APPROVED: Craig S. Neumann, Major Professor Kenneth W. Sewell, Committee Member and

Director of Clinical Training Martin Bink, Committee Member David Tam, Committee Member Linda Marshall, Chair of the Department of

Psychology Sandra L. Terrell, Dean of the Robert B. Toulouse

School of Graduate Studies

THE STRUCTURE OF INSIGHT IN PATIENTS WITH PSYCHOSIS

Andrea R. Gonterman, B.S., M.S.

Dissertation Prepared for the Degree of

DOCTOR OF PHILOSOPHY

UNIVERSITY OF NORTH TEXAS

December 2004

Gonterman, Andrea R., The structure of insight in patients with psychosis. Doctor of

Philosophy (Clinical Psychology), December 2004, 69 pp., 10 tables, 2 illustrations, references,

128 titles.

Failure to acknowledge their mental illness occurs in approximately half of all psychotic

patients. Interest has been recently been refocused on insight (i.e., awareness of mental illness),

and its associations with treatment compliance and better prognosis. Researchers have called

into question the traditional factor structure of insight, instead viewing and defining it as a

multidimensional and continuous construct. While factor analytic research has suggested that

insight is an independent feature of psychotic disorders rather than a secondary manifestation of

psychotic symptoms, several factor analytic studies have identified only one higher-order factor.

Furthermore, a significant amount of the research literature has assessed insight or analyzed its

relationships using only a single insight score. The current study evaluated the structural model

of insight and assessed the associations between the different proposed dimensions of insight and

psychotic symptoms. One hundred and six participants recruited from both inpatient and

outpatient settings with a diagnosis of schizophrenia, schizoaffective disorder, psychotic disorder

NOS, or bipolar disorder with psychotic features were rated on David’s Schedule for Assessing

Insight-Expanded Version (SAI-E), Birchwood’s Insight Scale (IS), and the Brief Psychiatric

Rating Scale (BPRS) or the Positive and Negative Syndrome Scale (PANSS). Structural

equation modeling (SEM) was utilized to provide stringent, confirmatory statistical tests of the

hypothetical factor models while accounting for measurement error. Principal findings from the

current study were that the three factor model of insight was supported and that the insight

factors were meaningfully correlated to the two symptom factors. Moreover, the three factor

insight model provided significantly better fit than a single factor model of insight.

ii

TABLE OF CONTENTS

Page

LIST OF TABLES.......................................................................................................................iii LIST OF ILLUSTRATIONS....................................................................................................... iv Chapter

1. INTRODUCTION ................................................................................................ 1

Proposed Dimensions of Insight

Structural Equation Modeling (SEM)

Symptom Correlates of Insight

Statement of Problem 2. METHOD ........................................................................................................... 16

Participants

Materials

Procedure

Data Analysis 3. RESULTS ........................................................................................................... 25

Sample Characteristics

Descriptive Analyses

Analyses of the Combined Sample

Structural Equation Modeling (SEM)

Goodness of Fit 4. DISCUSSION..................................................................................................... 33

Analyses of the Combined Sample

Model 1

Model 2

Clinical Implications

Limitations

Future Directions REFERENCES ........................................................................................................................... 54

iii

LIST OF TABLES

Page

1. Studies Showing Significant Associations across Insight Dimensions with Symptoms ......................................................................................................................................... 43

2. Descriptive Statistics for the Manifest Variables for M1 & M2..................................... 45

3. Spearman’s Rho Correlations between Model 1 & Model 2 Proposed Manifest Variables......................................................................................................................................... 46

4. Model Fit Results............................................................................................................ 47

5. Standardized Parameter Latent Variables and Error Loadings: Model 1 ....................... 47

6. Standardized Parameter Latent Variables and Error Loadings: Model 2 ....................... 48

7. Standardized Parameter Latent Variables and Error Loadings: Model 3 ....................... 49

8. Model 1 Confirmatory Factor Analysis (CFA) Latent Variable Intercorrelations ......... 50

9. Model 2 Confirmatory Factor Analysis (CFA) Latent Variable Intercorrelations ......... 50

10. Model 3 Confirmatory Factor Analysis (CFA) Latent Variable Intercorrelations ......... 51

iv

LIST OF ILLUSTRATIONS

Page

1. Model 1: Three Factor Model of Insight using SAI-E Items as Manifest Variables...... 52

2. Model 2: Three Factor Model of Insight and Psychopathology using SAI-E, BPRS, and PANSS Items as Manifest Variables .............................................................................. 53

1

CHAPTER 1

INTRODUCTION

Until the late 1980s, insight (or awareness of illness) into psychotic illness had received

little systematic examination (Collins, Remington, Coulter, & Birkett, 1997; David, 1990).

