ruled out the presence of coexisting psychiatric, physical, or neurological disorders, and patients had not used alcohol or illicit drugs during at least the 3 months prior to evaluation. Age at illness onset was 23.3 years (SD = 5.7) with a mean length of illness of 12.7 years (SD = 7.3). When tested, 18 subjects were medication free, 66 were treated with haloperidol alone, 58 with haloperidol
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333
and benztropine, and 22 with combinations of other antipsychotic, antidepressant, or antimanic medications. Ratings of psychosis severity at the time of evaluation using the Bunney–Hamburg Rating Scale (Bunney and Hamburg, 1963) were in the mild range (mean= 5.9; SD = 1.8), indicating that patients tended to be clinically stable and able to participate in testing. More detailed information regarding the sample can be found in Allen et al. (1998). 2.1. Measures and procedure The Wechsler Adult Intelligence Scale—Revised (WAIS-R; Wechsler, 1981) (Wechsler, 1997) was administered by either a clinical psychologist or doctoral intern. Symptoms were assessed using an extended version of the Brief Psychiatric Rating Scale (BPRS; Overall and Gorham, 1962). BPRS items were used to calculate scores representing Negative, Disorganized, and Psychotic symptom domains (Liddle, 1987; Mueser et al., 1997). A Social Functioning score was also calculated by combining the BPRS Social Competence and Loss of Functioning items. Trained raters completed symptom ratings based on a standardized interview with the patient, and observations of behavior on the inpatient unit. Raters participated in reliability rating sessions at least every other week to maintain reliability. The WAIS-R and BPRS were completed within the same week, typically on the same day. 2.2. Data analyses Seven models were examined using CFA including a one-factor model, two different three-factor models, and four different four-factor models. The one-factor model (M1) specified all WAIS-R subtests to load on a single factor. The three-factor models specified Verbal Comprehension (VC), Perceptual Organization (PO), and Working Memory (WM) factors. The Vocabulary, Information, Comprehension, and Similarities subtests were specified to load on the VC factor, the Block Design, Picture Completion, Picture Arrangement, and Object Assembly subtests on the PO factor, and Digit Span and Arithmetic subtests on the WM factor. In the first three-factor model, Digit Symbol was specified to load only on the WM factor (M3). Based on results of our prior investigation, for the second three-factor model, Digit Symbol was specified as a doublet (M3d), loading on both PO and WM factors (Allen et al., 1998). The four-factor models retained the VC, PO, and WM factors but also specified a Social Cognition (SC) factor. For the first two models, Comprehension and Picture Arrangement loaded on the SC factor, with Digit Symbol specified to load either on the
327
WM factor only (M4C), or on both PO and WM factors (M4Cd). These models were based on the long-standing interpretation of Comprehension and Picture Arrangement as measures of social reasoning (Rapaport et al., 1968; Schafer, 1948). For the second two models, Picture Completion and Picture Arrangement loaded on the SC factor, with Digit Symbol loading on the WM factor (M4PC), or on both the PO and WM factor (M4PCd). These models were based on recent CFA work with autism and the WAIS-III standardization samples (Allen and Park, 2007; Goldstein et al., 2006), and separated visual perceptual subtests with social content (Picture Arrangement, Picture Completion) from those with more neutral content (Block Design, Object Assembly). LISREL 8.7 (Jöreskog and Sörbom, 1993) was used to test the models with confirmatory factor analysis (CFA). CFA