141

Psychiatry Research, 44: 14 l-1 5 I Elsevier

Empirical Assessment of the Factorial Structure of Clinical Symptoms in Schizophrenic Patients: Formal Thought Disorder Philip D. Harvey, Mark F. Lenzenweger, Richard S.E. Keefe, David L. Pogge, Mark R. Serper, and Richard C. Mohs Received March 6, 1992; revised version received August 12, 1992; accepted October 6, 1992.

Abstract. Male schizophrenic patients (n = 142) were examined with a clinical assessment of their language dysfunctions with the Scale for the Assessment of Thought, Language, and Communication (TLC). Confirmatory factor analyses were conducted to test the relative fit of several differential theoretical models of the factorial structure of thought disorders. The models examined were positivenegative thought disorder, a three-factor model based on the results of an earlier exploratory factor analysis, and a simpler verbal productivity-disconnection model that can be extracted from other exploratory analyses and empirical studies. The positive-negative thought disorder model failed to fit the data, while the three-factor model fit the data, but no better than the simpler verbal productivity-disconnection model. Key Words. Language,

cognition,

factor analysis.

Although thought disorder is no longer seen to be the primary or pathognomonic sign of schizophrenia (Bleuler, 1950) because of its demonstrated lack of specificity to the disorder (Harrow and Quinlan, 1977; Harvey, 1983), its importance as a symptom is not in question. There are multiple means of assessment of thought disorder, including rating procedures based on projective tests (Johnston and Holzman, 1977), ratings of language in response to other types of evocative tests (e.g., Harrow et al., 1982), and clinical ratings of speech produced in structured interview contexts. The most commonly used clinical rating scale for thought disorder is the Scale for Assessment of Thought, Language, and Communication (TLC; Andreasen, 1979a, 19796). The TLC was designed to provide the user with specific definitions and quantitative ratings for 18 subtypes of aberrant language. These subtypes cover the broad range of communication deviance displayed by schizophrenic patients. A review by Andreasen and Grove (1986) documented the reliability and validity of the scale.

Professor; Richard SE. Keefe, Ph.D., is Assistant Professor; and Department of Psychiatry, Mt. Sinai School of Medicine, New York, NY. Mark Lenzenweger, Ph.D., is Associate Professor, Cornell University, Ithaca, NY. David L. Pogge, Ph.D., is Assistant Professor, Fairleigh Dickinson University, Teaneck, NJ. Mark R. Serper, Ph.D., is Assistant Professor, New Jersey Medical School, Newark, NJ. (Reprint requests to Dr. P.D. Harvey, Dept. of Psychiatry, Box 1230, Mt. Sinai School of Medicine, I Gustave L. Levy Pl., New York, NY 10029, USA.) Philip D. Harvey,

Ph.D.,

is Associate

Richard C. Mohs, Ph.D., is Professor,

0165-1781/92/SO5.00

@ 1992 Elsevier Scientific

Publishers

Ireland

Ltd.