However, the concept of insight has often been considered highly relevant to mental health

professionals who have to treat psychotic individuals with perceptions of self that are grossly

different from that of his or her community and culture (Amador & Kronengold, 1998). While

the recognition of having (or having had) a mental disorder is an important characteristic of

insight, it has been suggested that insight is not simply present or absent (i.e., categorical), but

appears to be a continuous and multidimensional construct (Amador et al., 1993, 1994; Amador

& David, 1998; David, 1990; McEvoy et al., 1989a). Important dimensions of insight include

not only the awareness or recognition of having a disorder, but also the realization of need for

treatment, as well as the ability to re-label unusual mental events as pathological and attribute

appropriate causes for mental illness (Amador & Kronengold, 1998; David, 1998). Each of these

three primary dimensions (awareness, need for treatment, and ability to re-label/attribute) are

considered important continuous factors which make up the higher order construct of “insight.”

Lack of insight is a distinctive feature of severe psychotic disorders (Amador & David,

1998; Carpenter et al., 1979; Lewis et al., 1934). Over half of all psychotic inpatients,

particularly those with schizophrenia, fail to acknowledge their mental illness (Amador et al.,

1994; Kemp & David, 1996; McEvoy, Schooler, Friedman, Steingard, & Allen, 1993a;

Neumann, Walker, Weinstein, & Cutshaw, 1996; Walker & Rossiter, 1989; Young et al., 1998).

Slightly higher rates of insight have been reported in outpatients, though it should be noted that

these participants were recruited from psychoeducational programs (Dickerson, Boronow,

2

Ringel, & Parente, 1997; McEvoy, Freter, Merritt, & Apperson, 1993b). Others have found

higher insight rates for outpatients were confined to the dimension of awareness for the need for

treatment (Kim, Sakamoto, Kamo, Sakamura, & Miyaoka, 1997).

As part of the DSM-IV field trials, Amador et al. (1994) found that nearly 60% of the

patients with schizophrenia evidenced moderate to severe unawareness of having a mental

disorder. Though patients with schizophrenia produced the poorest insight scores, average

awareness scores were not significantly different between patients with schizophrenia,

schizoaffective, and bipolar disorder. Pini, Cassano, Dell’Osso, and Amador (2001) found that

patients with schizophrenia had significantly lower insight scores than patients with

schizoaffective disorder or unipolar depression with psychosis, but that the schizophrenic sample

did not differ significantly from the patients with bipolar disorder. In contrast to these earlier

studies, Daneluzzo et al. (2002) reported that 65.1% of the patients with schizophrenia in their

sample had lack of insight (LOI) scores (based on use of the single PANSS item) of mild or

greater compared to only 32.4 % of the patients with bipolar disorder. However, the LOI item

was the only PANSS item to highly discriminate between these two diagnostic groups.

Meanwhile others have not found a significant difference between diagnostic groups regarding

level of insight (David, Buchanan, Reed, & Almeida, 1992; David et al., 1995).

Understanding the phenomenon of insight in psychosis is critical given the impact

awareness of illness has on patients’ clinical outcomes. In a large study of recent-onset, mixed

psychotic patients, lack of insight had a significant relationship with time spent in the hospital

and living independently regardless of the specific psychotic diagnosis or other prognostic

indicators (Van Os et al., 1996). Also, a positive association between level of insight and

treatment compliance has repeatedly been reported in the literature (Bartko, Herczeg, & Zador,

3

1988; David, 1998; David, Buchanan, Reed, & Almeida, 1992; McEvoy, 1998; McEvoy et al.,

1989b; Sanz, Constable, Lopez-Ibor, Kemp, & David, 1998; Van Putten, 1974). Thus, decreased

insight predicts poorer treatment compliance and prognosis (David, 1998: McEvoy, 1998;

McEvoy et al., 1989b).

Significant relationships have also been noted between insight and contact with mental

health services (Sanz et al., 1998), global functioning (Amador et al., 1994), and higher social

class (David et al., 1995). Similarly, good insight has been associated with better vocational

rehabilitation (Lysaker & Bell, 1994). Finally, a measure of social knowledge developed by

Cutting and Murphy (1988, 1990) has been found to be significantly correlated to insight

(Gonterman, 2002; McEvoy et al., 1996).