allows for testing hypothesized models against an actual set of data. The fit between the hypothesized models and the actual data is evaluated using a number of goodness-of-fit statistics. For the present study, we examined the χ2 statistic, goodness-offit index (GFI), adjusted goodness-of-fit index (AGFI), Bentler and Bonnett Normed Fit Index (NFI; Bentler and Bonnett, 1980), Tucker–Lewis Index (T-LI; Tucker and Lewis, 1973), Root Mean Square Error of Approximation (RMSEA), and the Comparative Fit Index (CFI). Because each of these statistics has its own strengths and weaknesses (Bentler, 1990; Bollen, 1990), decisions regarding adequacy of particular models were based on a consideration of all indexes. An adequate fit was indicated by a GFI of .90 and above, an AGFI of 0.80 or above (Cole, 1987). NFI and T-LI at .90 and above were also taken to indicate adequate fit, with values greater than .95 indicating good fit. For the CFI and RMSEA, an excellent fit was indicated by a CFI of .95 or greater, and an RMSEA of less than .05, with RMSEAs between .06 and .08 indicating adequate fit (Jöreskog and Sörbom, 1993). Models for schizophrenia were compared to the 35- to 44-year-old age group from the WAIS-R normative sample, because they most closely approximated the age of our schizophrenia group. The correlation matrix published in the WAIS-R manual was used for the CFA (Wechsler, 1981, p. 41). Determining improvement in model fit was also important for the current study. For nested models, improvement in model fit was determined by the magnitude of differences in χ2 between the various models which provides a statistical test of improvement in model fit (Bentler and Bonnett, 1980; Mulaik et al., 1989). To evaluate its validity, analyses were conducted to determine if the SC factor score would be significantly lower than the other factor scores. Such a finding would suggest the SC factor is sensitive to social cognitive deficits
328
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333
Table 1 Goodness-of-fit indexes for all models in schizophrenia (N = 169) and WAIS-R standardization sample (N = 250) groups Models a
Fit indexes b χ2
df
χ2 / df
GFI
AGFI
NFI
T-LI
RMSEA
CFI
243.37 98.48 95.64 140.73 138.12 86.41 83.91
44 41 40 38 37 38 37
5.53 2.40 2.39 3.70 3.73 2.27 2.27
.76 .91 .91 .87 .87 .92 .92
.64 .86 .86 .78 .78 .86 .86
.83 .94 .93 .90 .90 .94 .94
.82 .95 .94 .89 .89 .95 .95
.18 .09 .08 .12 .12 .08 .08
.86 .96 .96 .93 .93 .97 .97
192.25 118.49 117.52 107.88 107.87 69.69 68.60
44 41 40 38 37 38 37
4.37 2.89 2.94 2.84 2.92 1.83 1.85
.87 .91 .91 .93 .93 .95 .95
.80 .86 .86 .88 .87 .91 .91
.95 .97 .97 .97 .97 .98 .98
.96 .97 .97 .98 .97 .99 .99
.12 .09 .09 .09 .09 .06 .06
.96 .98 .98 .98 .98 .99 .99
c
SZ 1) M1 2) M3 3) M3 d 4) M4C 5) M4Cd 6) M4PC 7) M4PCd WSS d 1) M1 2) M3 3) M3d 4) M4C 5) M4Cd 6) M4PC 7) M4PCd a
M1 = one-factor model; M3 = three-factor model with Verbal Comprehension (VC), Perceptual Organization (PO), and Working Memory (WM) factors; M3d = three-factor model with Digit Symbol on PO and WM factors; M4C = four-factor model with VC, PO, WM, and Social Cognition (SC) factors, with Comprehension and Picture Arrangement on the SC factor; M4Cd = M4C with Digit Symbol on PO and WM factors; M4PC = four-factor model with VC, PO, WM, and SC factors, with Picture Arrangement and Picture Completion on the SC factor; M4PCd = M4PC with Digit Symbol on PO and WM factors. b GFI = Goodness-of-fit index; AGFI = adjusted GFI; RMSEA = Root Mean Square Error of Approximation; NFI = Normed Fit Index; T-LI = Tucker–Lewis Index; CFI = Comparative Fit Index. c SZ = schizophrenia group (n = 169), independence model chi square = 1441.23, df = 55. d WSS = WAIS-R standardization sample 35–44-year-old age group (n = 250), independence model chi square = 4268.98, df = 55.