142 In recent years, thought disorder has been conceptualized as consisting of various dimensions, including aspects that could be viewed as positive and negative (Andreasen, 1979b; Harvey et al., 1984; Pogue-Geile and Harrow, 1984). Negative thought disorder is often conceptualized to include aspects of reduced amount of information in speech, including poverty of speech, poverty of content of speech, and blocking, while positive thought disorder includes aspects of language dysfunction such as pressure of speech, derailment, and tangentiality (Andreasen, 1981, 1982). Positive and negative thought disorder appear to be statistically related to other positive and negative symptoms of schizophrenic pathology (Walker et al., 1988) and to manifest discriminant validity. The distinction between positive and negative thought disorder has been validated by the identification of differential correlations with a variety of cognitive functions (Walker and Harvey, 1986; Harvey et al., 1988) and by findings of diagnostic specificity of the two dimensions (Andreasen, 1979b; Harvey et al., 1984; Andreasen and Grove, 1986). Dichotomizing the varied aspects of schizophrenic thought disorder into a positive/ negative structure, however, may not accurately capture the complexity of schizophrenic communication deviance. There may potentially be more than two factors within the spectrum of thought disorder or the factors may reflect dimensions other than excess/deficit. Some investigators have found that the negative thought disorder items poverty of speech and poverty of content fail to correlate with each other (Pogue-Geile and Harrow, 1984; Harvey and Serper, 1990). Two exploratory factor analyses of TLC scores (Berenbaum et al., 1985; Andreasen and Grove, 1986) have yielded multifactorial solutions, in contrast to the two-factor solution that the positive versus negative conception would suggest. Both of those exploratory analyses produced factors that could be best labeled productivity and disconnection, with both reports suggesting that tendencies toward disconnection in speech can themselves be subdivided into two separate factors. In the present study, we examined a large group of schizophrenic patients and used a confirmatory factor-analytic approach to examine the fit of five alternative factor models. A three-factor model derived from exploratory analysis (Berenbaum et al., 1985) was examined, as were two different two-factor models: the positivenegative thought disorder model as described by Andreasen (19793, 198 1, 1982) and a verbal productivity and disconnection model. This final model can be conceptualized from both of the earlier exploratory analyses of TLC scores and earlier theoretical and empirical work on thought disorder. Finally, we estimated two models for comparison purposes: (1) a null model that assumed no underlying structure among the TLC variables and (2) a single-factor model that corresponded to a nonspecific severity (or general dysfunction) factor underlying the observed TLC variables. The rationale for use of the three-factor model is that it is an exploratory factoranalytic model that was based on the TLC. Variations of the positive-negative thought disorder model have been used extensively in the studies described above. The verbal productivity and disconnection model has its roots in three earlier lines of evidence. First is that earlier studies with the TLC have found that the severity levels of two components of negative thought disorder, poverty of speech and poverty of content of speech, are not positively correlated (Harvey et al., 1984; Pogue-Geile and

143

Harrow, 1984), suggesting that they are not likely to be indicators of a single underlying construct. Earlier exploratory studies (Berenbaum et al., 1985; Andreasen and Grove, 1986) have found that measures of excessive and deficient verbal productivity are reciprocally related to a single factor. The second is that this twofactor model is basically a simpler version of the three-factor models of Andreasen and Grove and Berenbaum et al., reducing all aspects of disconnected speech into a single factor. The third is that studies of schizophrenic symptoms in general (Liddle, 1987) have yielded factors reflecting psychomotor poverty and disorganization, processes similar to those in this model. This model is a direct reflection of the conceptualization initially suggested by Kraepelin (1919), who asserted that schizophrenic patients tended to manifest two general tendencies toward aberration in speech: “mutism” and “derailment” (pp. 64-66). Methods Subjects. Subjects were 142 male schizophrenic inpatients who participated in biological research protocols under the auspices of the Bronx VA Medical Center Schizophrenia Biological Research Center. All patients met DSM-ZZZ criteria for schizophrenia (American Psychiatric Association, 1980) on the basis of information collected with a structured diagnostic interview (Schedule for Affective Disorders and Schizophrenia [SADS]; Spitzer et al., 1978) and extensive medical record review. The SADS was evaluated by two trained raters simultaneously, with the raters generating independent DSM-ZZZ diagnoses that were then presented at a meeting with a senior consensus clinician. The independent diagnoses had a K coefficient of 0.87, with all cases receiving their final diagnoses at the consensus meeting. All patients in this study met the active phase of illness criteria from section A of the DSM-ZZZ. Table 1 presents descriptive information about the patients. Of these patients, 43 (3oo/o) had been receiving neuroleptic medication in a chlorpromazineequivalent dose of at least 1000 mg/day for at least 1 week and 35 (25%) had been free of all neuroleptic medication for at least 2 weeks. The other 64 patients (45%) had either been receiving neuroleptic medication for < 1 week or were medication free, but for < 2 weeks. Thought Disorder Ratings. The two raters who completed the SADS interview used the Scale for Assessment of Thought, Language, and Communication (TLC; Andreasen, 1979~) to make thought disorder ratings at the time of the interview on the basis of the subjects’ language during the diagnostic interview. The thought disorder ratings and SADS interview were performed within 3 weeks of the patient’s admission to the research unit. Raters collected information on the types of communication disorder manifested by the patients during the interview with notations and examples. Ratings were completed by each rater on all 18 items from the TLC and were then subjected, item by item, to a consensus procedure at the time of the consensus diagnosis of the subjects. Each item, therefore, was initially rated independently at the time of assessment of the patients and then was subjected to consensus. Of the 18 thought disorder variables in the TLC, only eight were included in the Berenbaum et al. (1985) study.’ Direct comparisons in confirmatory factor analysis between competing models require that the factor structures for the various models be derived from a single correlation matrix. Thus, the limiting case for the number of variables across all models is the model with the fewest variables. It was possible to test a three-factor model as well as to test two alternative models with those eight variables, since enough variables to test the verbositydisconnection model and the positive-negative thought disorder model were included in that three-factor model. Thus, the decision was made to use those eight variables for all subsequent