Presently, there are several reliable and valid multidimensional assessments of insight

that utilize a semi-structured interview format. Three of the most widely used include: (a) the

Scale to Assess the Unawareness of Mental Disorder (SUMD; Amador et al., 1993) with higher

scores indicating poorer insight; (b) the Schedule for the Assessment of Insight – Expanded

Version (SAI-E; Kemp & David, 1997); and (c) the Insight and Treatment Attitudes

Questionnaire (ITAQ; McEvoy et al., 1989). Higher scores indicate better insight on the SAI-E

and ITAQ. The ITAQ has been criticized as not being truly multi-dimensional since it consists of

only two dimensions (awareness of general mental illness and treatment compliance) with eight

of its eleven items mentioning treatment and six of these items specifically measuring treatment

compliance (Cuesta, Peralta, and Zarzuela, 2000; David, 1998; Weiler, Fleisher, & McArthur-

Campbell, 2000).

Two self-report measures have been reported as psychometrically sound (Birchwood et

al., 1994; Markova & Berrios, 1992) and are significantly correlated to the interviewer-based

4

assessments (Garavan et al., 1998; Gonterman, 2002; Sanz et al., 1998). These self-report

measures are also based on a continuous and multidimensional approach. Given that insight is

related to disturbances in self-perception, Strauss (1998) has suggested that research involving

both subjective (self-report) and more objective (interview) assessments as an area or further

research. Indeed, little research has been conducted comparing clinician ratings of insight with

client self-reports to determine whether the dimensions gathered from each correlate to

symptoms and cognitive factors in a similar manner and if type of assessment impacts the factor

structure of insight.

Aside from the impact on clinical course and outcome, insight has been most often

associated with psychotic patients’ symptoms of psychopathology. For instance, Peralta and

Cuesta (2001) argue that there is evidence to support eight major dimensions of psychopathology

in schizophrenia, including a dimension for lack of insight. Moreover, several exploratory factor

analytic studies suggest that insight loads on an independent factor rather than merely being a

secondary manifestation of psychotic symptomatology (David, 1998). In their study of 115

inpatients diagnosed with schizophrenia, Peralta and Cuesta (1994) found that three proposed

dimensions of insight (awareness of illness and specific symptoms and need for treatment) all

loaded onto their own factor, separate from positive, negative, and disorganized symptoms. In

another study, Cuesta and Peralta (1994) found lack of insight to be a factor independent of

several measures of neuropsychological functioning commonly associated with schizophrenia, as

well as, positive and negative symptoms. Research with a sample of 150 patients with recent-

onset, mixed psychotic diagnoses found lack of insight formed an independent factor from

several psychopathological symptoms (Van Os et al., 1996).

5

Furthermore, changes in psychotic symptoms are not uniformly associated with changes

in insight (Carroll et al., 1999). Few significant relationships between insight and

psychopathology remained stable after six months, and surprisingly those that did were

relationships between negative symptoms and disorganization with lack of insight (Cuesta,

Peralta, & Zarzuela, 2000). Additionally, these researchers found no predictive value for

variability in psychopathological symptoms for the dimensions of insight. Finally, a review of

the literature suggests that insight is largely independent of global severity of psychopathology,

with the correlations between insight and total symptom scores being modest at best (David,

1998). Thus, there is reasonable support for the theory that insight is an independent feature of

psychotic illness rather than merely a secondary manifestation of psychotic symptoms. An

assumption that remains untested, however, is whether the insight construct is multidimensional

and can be represented by separate (but correlated) factors.

Proposed Dimensions of Insight

As mentioned briefly above several dimensions of insight have been proposed, but the

dimensions most commonly discussed in the literature are those suggested by David (1990) and

Amador (1991). David (1990) has suggested there are three main components of insight:

awareness of illness, ability to relabel psychotic experiences correctly, and treatment compliance.

The Schedule for Assessing Insight (SAI) and the SAI-E were created to measure these domains

along with a supplementary question referred to as the hypothetical contradiction, which asks

patients how they feel when others don’t believe them in order to determine whether this cues

patients that they are getting sick.

The SUMD (Amador et al., 1991, 1994) assesses domains of insight including global

insight, insight into symptoms, need for treatment, and social consequences of the disorder.

6

Seventeen specific signs and symptoms are assessed providing four additional scales for current

and past awareness and current and past attributions.

The subscales (which represent separate dimensions of insight) for each of the insight

measures are intercorrelated (Amador et al., 1993; Birchwood et al., 1993; David et al., 1992).

For example, Amador et al. (1993) reported that the SUMD subscales representing awareness

and attribution for present and past episodes were moderately to strongly correlated (r = .55, p =

.001 and r = .67, p = .001, respectively). David, Buchanan, Reed, and Almeida (1992) related

correlations between the insight subscales from the SAI that ranged from .75 (p < .001) for the

hypothetical contradiction and relabeling to .26 (p < .02) for treatment compliance and relabeling

of psychotic symptoms. Given the generally high inter-correlations between the subscales of

insight, it is not surprising that several studies conducting exploratory factor analysis, using

principle component analysis with orthogonal rotation, suggest that the separate dimensions of

insight all load onto a single higher-order factor (Birchwood et al., 1994; David et al., 1992;

McEvoy et al., 1989a; Peralta & Cuesta, 1994).