in schizophrenia. Prior investigations have demonstrated that the VC factor is the least affected by neurocognitive dysfunction in schizophrenia, so comparisons to the VC factor were of particular importance. We also compared patients' performances on the WAIS-R factors to the performance of the WAIS-R normative sample. To conduct these comparisons, factor scores were calculated based on a LISREL maximum likelihood solution derived from the age-corrected scaled scores for the patients with schizophrenia. Pearson correlations were used to examine associations among the WAIS factors and the BPRS negative, disorganized, and psychotic symptoms, as well as social functioning. Finally, factor scores were examined across various IQ categories to determine the relative stability of social cognitive deficits across different levels of intellectual ability. 3. Results Goodness-of-fit statistics for the models are presented in Table 1 for the schizophrenia and WAIS-R groups. For both groups, the three- and four-factor models provide better fit than the one-factor model. In comparing the various three- and four-factor models,
there is little difference in the χ2 and goodness-of-fit statistics for models specifying Digit Symbol to load on the WM factor alone (M3, M4C and M4PC) compared to the more complex models specifying that Digit Symbol would load on both the WM and PO factors (M3d, M4Cd and M4PCd). Additionally, four-factor models specifying Picture Completion and Picture Arrangement to load on the SC factor (M4PC and M4PCd) provided better fit than the four-factor models specifying Comprehension and Picture Arrangement on the SC factor (M4C and M4Cd). To examine improvement in model fit moving from the less complex to more complex models, χ2 differences and NFIs were calculated for nested models (see Table 2). For these comparisons, a significant χ2 difference indicates that one model provides a significant improvement in fit over the other. Only optimal nested models (M3, M4C and M4PC) were selected for comparison purposes. The threeand four-factor models provided significant improvement in model fit when compared to the one-factor model, and model M4PC provided the greatest improvement in fit for both the schizophrenia and standardization samples (Table 2). Also, for both groups, model M4PC provided a significant improvement in fit when compared with the three-factor model, making model M4PC the preferable
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333 Table 2 Incremental fit for confirmatory factor analysis models Model comparisons a
χ2 difference
df
NFI
144.89⁎⁎⁎ 147.73⁎⁎⁎ 102.64⁎⁎⁎ 156.96⁎⁎⁎ 12.07⁎⁎
3 4 6 6 3
.60 .61 .42 .64 –
73.76⁎⁎⁎ 74.73⁎⁎⁎ 84.37⁎⁎⁎ 122.56⁎⁎⁎ 48.80⁎⁎⁎
3 4 6 6 3
.38 .39 .44 .64 –
b
SZ M1–M3 M1–M3d M1–M4C M1–M4PC M3–M4PC WSS b M1–M3 M1–M3d M1–M4C M1–M4PC M3–M4PC
⁎⁎p b .01, ⁎⁎⁎p b .001. a M1 = one-factor model; M3 = three-factor model with Verbal Comprehension (VC), Perceptual Organization (PO), and Working Memory (WM) factors; M4C = four-factor model with VC, PO, WM, and Social Cognition (SC) factors, with Comprehension and Picture Arrangement on SC factor; M4PC = four-factor model with VC, PO, WM, and SC factors, with Picture Arrangement and Picture Completion on SC factor. b SZ = schizophrenia group (n = 169); WSS = WAIS-R standardization sample 25–34-year-old age group (n = 250).