1.Berenbaum et al. (1985) modified the TLC to separate "tangentiality"from “non sequitur”responses to questions. Since we did not modify the TLC in that way, that variable was not included in the present analyses.

144

Table 1. DescriDtive

data on Datients

Variable Age (yrl Age at first admission

(yr)

Years of education Number

of prior admissions

Total months hosoitalized

Mean

SD

37.71

11.31

23.49

5.49

11.92

1.96

7.09

6.52

55.00

77.64

model testing and to prioritize direct contrasts between models versus evaluation of all possible factor models, with no direct comparisons. Table 2 presents the severity scores for those eight signs of thought disorder. The independent ratings for these eight signs were compared across all 142 patients and the interrater reliabilities of the items were computed with intraclass correlation coefficients (Bartko, 1966). Table 2 presents these interrater reliabilities. As can be seen in Table 2, the reliabilities of the items are adequate, especially when it is considered that all items for each subject were subjected to consensus after independent rating. The intraclass correlation for the total score of the TLC, including all 18 items, was 0.8 1.

Table 2. Reliability (n= 142)

Poverty

of speech

Poverty

of content

and severity of thought

of speech

disorders

in the patients ICC

Mean

SD

0.94

1.10

0.78

1.25

1.21

0.79 0.88

0.47

0.75

Tangentiality

1.83

1.26

0.70

Derailment

1.64

1.38

0.68

Pressure

of speech

Incoherence

0.74

1.15

0.50

Circumstantiality

1.14

1.06

0.87

Loss of goal

1.32

1.09

0.90

Note. ICC = intraclass correlation coefficient. Thought

Disorder Models. Five models of thought disorder were tested, with the models varying in the degree of restrictiveness. First, a null model which assumed the absence of latent structure was estimated. Second, a unidimensional “severity” model was estimated. The third model was that of positive and negative thought disorder (Andreasen, 1979b). This model was specified by grouping poverty of speech and poverty of content of speech on one factor (negative thought disorder) and the other six items on the other factor (positive thought disorder).2 Another two-factor model, which we label as “verbal productivity and disconnection,” was specified with poverty of speech and pressure of speech loading on the verbal productivity factor and all other items loading on the disconnection factor. A fifth model, the three-factor model (Berenbaum et al., 1985), was specified by grouping poverty of speech, pressure of speech, circumstantiality, and loss of goal as factor 1; incoherence and 2. Note that this definition of positive and negative thought disorder, based on the eight items examined in this study, varies from the original TLC definition of positive and negative thought disorder (Andreasen, 19796), from the redefinition of positive and negative thought disorder presented in the Scale for the Assessment of Positive Symptoms (SAPS; Andreasen, 1982), as well as from the more abbreviated versions of positive thought disorder used in our previous research (Harvey et al., 1988; Harvey and Serper, 1990). What is retained is the concept that poverty of speech and poverty of content of speech form the core of negative thought disorder and that derailment, tangentiality, and other similar language impairments form the core of positive thought disorder.