Based upon these results, one could reasonably suggest that insight is a unidimensional

construct. However, supporters of the multidimensional approach have countered that the low

correlations found between some dimensions of insight support the hypothesis that insight is not

unitary, but made up of two or more related but partially independent elements (David et al.,

1992; David, 1998). Amador et al. (1993) has argued that moderate to strong correlations

between awareness of illness and attributions should be expected since attribution scales are

hierarchical, and the findings that these scales are not more highly correlated supports the idea

that appropriate attributions is a phenomenon independent of awareness. Additionally, research

findings that impaired awareness can be limited to specific dimensions of insight and that

7

percentage of improvement from baseline can vary across dimensions could be used to support a

multi-dimensional approach (Amador et al., 1994; Kemp, Hayward, & David, 1996).

Another argument that could be made for the multidimensional approach involves

research findings that the dimensions of insight show differential relationships with symptoms

and neurocognitive correlates, but these results, which will be reviewed in the sections related to

correlates of insight, are not entirely convincing.

In terms of factor analytic research on the dimensions of insight, one study used principal

component analysis with an orthogonal rotation at the level of insight scales and found evidence

for two insight factors. Cuesta et al. (2000) entered 14 subscale scores from 75 psychotic

patients with mixed diagnoses and found a factor representing general awareness and another

reflecting attitudes to treatment, which accounted for 76.9% of the variance. These factors

demonstrated considerable overlap, though, with an absolute correlation of .42. As mentioned

above, these two factors were found again when the insight scores were included in a factor

analysis that also included psychopathological symptoms.

One of the main limitations of the exploratory factor analytic studies mentioned above is

that they did not examine the factor structure of insight at the level of individual items. To date,

only one study has explored the “dimensions” of insight utilizing individual items. Kim et al.

(1997) administered the SAI (David, 1990) to 63 patient diagnosed with schizophrenia. Their

results suggested three factors which corresponded with the three hypothesized dimensions of

insight (awareness of illness, need for treatment, relabeling psychosis).

While the findings of Kim et al. (1997) and Cuesta et al. (2000) do support the

multidimensional approach to insight, a definitive statement cannot be made due to several

limitations in the analytical methods used in these two studies. First, both studies relied upon

8

exploratory and not confirmatory factor analysis, and thus a stringent a priori statistical test of

the factor structure of insight was not possible (Bentler, 1995). Second, Cuesta et al. (2000)

relied upon their insight subscales for their factor analysis, rather than the individual items which

made up each scale, which could have potentially obscured the relationship between individual

insight items and the factors they were designed to represent. Finally, the Kim et al. (1997)

study is limited in that the authors relied upon a principal component analysis (PCA) with

orthogonal rotation. The PCA is based upon the assumption of perfect calculation of the

variables from the components (Gorsuch, 1983). In addition, PCA is often completed with the

assumptions that the factors are uncorrelated (Gorsuch, 1983; Peralta & Cuesta, 2001).

As previously mentioned, the dimensions of insight have been found to be significantly

intercorrelated, and thus the assumption of orthogonal insight factors seems unrealistic.

Furthermore, it is reasonable to suggest that there would be some degree of measurement error

when clinicians are assessing a high base rate condition such as lack of insight among psychotic

patients (Meehl, 1973). Thus, there is currently a gap in the literature regarding statistical

analysis of the multidimensional (correlated) factors of insight, which also takes into account

measurement error.

Confirmatory factory analysis (CFA), which has been recognized as one of the best

methods for establishing construct validity (Embretson & Hershberger, 1999), could help address

some of the limitations discussed above. Moreover, if the structure of insight was found to be

‘truly’ multidimensional, and it was determined that these dimensions were differentially related

to psychotic patients’ symptoms of psychopathology, then such findings would have important

treatment implications. As it turns out, latent variable model-based analytical methods such as

9

CFA have been used with good success to study numerous constructs in psychology and related

fields (Bentler, 1995; Embretson & Hershberger, 1999).

Structural Equation Modeling (SEM)

SEM is a latent variable method of analysis, which tests the structural theories of various

phenomenon. It has also been referred to as causal modeling, covariance structure analysis, path

analysis, and linear structural relations. “A structural theory is intended to represent the ‘causal’

process that gives rise to the correlations among the observed variables” (Bentler, 1988, p.317).