model. The NFI also supports M4PC as the optimal solution in both groups. The maximum likelihood factor loadings for M4PC are presented in Table 3, and indicate a similar pattern of loadings for the schizophrenia group and standardization sample. However, the schizophrenia group evidences a reduced magnitude of loadings across all subtests compared to the standardization sample, a pattern which is also evident from the goodness-of-fit indexes, which were uniformly lower for the schizophrenia group. As expected, the SC factor score was significantly correlated with all other factors, with its strongest correlation with PO (r= .70, n = 169, p b .001), followed by VC (r = .53, n = 169, pb.001), and then WM (r= .41, n= 169, p b .001). Validity of the SC factor was investigated in a number of ways. Within the schizophrenia group, paired t-tests comparing it to the other factors indicated the SC factor score was significantly lower than the VC (t = 3.52, df = 168, p = .001) and PO factors (t = 2.39, df = 168, p = .02), but not the WM factor (t =.73, df =168, p =.47). Mean scores for the VC, PO, WM, and SC factors were 9.02 (SD =2.20), 8.78 (SD = 2.34), 8.33 (SD = 1.78), and 8.45 (SD = 2.10), respectively. Factor scores from the schizophrenia group were also compared to the WAIS-R standardization sample using one-sample t-tests. The schizophrenia group performed significantly worse on all factors, reflecting the generalized nature of cognitive deficit in schizophrenia (pb .01). However, the WM factor score showed the greatest discrepancy from the standardization
329
Table 3 Optimal maximum likelihood factor solution for the four-factor model for the schizophrenia and WAIS-R standardization sample groups (M4PC)a WAIS factor scoresb
Subtest
Schizophrenia sample Vocabulary Comprehension Information Similarities Block Design Object Assembly Arithmetic Digit Span Digit Symbol Picture Completion Picture Arrangement WAIS-R standardization sample Vocabulary Comprehension Information Similarities Block Design Object Assembly Arithmetic Digit Span Digit Symbol Picture Completion Picture Arrangement
VC
PO
WM
SC
0.87 0.78 0.76 0.72 – – – – – – –
– – – – 0.88 0.73 – – – – –
– – – – – – 0.67 0.57 0.52 – –
– – – – – – – – – 0.69 0.63
0.92 0.82 0.84 0.81 – – – – – – –
– – – – 0.91 0.78 – – – – –
– – – – – – 0.84 0.70 0.63 – –
– – – – – – – – – 0.74 0.75
a
Model M4PC = four-factor model with Picture Arrangement and Picture Completion specified to load on factor 4. b VC = Verbal Comprehension, PO = Perceptual Organization, WM = Working Memory, SC = Social Cognition.
sample mean (t=12.19, df= 168, pb .0001), followed by the SC factor score (t= 9.59, df= 168, pb .0001). Associations among factor scores and symptom dimensions were also examined for the 114 patients who were administered the BPRS (Table 4). The SC factor was significantly correlated with Negative Symptoms, Disorganization, and Social Functioning, while the VC factor was significantly correlated with Psychotic Symptoms and Table 4 Correlations among the WAIS factors and symptom dimensions for the schizophrenia group (n = 114) Symptom dimension
Psychotic symptoms Disorganization Negative symptoms Social function
WAIS factor a VC
PO
WM
SC
− .27⁎⁎ − .20⁎ − .14 − .14
.01 − .07 .02 .03
− .15 − .09 − .13 − .02
− .16 − .19⁎ − .20⁎ − .20⁎
⁎p b .05; ⁎⁎p b .01. a VC = Verbal Comprehension, PO = Perceptual Organization, WM = Working Memory, SC = Social Cognition.
330
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333
Fig. 1. Interaction effect for the WAIS-R factor scores by IQ group in the schizophrenia sample (n = 169). aVC = Verbal Comprehension factor, PO = Perceptual Organization factor, WM = Working Memory factor, SC = Social Cognition factor.