145

derailment as factor 2; and poverty of content of speech and tangentiality loaded equally strongly on factors 2 and 3 in analyses twice with that model, once with tangentiality loading tangentiality loading on factors 2 and 3. Fig. 1 visually depicts

tangentiality as factor 3. Since the original study, we ran the only on factor 3 and once with the factor models.

Fig. 1. Factorial models of thought disorder WDEL3: Verbal

MODEL 2: Positive 8 Negative Thought Disorder

Productivity 8 Disconnection

P-wdspeecho

+

Pressureof speech0

+

MODEL 4: Berenbaum 3

- Factor

Model

EE=z!I12s+F_rt

oar-0 lnmhe8mce0

+ +

-Factor2

Data Analysis. To evaluate the five latent structure models of thought disorder in schizophrenia, the LISREL VI (Linear and Structural Relations VI) (Joreskog and Sorbom, 1984) program was used to perform a confirmatory factor analysis (CFA). In CFA an investigator constructs a factor structure that is derived from assumptions of the theory of interest and specifies an implied model of how correlations between a group of variables should have been caused by latent variables (i.e., those factors). In this case, a Pearson correlation matrix was used as the input data. The LISREL program estimates model parameters based on the measurement model and then compares, on the basis of maximum likelihood-based procedures, the estimated correlation matrix to the actual input correlation matrix. The investigator must specify whether the factors are allowed to correlate with each other and in this case, on the basis of the principle of clinical reality-symptoms co-occur in the same patient-correlations between the factors were allowed to be free to vary. Since there were no strong empirical grounds for specifying the exact loadings of each TLC item on its respective factor, those loadings were allowed to vary as well. Similarly, the error terms were not permitted to correlate, in the absence of empirical evidence suggesting that they should be related. The quality of the fit between estimated and solution models can be evaluated statistically with the x2 test (Long, 1983; Joreskog and Sorbom, 1984). The goal of a theoretical model is to explain as much of the covariance present in the obtained data as possible within the specifications of the model. In this instance, the null hypothesis is that all of the population covariance has been extracted from the correlation matrix by the prespecified measurement model. If the x2 value is statistically significant (e.g.,p < 0.05) then the residual matrix still has significant covariance in it, and one may conclude that the model being tested does not fit the data well (Gorsuch, 1983, p. 129). If thex2 value is not statistically

146 significant, then the null hypothesis is not rejected, and one may conclude that the prespecified model fits the observed data, leaving little covariance in the residual matrix. In a study that involves large groups of subjects, a model may provide a good fit to observed data but will generate a statistically significant x2 value (Bentler and Bonett, 1980; Marsh et al., 1988). The x2 contrasts and incremental fit indexes are then used to assess the relative fits of competing models (Bentler and Bonett, 1980; Marsh et al., 1988). Extensive reviews and introductions to the mathematical approach to parameter estimation involved in CFA and the CFA approach to statistical comparisons of competing substantive models are readily available elsewhere (see Long 1983; Joreskog and Sorbom, 1984; Hayduk, 1987; Bollen, l989), and extensive discussions of applications of CFA in psychopathology research can be found in Lenzenweger et al. (1989, 1991). Using the data contained in Table 3, we estimated the five models described above.