The intent of causal modeling is to rule out implausible connections. Additionally, SEM can be

used to confirm the factor structure of a phenomenon, because the researcher has complete

control over the specification of indicators for each construct. Finally, SEM allows for statistical

tests to assess the goodness-of-fit of the proposed model, which is not possible with principal

component analysis or factor analysis (Hair, Anderson, Tatham, & Black, 1995).

In conducting SEM, a distinction is made between measured (manifest) (MVs) and latent

variables (LVs). A LV represents a theoretical construct (e.g., insight) and represents MVs (e.g.,

subscale scores or items from self-reports or interview ratings). It is hypothesized that

correlations between observed variables (MVs) are attributable to the LVs (Bentler, 1980).

Moving to the level of latent variable analysis also has certain advantages. For instance, MV

causal models tend to be less reliable than LV models due to the bias that can be introduced by

the level of measurement error contained in each variable, which strongly influences the

correlation coefficient between two MVs. In contrast, LV models can represent the common

factor variance separately from the error (plus unique) variance in a set of MVs representing a

particular theoretical construct (Bentler, 1980). Finally, LV approaches allow investigators to

hypothesize on how a pattern of correlations among a set of observed variables can be explained

10

in terms of a smaller set of factors (LVs), and then statistically estimate the fit between the model

and the observed (manifest) data (Bentler, 1995).

The primary statistical issues within SEM are to estimate the unknown parameter values

of the model (e.g., factor loading and correlations) and then determine the goodness-of-fit of the

model to the sample data of measured variables (Bentler, 1980). Goodness-of-fit indices such as

chi-square, normed-fit-index, and others, indicate how likely the proposed model of LVs

represents the causal structure of the MVs. If there is a good fit, the model is considered a

plausible representation of the hypothesized causal structure of the observed variables.

Symptom Correlates of Insight

Between the late 17th and early 19th centuries, lack of insight was viewed as being

synonymous with mental illness and interest in exploring the implications of varying degrees of

insight in the mentally ill would have been considered nonsensical (Berrios & Markova, 1998).

Even into the present century, the widely held belief that insight and acute psychosis were

incompatible has hindered research and assessment of insight (David, 1990). Within the last

fifteen years the assessment of insight has become more rigorous, yet the correlates of insight

have yet to be fully articulated (David & Amador, 1998).

The view that symptoms of severe psychopathology are redundant with poor insight or

that such symptoms precludes insight has not been supported (Cuesta & Peralta, 1994; David,

Buchanon, Reed, & Almeida, 1992; David et al., 1995; McEvoy et al., 1996). David et al.

(1992) found a correlation of r = -.31 between the Present State Examination (PSE; Wing,

Cooper, & Sartorius, 1974) total score and the SAI total score. Similarly, comparing the Brief

Psychiatric Rating Scale (BPRS; Overall & Gorham, 1962) total score with the SAI total score,

Aga, Agarwal, and Gupta’s study (as cited in David, 1998) found a correlation of r = .28.

11

McEvoy et al. (1996) did not find a significant correlation between the BPRS total symptom

score and insight. Cuesta and Peralta (1994) demonstrated that insight was not correlated with

either the total score for the Scale for the Assessment of Negative Symptoms (SANS;

Andreason, 1984a) or the total score for the Scale for the Assessment of Positive Symptoms

(SAPS; Andreasen, 1984b). Finally, David et al. (1995) determined that the mean PSE total

score was not correlated with the PSE insight item. The sum of these findings suggests that

insight is largely independent of global psychopathology severity (David, 1998). As such,

investigators have focused on more specific symptom dimensions.

Positive symptoms and insight. One particular area of interest has been the

meaningfulness of the relationship between insight and positive, or psychotic, symptoms.

Positive symptoms are phenomenon present in individuals with psychotic disorders (e.g.,

schizophrenia) that are not present in healthy individuals and include such experiences as

delusions and hallucinations. The majority of investigations have found significant, if frequently

moderate, inverse relationships between insight and positive symptoms (Amador et al., 1994;

Carroll et al., 1999; Collins, Remington, Coulter, & Birkett, 1997; David et al., 1992; Dickerson

et al., 1997; Kemp & Lambert, 1995; Lysaker, Bell, & Bioty, 1995; Markova & Berrios, 1992;

McEvoy, Schooler, Friedman, Steingaard, & Allen, 1993; Michalakeas et al., 1994; Takai,

Uematsu, Hirofumi, Sone, & Kaiya, 1992; Young, Davila, & Scher, 1993). While few studies

have failed to find such a relationship (but see Cuesta & Peralta, 1994, for an example), some

have demonstrated a dissynchrony between improvement in positive symptoms and insight

(McEvoy et al., 1989b).