Disorganization. The PO and WM factors were not significantly correlated with any symptom dimensions, although some correlations for the WM factor did approach significance. The more general pattern of associations across WAIS factors indicated that VC and SC had the strongest associations with symptom dimensions, followed by the WM factor, with the PO factor generally exhibiting minimal to no association with symptoms. Factor scores were examined across IQ categories by dividing the sample into groups based on 10 Full Scale IQ point increments, forming five groups with IQs of 79 or below (n = 27, mean = 75.0, SD = 3.4), 80–89 (n = 57, mean = 85.4, SD = 2.5), 90–99 (n = 59, mean = 94.1, SD = 2.7), 100–109 (n = 20, mean = 104.7, SD = 3.2), and 110 or greater (n = 6, mean = 112.5, SD = 3.2). Repeated measures ANOVA, using the WAIS factor scores as the repeated measure and IQ groups as a between subjects variable, indicated a significant factor score by IQ group interaction effect, F = 2.61, df = 12,492, p = .002. This interaction shown in Fig. 1 appears largely due to a relative decrease in the Social Cognition factor for the highest IQ group. 4. Discussion Confirmatory factor analysis of the WAIS-R provided support for a four-factor model consisting of Verbal Com-
prehension (VC), Perceptual Organization (PO), Working Memory (WM), and Social Cognition (SC) factors. This factor structure differs from those previously reported for schizophrenia (Allen et al., 1998; Dickinson et al., 2002), by identifying an SC factor composed of Picture Arrangement and Picture Completion. To our knowledge, this is the first study to identify a social cognition factor for the Wechsler scales in patients with schizophrenia, and more generally one of the few that has used factor analysis to identify this construct as separate from nonsocial visuospatial cognitive processes. Therefore, demonstrating the validity of a SC factor was of primary importance in the factor analyses. In this regard, as predicted, two subtests with social content formed the SC factor, with each having relatively high loadings on the factor (N .62). These factor loadings, combined with improvement in model fit when the SC factor was incorporated, provide support for the SC factor's construct validity. Additional support was provided by comparisons among the various factor scores. The SC factor was differentially sensitive to neurocognitive deficit in schizophrenia, and significantly lower than VC and PO factors. The difference between the SC and PO factors was important to demonstrate because based on the high correlation between the two factor scores (r = .70), it might be expected that patients would exhibit comparable levels of performance on both factors; the fact that they did not suggests the SC factor accounts for additional unique variance separate from the more basic perceptual organizational abilities measured by the PO factor. Similarly, while the SC factor was significantly correlated with disorganization (r = −.19), negative symptoms (r = −.20), and social functioning (r = −.20), the PO factor was not (range: r = −.07 to r = .03). These observed correlations were consistent with predictions based on previous studies. For example, Picture Arrangement and similar social sequencing tasks were associated with social withdrawal, negative symptoms, and disorganization, but not psychotic symptoms (Corrigan and Addis, 1995; Shean et al., 2005). Thus, the current SC factor findings parallel these prior studies with regard to symptom dimensions. While the magnitude of correlations between the SC factor and symptoms dimensions were low, suggesting a small percentage of shared variance between variables, the finding is much like that observed in studies reporting minimal to moderate correlations between symptom and neurocognitive variables (see Green, 1998 for review). The high correlation between the SC and PO factors is also to be expected, and is consistent with the view that social and nonsocial cognition represent different levels of processing rather than orthogonal constructs (Penn et al., 1997; Spaulding, 1986). So, while
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333