Table 3. Intercorrelations of variables 1. Poverty

of speech

2. Poverty

of content of speech

3. Pressure

of speech

4. Tangentiality 5. Derailment 6. Incoherence 7. Circumstantiality 6. Loss of goal

1

2

-

0.22

-0.32

-0.20

-0.22

-0.10

-0.09

-

0.10

0.45

0.44

0.35

0.34

0.35

-

0.32

0.32

0.09

0.05

0.23

-

0.71

0.51

0.20

0.54

-

0.59

0.31

0.45

-

0.27

0.19

-

0.27 -

3

4

5

6

7

8 -0.15

Results Table 4 (top) presents the results of the LISREL analyses evaluating goodness of fit for each model. The null model revealed a relatively large and significant x2 value and a relatively low LISREL goodness-of-fit index (a “perfect” fit between model and data would generate a fit index of 1.00). These data indicate that a model assuming no latent structure underlying schizophrenic thought disorder symptomatology fits the data poorly. The unidimensional model produced a smaller x* value relative to the null model and an improved goodness-of-fit index. The LISREL program failed to converge to a solution for the two-factor “positive and negative thought disorder” model.. which suggested the model fit the observed TLC data quite poorly. The two-factor “verbal productivity and disconnection” model revealed a smaller x2 value relative to both the null and unidimensional models and yielded an improved goodness-of-fit index. Finally, with tangentiality loading only on factor 3, the three-factor model generated a x2 value that was smaller than that observed for the null and unidimensional models. When we recomputed the LISREL with the tangentiality item loading on both factors 2 and 3, the program failed to converge to a solution, again suggesting a very poor fit of that variant of the model. Thus, despite the significant x2 value associated with both of the multifactorial models where the program generated a solution (a function of large sample size: Bentler and Bonett, 1980; Marsh et al., 1988), they appeared initially to provide a better fit to the data than the null and unifactorial models. The four models that could be estimated (i.e., the positive-negative model and the three-factor model with tangentiality loading on both factors could not be estimated) were compared sequentially where appropriate by contrasting their respective x2

147 Table 4. Results of confirmatory factor analyses on the Scale for the Assessment of Thought, Language, and Communication Model fits Model Null, Unidimensional,

df

GFI

P

342.55

28

0.581

0.001

X2 86.07

20

0.897

0.001

Verbalproductivity/disconnection255.49

19

0.915

0.001

Three-factormodels

58.33

17

0.910

0.001

X’

df

P

b*t

AId

&,-MI

276.48

8

0.001

0.795

0.807

MO-M2

287.06

9

0.001

0.829

0.838

MO-M3

284.22

11

0.001

0.784

0.829

10.58

1

0.01

0.034

0.031

7.74

3

0.01

-0.011

0.022

Model contrasts

MA& Ml-M3

,.

Note. GFI is the LISREL goodness of fit index, which ranges from 0.00 to 1.00, with 1 .oO being a perfect fit. p*, = nonnormed incremental M index (Tucker and Lewis, 1973; &I = normed incremental fit index (Bentfer and Sonett, 1960).

values.

The x* differences were calculated in all of the nested comparisons (i.e., where a subsequent model is derived by fixing a free parameter in the previous model). Note that the unifactorial model is nested within the null model and the verbal-productivity disconnection model and the three-factor model are nested within the unifactorial model. The multifactorial models are not, however, nested and cannot be compared directly with statistical tests. The dzfferences between the x2 values and degrees of freedom associated with each model were calculated and evaluated for statistical significance (Bentler and Bonett, 1980). To assess the amount of information gained in the comparison of two competing models and to generate an estimate of the improvement in fit, incremental fit indexes were calculated (Tucker and Lewis, 1973; Bentler and Bone& 1980; Marsh et al., 1988). Table 4 (bottom) presents the results of the LISREL-based model comparisons, as well as the incremental fit indexes. All three estimated models provided significantly better fits to the observed data than the null model (p < 0.001 for all comparisons). As can be seen from Table 4, both of the more complex models provided a significantly better fit to the data than did the unidimensional “severity” model. Furthermore, the two-factor “verbal productivity and disconnection” model yielded a smaller x2 value relative to the null and unifactorial models than did the three-factor model. The incremental fit indexes in the bottom of Table 4 are consistent with the sequential x* contrasts, and it is most important to note that the increment in fit provided by the “verbal productivity and disconnection” model over the more restricted models is substantial (Bentler and Bonett, 1980). For the purposes of later direct testing of the fit of this model to other sets of data, the factor loadings for the two-factor verbal productivity and disconnection model are presented in Table 5.