While research on specific symptom dimensions has been helpful, there are problems

with simply studying the associations of insight and global positive symptoms. First, there is

12

considerable variability in the number of symptoms included in scales designed to assess

positive, as well as negative, symptoms and there are still symptoms that are inconsistently

classified between these two categories (Walker & Lewine, 1988). Additionally, positive

symptoms are too frequently applied to schizophrenia alone while they are applicable to other

psychotic disorders. Finally, David (1998) argues that continued attempts to understand the

relationship between insight and different psychotic symptoms will hopefully further our

understanding of both insight and psychotic symptomatology.

The current body of literature indicates that several specific symptoms of psychosis are

significantly associated with insight. For instance, moderate to large inverse correlations have

been found between insight and: delusions (Dickerson et al., 1997; Neumann, Walker,

Weinstein, & Cutshaw, 1996; Sanz et al., 1998), grandiosity (Kemp & Lambert, 1995: Neumann

et al., 1996; Neumann & Walker, 1998; Sanz et al., 1998; Van Putten, Crumpton, & Yale, 1976),

disordered thought (Amador et al., 1994; Kim, Sakamoto, Kamo, Sakamura, & Miyaoka, 1997),

and disorganized behavior (Amador et al., 1994). Interestingly, researchers have found no

relationship between hallucinations and insight (David et al., 1992; Nayani & David, 1996),

while others have found that hallucinatory behavior was related to increased insight (Neumann et

al., 1996). These results suggest that it may be important to study positive symptoms separately,

and that not all positive symptoms are specifically associated with deficits in insight.

Affective symptoms and insight. Another area of interest is the association between

specific affective symptoms and insight. Several researchers have noted an inverse relationship

between elated mood (or more specifically mania) and insight (Bartko et al., 1988; David, 1990;

Heinrichs, Cohen, & Carpenter, 1986; Kemp & Lambert, 1995; Sanz et al., 1998; Van Putten et

al., 1976). Additionally, the majority of studies have found positive correlations between

13

depression, and to some extent anxiety, and insight in psychotic patients (Amador et al., 1994;

Carroll et al., 1999; David et al., 1995; Dickerson et al., 1997; Kemp et al., 1995; Michalakeas et

al., 1994; Mintz, Addington, & Addington, 2004; Moore et al., 1999; Neumann et al., 1996;

Peralta et al., 1994; Sanz et al., 1998; Smith et al., 1998; Smith et al., 2000). Across these

studies, (absolute) correlations were found to range from .39 to .58 and averaging approximately

.48.

Symptoms and the factor structure of insight. It could be argued that support is lent to a

multi-dimensional model of insight if it were determined that specific symptoms were associated

with specific dimensions of insight. Table 1 is a sampling of research studies that have explored

the relationship between proposed dimensions of insight and symptoms correlates. Note that this

sampling is not meant to be exhaustive, but rather represent those studies that have been

published since the introduction of more modern conceptualizations of insight.

Overall, a review of Table 1 suggests consistent relationships between symptoms and the

different subscales of insight, at least in patients who have recently experienced an exacerbation

of symptoms. More specifically, the findings indicate a positive association between (better)

insight and depression and inverse relationships between the other symptoms, delusions and

disorganized symptoms in particular. It should be noted, though, that none of the studies found a

significant relationship between negative symptoms and relabeling and that the relationship

between negative symptoms and treatment compliance is ambiguous with two studies suggesting

an inverse relationship while a third reported a positive relationship. Thus, the relationship

between negative symptoms and the insight dimensions is mixed. Notably, some longitudinal

research (Cuesta, Peralta, & Zarzuela, 2000) has suggested that a stable relationship exists

14

between insight and negative symptoms, though further longitudinal research will be necessary

to replicate these limited findings.

Statement of Problem

Interest has increased in recent years on the nature of insight (or awareness of mental

disorder) in psychotic disorders. What was once viewed as a secondary feature fundamental to

psychosis, insight is now considered an independent factor separate from symptoms and

neurocognitive deficits of psychosis, and this notion has been supported by research (Cuesta &

Peralta, 1994; David 1998; Peralta & Cuesta 1994; Van Os et al., 1996). What is less clear is

whether insight is a truly multi-dimensional phenomenon as it is currently defined in the

literature, and whether symptoms of psychopathology are differentially related to these

dimensions.