related to more basic cognitive and perceptual processes and sharing substantial variability with them, the perception and understanding of social information is distinct from these non-social cognitive operations. Because social cognition is itself a complex construct, the question remains as to what aspect of social cognition the SC factor actually measures. The Picture Arrangement and Picture Completion subtests both require perception and analysis of simple line drawings depicting various social situations, people, and objects, with identification of specific details crucial to successful performance. Additionally, the social content of Picture Arrangement and some items from the Picture Completion subtest require knowledge of common social situations. Given these considerations, it appears that the areas of social cognition assessed by this factor would be social perception and social knowledge. Green et al. (2005) noted that these two areas closely interface, as knowledge of social situations is necessary for interpreting social cues. Consequently, because it provides only one indication of the intactness of social cognitive processes, some individuals with schizophrenia might exhibit adequate performance on the SC factor, but still exhibit severe impairment in social functioning as a result of impaired abilities in other aspects of social cognition and depending on specific situational demands. When examining performance across IQ groups, diminished performance on the SC factor for the highest IQ group was noted. This finding should be interpreted with caution given the small number of subjects in that group (n = 6). However, impairment in social functioning at the highest IQ level might help explain why even patients with above average intellectual abilities experience poor social adjustment. There is a group of patients with schizophrenia who exhibit near-normal neurocognitive function referred to as “high-functioning” (Allen et al., 2003; Wilk et al., 2005). It would be of interest to extend the current findings to that highfunctioning group to determine if indeed social cognitive deficits are differentially impaired. In this study and others (Allen and Park, 2007; Goldstein et al., 2006), the two subtests traditionally viewed as requiring the greatest amount of social reasoning, Comprehension and Picture Arrangement, failed to load together to form the Social Cognition factor. Two possible explanations are apparent. One is that the substantial differences in stimulus type (auditory/verbal vs. visual/spatial) precluded Comprehension and Picture Arrangement from loading together. A second is suggested by the findings of Cutting and Murphy (1988, 1990) who demonstrated that patients with schizophrenia had more difficulty on questions designed to assess understanding of
331
social behaviors and situations (e.g., “How would you tell a friend politely that they had stayed too long?” p. 357) than questions assessing everyday knowledge (e.g., “Why is it unsafe to drink tap water in some countries?” p. 358). The Comprehension subtest items are most like the latter form of questions that assess everyday knowledge, and similar in some ways to the content of the Information subtest. On the other hand, because Picture Arrangement requires analysis of social situations and behaviors, it appears to require an ability that is fundamentally different than that required for the Comprehension subtest. We favor this latter explanation, but a definitive answer awaits future investigation. In a more general sense, the current results support the factorial validity of social cognitive tasks in both the schizophrenia group and the WAIS-R standardization sample, and indicate that the social cognition factor is not unique to schizophrenia, but generally descriptive of the Weschler scales when that model is examined. In developing the NIMH-MATRICS Consensus Cognitive Battery, review of existing factor analytic studies did not provide support for social cognition as a separate neurocognitive domain (Nuechterlein et al., 2004). The current results demonstrate that at least some aspects of social cognition do represent separable cognitive domains, and can be distinguished from nonsocial cognitive abilities, such as visuoconstructional abilities. Thus, the results provide factor analytic support for inclusion of social cognition as a separable cognitive domain in the NIMHMATRICS Consensus Cognitive