148 Table 5. Factor loadings for the verbal-productivity/disconnection Verbal productivity Item

Loading

Pressure

of speech

Poverty

of speech

Poverty

of content of speech

model

Disconnection

t

Loading

0.733

4.32

0.00

-0.438

-3.59

t

0.00

0.00

0.519

6.22

Circumstantiality

0.00

0.319

3.64

Loss of goal

0.00

0.557

6.76

Incoherence

0.00

0.628

7.83

Derailment

0.00

0.868

12.07

Tangential&y

0.00

0.833

11.38

Note. All p values are significant at p < 0.001. The correlation between the two factors is r = 0.47, p < 0.001.

Discussion The best fitting model of thought disorder divided the language disturbances of schizophrenic patients into those associated with disturbances in the amount of speech produced and those associated with tendencies toward disconnectedness in speech. Positive and negative thought disorder as a model failed to explain the data. The most obvious reason for this failure is that negative thought disorder is not a statistically supported concept: the two indicators were found in this study, as in others (Pogue-Geile and Harrow, 1984; Harvey and Serper, 1990), to be poorly correlated. In addition, poverty of content was correlated with measures of disconnection in speech, a result similar to earlier studies of schizophrenic symptoms (Liddle, 1987) where this symptom was correlated with other measures of disorganized behavior. Another possibility, albeit a less plausible one, is that somehow the different definition of positive thought disorder in this study led to a failure to confirm the model. The definition of what constitutes positive thought disorder has varied substantially in the past (Andreasen, 19796, 1982; Harvey et al., 1988; Harvey and Serper, 1990) and the indicators of positive thought disorder included in this study were well correlated with each other. Finally, the three-factor model, although fitting the data reasonably well, is less parsimonious and probably less desirable at equivalent levels of fit than the twofactor model. It is possible, however, that the three-factor model derived by Andreasen and Grove (1986) would explain more of the variance in TLC scores, because of its greater comprehensiveness, and be a more suitable model than the present study’s two-factor model. This possibility is underscored by the fact that neither of the present multifactorial models provided an ideal fit to the data. Experts in the area of CFA suggest that indexes of goodness of fit of 0.90 are at the lower limit of a well-fitting model (Bentler and Bonnet, 1980; Marsh et al., 1985), such that both of the multifactorial models in the present study are operating at or below the lowest suitable level of goodness of fit. It is possible that the items used in this analysis are different from the other 10 TLC items and that their inclusion would alter the factor structure. This possibility is contradicted by the fact that these items were initially selected for analysis because they were of higher prevalence and reliability than the other items (Berenbaum et al., 1985).