Several studies have explored this issue with exploratory factor analysis employing

principle component analysis with orthogonal rotation with results indicating that all the separate

dimension subscales load onto one higher order factor (Birchwood et al., 1994; David et al.,

1992; McEvoy et al., 1989a; Peralta & Cuesta, 1994). It is worthy to note that all of these

studies entered subscale scores into the analysis. Since some insight subscales have been found

to be highly intercorrelated, something principle component analysis often does not allow for,

this failure to find dimensions of insight is not conclusive.

Other research has lent support to the multi-dimensional approach to insight. One study

utilizing significantly more subscale scores (fourteen versus three to five in the latter mentioned

studies) did find two dimensions for insight: general awareness and treatment compliance

(Cuesta et al., 2000). The work of Kim et al. (1997), which explored the factor structure of

insight at the level of individual items, also supports a multi-dimensional model with their

15

findings of three insight domains that closely resemble the dimensions proposed by David

(1990).

To date, all research into the factor structure of insight has been exploratory with

limitations such as bias from exploratory statistical methods that do not allow for error terms or

high intercorrelations amongst the data entered. Therefore, further research is needed utilizing

statistical methods that can test a proposed multi-dimensional model of insight while allowing

for the intercorrelations known to be present among insight subscales and controlling for the

effects of error.

The current study proposes to test the multi-dimensional model of insight underlying the

SAI-E using structural equation modeling. The SAI-E is assumed to have three main

dimensions: awareness of illness, relabeling of symptoms as pathological, and treatment

compliance. The current model of insight tested can be found in Figure 1.

Based on the literature review and conclusions drawn, the following model hypotheses

were tested:

Model hypothesis 1. The three-factor model of insight currently proposed will have an

acceptable goodness of fit score for the current sample. The goodness of fit score for the three-

factor model will be significantly better than the goodness-of-fit score for the one factor (null)

model of insight.

Model hypothesis 2. A five factor model will result in good model fit when the three insight

factors are modeled in conjunction with a depression symptom factor and a delusion symptom

factor. The depression factor will correlate positively with insight, while the delusion factor will

correlate negatively with insight.

16

CHAPTER 2

METHOD

Participants

The sample for this study consists of 106 patients with one of the following diagnoses:

schizophrenia, schizoaffective disorder, psychotic disorder not otherwise specified (NOS), or

bipolar disorder with psychotic features. Patients were recruited from the North Texas Veterans

Health Care System (NTVHCS), Terrell State Hospital (TSH), and Vernon State Hospital

(VSH). The current study is part of a large project on the biological and cognitive factors

associated with insight in patients with psychotic symptoms.

The NTVHCS, formerly known as the Dallas Veterans Affairs Hospital, serves veterans

and their qualified family members living in the Dallas/Forth Worth area, as well as the

surrounding rural areas. The sample from this site was comprised of both inpatients

(hospitalization was voluntary) and outpatients with diagnoses of schizophrenia, schizoaffective

disorder, or psychotic disorder NOS.

TSH located in Terrell, Texas is a modern psychiatric facility serving 22 counties in the

northeastern part of Texas. All participants in this sample were recruited from locked acute

inpatient wards, at least some of whom had been placed there under civil commitment.

Diagnoses present in this sample include schizophrenia, schizoaffective disorder, psychotic

disorder NOS, and bipolar disorder.

The final study site, VSH, is a maximum security forensic facility to which patients are

committed for various criminal offenses. Participants with diagnoses of schizophrenia or

schizoaffective disorder were recruited from a ward of relatively stabilized patients. This unit is

17

designated for those patients who have earned the highest levels of privileges available to that

population.

The presence of a recent exacerbation (within two weeks of testing) of psychotic

symptoms served as the primary inclusion criteria for the samples collected at the NTVHCS and

TSH. Those patients found to have at least one psychotic symptom (i.e. hallucination, delusion,

thought disorder, grandiosity) occurring in absence of delirium were included in the data set.

Psychotic symptoms were measured and a diagnosis was determined based upon the results of

the either the Brief Psychiatric Rating Scale (BPRS) (Terrell and Vernon State Hospitals) or the

Positive and Negative Syndrome Scale (PANSS) (North Texas Veterans Health Care System),

the Positive and Negative Affect Scale (PANAS), chart notes, and past diagnoses. Exclusion

criteria included participants with the following: substance-induced psychotic disorder, psychotic

disorder due to a general medical condition, delirium, dementia, amnestic and other cognitive

disorders, dual diagnosis of mental retardation, epilepsy, thyroid disease, an inability to speak

English, or those receiving electroconvulsive therapy.

Materials

Symptoms. The BPRS, Expanded Version (4.0) was used to assess the presence of

psychopathological symptoms which were then cross referenced with the participant’s charts.

The BPRS is a semi-structured interview consisting of 24 items rated on 7-point ordinal scales.