Battery. A limitation of this study was use of the WAIS-R rather than the most recent version, the WAIS-III (Wechsler, 1997). CFA of the WAIS-III would be a useful next step in further validating the SC factor in schizophrenia, particularly since addition of some new subtests to the WAIS-III have allowed for identification of a Processing Speed (PS) factor in addition to the VC, PO, and WM factors. A fifth SC factor is expected in schizophrenia given the findings of two recent studies, one of the WAIS-III normative sample (Allen and Park, 2007), and another of children and adults with highfunctioning autism (Goldstein et al., 2006). Both studies identified an SC factor composed of Picture Arrangement and Picture Completion. It would also be important to identify the SC factor in females with schizophrenia, as the current sample was composed entirely of males. Future studies may determine whether meaningful associations exist between the SC factor and other important features of schizophrenia. For example, the SC factor might be useful as a risk indicator, given that Picture Arrangement performance is significantly lower in children who later go on to develop schizophrenia
332
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333
(Niendam et al., 2003), and it is one of the most sensitive Wechsler subtests to post-psychotic cognitive decline in childhood onset schizophrenia (Bedwell et al., 1999). It will also be important to continue to develop measures to assess various subcomponents of social cognition, and determine associations between these specific subcomponents and behavioral and functional outcomes, much like what has been accomplished with nonsocial neurocognitive measures. Acknowledgement Indebtedness is expressed to the Medical Research Service, Department of Veterans Affairs, and to the Center for Advanced Research, University of Nevada Las Vegas, for support of this research. References Allen, D.N., Park, B.S., 2007. Using the Wechsler Adult Intelligence Scale to assess social cognition. International Neuropsychological Society. February 4–7, Portland, OR. Allen, D.N., Seaton, B.E., Huegel, S.G., Goldstein, G., Gurklis, J.A., van Kammen, D.P., 1998. Confirmatory factor analysis of the WAIS-R in patients with schizophrenia. Schizophr. Res. 34, 87–94. Allen, D.N., Goldstein, G., Warnick, E., 2003. A consideration of neuropsychologically normal schizophrenia. J. Int. Neuropsychol. Soc. 9, 56–63. Aylward, E., Walker, E., Bettes, B., 1984. Intelligence in schizophrenia: meta-analysis of the research. Schizophr. Bull. 10, 430–459. Bedwell, J.S., Keller, B., Smith, A.K., Hamburger, S., Kumra, S., Rapoport, J.L., 1999. Why does postpsychotic IQ decline in childhood-onset schizophrenia? Am. J. Psychiatry 156, 1996–1997. Bentler, P.M., 1990. Comparative fit indexes in structural models. Psychol. Bull. 107, 238–246. Bentler, P.M., Bonnett, D.G., 1980. Significance tests and the goodness of fit in analysis of covariance structures. Psychol. Bull. 88, 588–606. Bollen, K.A., 1990. Overall fit in covariance structure models: two types of sample size effects. Psychol. Bull. 107, 256–259. Bunney, W.E., Hamburg, D.A., 1963. Methods for reliable longitudinal observation of behavior. Arch. Gen. Psychiatry 9, 280–294. Cohen, J., 1957. The factorial structure of the WAIS between early adulthood and old-age. J. Consult. Clin. Psychol. 21, 283–290. Cole, D.A., 1987. Utility of confirmatory factor analysis in test validation research. J. Consult. Clin. Psychol. 55, 584–594. Corrigan, P.W., Addis, I.B., 1995. The effects of cognitive complexity on a social sequencing task in schizophrenia. Schizophr. Res. 16, 137–144. Cutting, J., Murphy, D., 1988. Schizophrenic thought disorder: a psychological and organic interpretation. Br. J. Psychiatry 152, 310–319. Cutting, J., Murphy, D., 1990. Impaired ability of schizophrenics, relative to manics or depressives, to appreciate social knowledge about their culture. Br. J. Psychiatry 157, 355–358. Dickinson, D., Iannone, V.N., Gold, J.M., 2002. Factor structure of the Wechsler Intelligence Scale III in schizophrenia. Assessment 9, 171–180.