149 There are several implications of these findings. Earlier conclusions about the characteristics of negative thought disorder, including temporal stability (Harvey et al., 1984; Ragin and Oltmanns, 1987; Docherty et al., 1988) and correlations with other negative symptoms (Walker et al., 1988) probably apply to verbal underproductivity, as measured by poverty of speech. Earlier studies have suggested that “positive thought disorder” often is related to the severity of deficits in verbal information processing and attentional functioning (e.g., Harvey et al., 1988). Poverty of content of speech might be investigated in the future to determine if its psychological correlates are similar to those of the rest of the disconnection factor. These results also suggest that clinical thought disorder has a factor structure similar to that of other schizophrenic symptoms, consistent with the concepts of “psychomotor poverty” and “disorganization” found by Liddle (1987). Future research should carefully focus on variables that affect linguistic underproductivity in schizophrenic patients. Among the possible models of reduced verbal output in schizophrenia worthy of future study are the following: (1) social anxiety and resultant reduction in language output; (2) quasiaphasic phenomena, whereby ability to produce speech is impaired such as is often manifested in postinfarct cases with reduced verbal fluency; and (3) “poverty of thought,” wherein schizophrenic patients with reduced verbal output also manifest reduced, slowed, or impoverished spontaneous cognitive productions. Understanding whether verbal underproductivity in schizophrenic patients is neurologically, psychologically, or socially mediated may provide insights into the origin of other aspects of behavioral poverty (e.g., affective nonresponse) in the disorder. In addition, research on thought disorder might benefit from attempting to conceptualize communication disturbances from the perspective of the empirically confirmed model in this study and attempting to begin to identify response to treatment and cognitive correlates, in order to see how the present models truly diverge in correlates from the positivenegative model on which earlier studies were based. In evaluating this study, a limitation of the research design, in addition to the imperfect fit of the best supported model, is that all of the patients were men. Possibly the addition of female patients would affect the outcome of the analyses. A second limitation is that the subgroups of patients within each of the three drug statuses were too small to permit empirical confirmation of the models within them. Clinical thought disorder is a particularly medication-responsive symptom and might have been affected in our medicated group. When our sample size increases in those groups, those replications will be attempted. A final limitation is that these models do not encompass all possible aspects of communication impairment in

schizophrenic patients, in that alternative means of assessment of communication disorder might yield other results, including identification of other aberrant processes such as unusual content of speech. This research was supported by a Schizophrenia Biological Research Center grant from the General Medical Research Service of the Veterans Administration to the Bronx VA Medical Center (Kenneth L. Davis, M.D., principal investigator). Computing resources were provided by the Cornell Institute for Social and Economic Research (CISER) at Cornell University, Ithaca, NY. Acknowledgment.

150

References American Psychiatric Association. DSM-III: Diagnostic and Statistical Manual of Mental Disorders. 3rd ed. Washington, DC: APA, 1980. Andreasen, N.C. Thought, language, and communication disorders: I. Clinical assessment, definition of terms and evaluation of their reliability. Archives of General Psychiatry, 36:13151321, 1979a. Andreasen, N.C. Thought, language, and communication disorders: II. Diagnostic significance. Archives of General Psychiatry, 36: 13251330, 19796. Andreasen, N.C. The Scale for the Assessment of Negative Symptoms, Iowa City: University of Iowa, 198 I. Andreasen, N.C. The Scalefor the Assessment of Positive Symptoms. Iowa City: University of Iowa, 1982. Andreasen, N.C., and Grove, W.M. Thought, language, and communication in schizophrenia: Diagnosis and prognosis. Schizophrenia Bulletin, 12:348-359, 1986. Bartko, J. The intraclass correlation coefficient as a measure of reliability. Psychological Reports, 19:3-l I, 1966. Bentler, P.M., and Bonett, D.G. Significance tests and goodness-of-fit in the analysis of covariance structures. Psychological Bulletin, 88588-606, 1980. Berenbaum, H.; Oltmanns, T.F.; and Gottesman, 1.1. Formal thought disorder in schizophrenic patients and their twins. Journal of Abnormal Psychology, 94:3-16, 1985. Bleuler, E. Dementia Praecox or the Group of Schizophrenias. New York: International Universities Press, 1950. Bollen, K.A. Structural Equations With Latent Variables. New York: Wiley, 1989. Docherty, N.M.; Schnur, M.; and Harvey, P.D. Reference performance and positive and negative thought disorder: A followup study of manics and schizophrenic patients. Journal of Abnormal Psychology, 971437-442, 1988. Gorsuch, R.L. Factor Analysis. 2nd ed. Hillsdale, NJ: Lawrence Erlbaum Associates, 1983. Harrow, M.; Grossman, L.; Silverstein, M.; and Meltzer, H.Y. Thought pathology in manic and schizophrenic patients: Its occurrence at admission and seven weeks later. Archives of General Psychiatry, 39:665-67 1, 1982. Harrow, M., and Quinlan, D. Is disordered thinking unique to schizophrenia? Archives of General Psychiatry, 34: 15-21, 1977. Harvey, P.D. Speech competence in manic and schizophrenic psychoses: The association between clinically rated thought disorder and cohesion and reference performance. Journal of Abnormal Psychology, 92:368-377, 1983. Harvey, P.D.; Earle-Boyer, E.A.; and Levinson J.C. Cognitive deficits and thought disorder: A retest study. Schizophrenia Bulletin, 14:57-66, 1988. Harvey, P.D.; Earle-Boyer, E.A.; and Wielgus, M.S. The consistency of thought disorder in mania and schizophrenia: An assessment of acute psychotics. Journal of Nervous and Mental Disease, 1721458-463, 1984. Harvey, P.D., and Serper, M.R. Linguistic and cognitive failures in schizophrenia. A multivariate analysis. Journal of Nervous and Mental Disease, 172:487-493, 1990. Hayduk, L.A. Structural Equation Modeling With LISREL: Essentials and Advances. Baltimore, MD: The Johns Hopkins University Press, 1987. Johnston, M.H., and Holzman, P.S. Assessing Schizophrenic Thinking: A Clinical Research Instrument for Measuring Thought Disorder. San Francisco: Jossey-Bass, 1977.

151 Joreskog, K.G., and Sorbom, D. LISREL VI: Analysis of Linear Structural Relationships by Maximum Likelihood, Instrumental Variables, and Least Squares Methods. 3rd ed. Mooresville, IN: Scientific Software, 1984. Kraepelin, E. Dementia Praecox and Paraphrenia. Edinburgh: E. and S. Livingstone, 1919. Lenzenweger, M.F.; Dworkin, R.H.; and Wethington, E. Models of positive and negative symptoms in schizophrenia: An empirical evaluation of latent structures. Journal of Abnormal Psychology, 98:62-70, 1989. Lenzenweger, M.F.; Dworkin, R.H.; and Wethington E. Examining the underlying structure of schizophrenic phenomenology: Evidence for a three-process model. Schizophrenia Bulletin, 17515-524, 1991. Liddle, P.F. The symptoms of chronic schizophrenia: A re-examination of the positivenegative dichotomy. British Journal of Psychiatry, 151: 145-151, 1987. Long, J.S. Confirmatory Factor Analysis (Sage University Paper Series on Quantitative Applications in the Social Sciences, Series No. 07-033) Beverly Hills, CA: Sage Publications, 1983. Marsh, H.W.; Balla, J.R.; and McDonald, R.P. Goodness-of-fit indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 103:391410, 1988. Pogue-Geile, M.F.. and Harrow, M. Negative and positive symptoms in schizophrenia and depression: A follow-up. Schizophrenia Bulletin, 10:371-387, 1984. Ragin, A.B., and Oltmanns, T.F. Communicability and thought disorder in schizophrenic patients and other groups: A followup study. British Journal of Psychiatry, 150:494-500, 1987. Spitzer, R.L.; Endicott, J.; and Robins, L. The Schedule for Affective Disorders and Schizophrenia. New York: Biometrics Research, 1978. Tucker, L.R., and Lewis, C.A. A reliability coefficient for maximum likelihood factor analysis. Psychometrika, 38:1-10, 1973. Walker, E., and Harvey, P.D. Positive and negative symptoms in schizophrenia: Attentional performance correlates. Psychopathology, 19:294-302, 1986. Walker, E.; Harvey, P.D.; and Perlman, D. The positive and negative symptom distinction in psychosis: A replication and extension of previous findings. Journal of Nervous and Mental Disease, 176:359-363, 1988.

Empirical assessment of the factorial structure of clinical symptoms in schizophrenic patients: formal thought disorder.

Male schizophrenic patients (n = 142) were examined with a clinical assessment of their language dysfunctions with the Scale for the Assessment of Tho...
853KB Sizes 0 Downloads 0 Views