Fioravanti, M., Carlone, O., Vitale, B., Cinti, M.E., Clare, L., 2005. A meta-analysis of cognitive deficits in adults with a diagnosis of schizophrenia. Neuropsychol. Rev. 15, 73–95. Fiske, S.T., Taylor, S., 1991. Social Cognition, second ed. McGraw Hill, New York. Frith, C.D., 1992. The cognitive neuropsychology of schizophrenia. Lawrence Erlbaum Associates, Hove, UK. Goldberg, T.E., Gold, J.M., Greenberg, R., Griffin, S., Schulz, S.C., Pickar, D., Kleinman, J.E., Weinberger, D.R., 1993. Contrasts between patients with affective disorders and patients with schizophrenia on a neuropsychological test battery. Am. J. Psychiatry 150, 1355–1362. Goldstein, G., Allen, D.N., Minshew, N.J., Williams, D., Volkmar, F., Klin, A., Schulz, R., 2006. Confirmatory factor analysis of the Wechsler scales in high functioning autism. International Meeting for Autism Research, June 1–3, Montreal, Canada. Green, M.F., 1998. Schizophrenia From a Neurocognitive Perspective: Probing the Impenetrable Darkness. Allyn and Bacon, Boston. Green, M.F., Olivier, B., Crawley, J.N., Penn, D.L., Silverstein, S., 2005. Social cognition in schizophrenia: recommendations from the measurement and treatment research to improve cognition in schizophrenia new approaches conference. Schizophr. Bull. 31, 882–887. Heinrichs, R.W., Zakzanis, K.K., 1998. Neurocognitive deficit in schizophrenia: a quantitative review of the evidence. Neuropsychology 12, 426–445. Jöreskog, K.G., Sörbom, D., 1993. LISREL 8 User's Reference Guide. Scientific Software International, Chicago. Liddle, P.F., 1987. The symptoms of chronic schizophrenia. A reexamination of the positive–negative dichotomy. Br. J. Psychiatry 151, 145–151. Lipsitz, J.D., Dworkin, R.H., Erlenmeyer-Kimling, L., 1993. Wechsler comprehension and picture arrangement subtests and social adjustment. Psychol. Assess. 5, 430–437. Mueser, K.T., Curran, P.J., McHugo, G.J., 1997. Factor structure of the Brief Psychiatric Rating Scale. Psychol. Assess. 9, 196–204. Mulaik, S.A., James, J.R., van Alstine, J., Bennett, N., Lind, S., Stilwell, C.D., 1989. Evaluation of goodness-of-fit indices for structural equation models. Psychol. Bull. 105, 430–445. Niendam, T.A., Bearden, C.E., Rosso, I.M., Sanchez, L.E., Hadley, T., Nuechterlein, K.H., Cannon, T.D., 2003. A prospective study of childhood neurocognitive functioning in schizophrenic patients and their siblings. Am. J. Psychiatry 160, 2060–2062. Nuechterlein, K.H., Barch, D.M., Gold, J.M., Goldberg, T.E., Green, M.F., Heaton, R.K., 2004. Identification of separable cognitive factors in schizophrenia. Schizophr. Res. 72, 29–39. Ostrum, T.M., 1984. The sovereignty of social cognition. In: Wyer, R.S., Skrull, T.K. (Eds.), Handbook of Social Cognition, vol. 1. Laurence Erlbaum, Hillsdale, NJ, pp. 1–37. Overall, J.E., Gorham, D.R., 1962. The Brief Psychiatric Rating Scale. Psychol. Rep. 10, 799–812. Penn, D.L., Spaulding, W.D., Reed, D., Sullivan, M., 1996. The relationship of social cognition to ward behavior in chronic schizophrenia. Schizophr. Res. 20, 327–335. Penn, D.L., Corrigan, P.W., Bentall, R.P., Racenstein, J.M., Newman, L., 1997. Social cognition in schizophrenia. Psychol. Bull. 121, 114–132. Rapaport, D., Gill, M.M., Schafer, R., 1968. Diagnostic Psychological Testing (rev. ed.). International Universities Press, New York. Schafer, R., 1948. The Clinical Application of Psychological Tests. International Universities Press, New York.
D.N. Allen et al. / Schizophrenia Research 93 (2007) 325–333 Seaton, B.E., Goldstein, G., Allen, D.N., 2001. Sources of heterogeneity in schizophrenia: the role of neuropsychological functioning. Neuropsychol. Rev. 11, 45–67. Shean, G., Murphy, A., Meyer, J., 2005. Social cognition and symptom dimensions. J. Nerv. Ment. Dis. 193, 751–755. Spaulding, W.D., 1986. Assessment of adult-onset pervasive behavior disorders, In: Ciminero, A., Calhoun, K., Adams, H. (Eds.), Handbook of Bhaviorla Assessment, 2nd ed. Wiley, New York, pp. 631–669.
333
Tucker, L.R., Lewis, C., 1973. A reliability coefficients for maximum likelihood factor analysis. Psychometrica 38, 1–10. Wechsler, D., 1997. Wechsler Adult Intelligence Scale, Third edition. Psychological Corporation, San Antonio, TX. Wechsler, D., 1981. Manual for the Wechsler Adult Intelligence Scale— Revised. Psychological Corporation, New York. Wilk, C.M., Gold, J.M., McMahon, R.P., Humber, K., Iannone, V.N., Buchanan, R.W., 2005. No, it is not possible to be schizophrenic yet neuropsychologically normal. Neuropsychology 19, 778–786.