The first fourteen items are rated based on self-reports and observed behavior; while the

remaining items are rated based on observed behavior and speech (Ventura et al., 1993).

The BPRS has been used extensively in American and European psychiatric research for

almost three decades and has been found to be a reliable instrument (Ventura et al., 1993).

Ventura et al. (1993) studied interrater reliability on the expanded BPRS for researchers with

18

advanced training degrees and those without. Interviewers participating in the study were trained

in the guidelines for rating as well as provided with structured questions to illicit the necessary

information from interviewees. They reported Median ICCs of .81 and .83 respectively.

Interrater reliability was measured for the first 16 items (Overall & Gorham, 1962) on a group of

newly admitted schizophrenic patients and was found to range from .56 to .87 with 10 of the

items falling at .80 or above. Suggestions for interview style were provided, but the raters in this

study did not use a structured interview format.

The BPRS has also been found to be associated with other instruments used to assess

psychopathology. Newcomer et al. (1990) reported that the BPRS depression subscale was

highly correlated with the Hamilton Rating Scale for Depression (HRSD) total score (rho = .80).

To examine the potential impact of single symptom factors being defined by different

combinations of BPRS symptoms, Nicholson, Chapman, and Neufeld (1995) pulled several

different BPRS definitions of positive and negative symptoms from research literature and

compared them to each other and to the Schedule for the Assessment of Negative Symptoms

(SANS) and the Schedule for the Assessment of Positive Symptoms (SAPS; Andreasen, 1984).

They found all the negative symptom definitions to be highly correlated with each other and the

SANS, and all the positive symptoms definitions to be highly correlated with each other and the

SAPS. Czobor, Bitter, and Volavka (1991) found the SANS and the BPRS to be highly

intercorrelated with the SANS composite score being highly redundant with the anergia factor of

the BPRS. Another study also found the SANS to be highly correlated with the BPRS

withdrawal-retardation scale, and after a comparison of the interrater reliabilities proved

unfavorable for the SANS, deemed little was gained by using the SANS instead of the BPRS. In

a comparison of the BPRS and the Positive and Negative Syndrome Scale (PANSS; Kay,

19

Fiszbein, & Opler, 1987), Bell, Milstein, Beam-Goulet, Lysaker, and Cicchetti (1992) found the

negative syndromes were correlated at .82, the positive syndromes correlated at .92, the general

scales correlated at .61, and the scale total scores were correlated .84.

The BPRS has been previously utilized in insight research to provide measures of

psychopathology and specific symptoms (McEvoy et al., 1989a; McEvoy, Applebaum,

Apperson, Geller, & Freter, 1989c; McEvoy et al., 1993; Neumann et al., 1997; Sanz et al., 1998;

Takai et al., 1992; Young et al., 1998).

Symptoms were measured in the veteran sample using the PANSS. The PANSS is a 30-

item, 7-point severity scale designed to measure positive and negative syndromes as well as

general psychopathology among inpatients with schizophrenia. It follows the same general

format as the BPRS (Overall & Gorham, 1962) and Psychopathology Rating Schedule (Singh &

Kay, 1975) with the addition of strict operational criteria for the interview and detailed rating

criteria for each level of severity (Kay, Opler, & Lindenmayer, 1988). The PANSS, with its

combination of measures of psychopathology as well as an item measuring lack of insight, has

been widely used in the recent insight literature (e.g. Carroll et al., 1999; Collins, Remington,

Coulter, & Birkett, 1997; Debowska et al., 1998; Kemp & Lambert, 1995; Schwartz, 1998).

Kay et al. (1988) reported interrater reliability on the individual items ranging from .69 to

.94 and mean interrater correlations for the four PANSS scales ranging from .83 to .87. Bell,

Milstein, Beam-Goulet, Lysaker, &Cichetti (1992) also reported interrater reliability for

individual items ranging from .54 to .93. To determine criterion-related validity, the same

authors compared the PANSS to the SAPS and SANS and found a correlation of .77 between

both the positive scales and negative scales of each test. Norman, Malla, Cortese, and Diaz

(1996) reported high convergent validity between the positive and negative subscales of the

20

PANSS and the summary total scores of the SAPS (.91) and SANS (.88). The PANSS has also

been found to reliable and valid for use with outpatients diagnosed with schizophrenia or major

mood disorders (Purnine, Carey, Maisto, & Carey, 2000).

Insight. Patient insight was measured using the Schedule for the Assessment of Insight –

Expanded Version (SAI-E; Kemp & David, 1997). This measure is a semi-structured interview

where the number of questions answered is determined by the answers given to key questions.

The number of items answered on the test can range from six to nine. The first six items are

rated on a 3-point scale with ratings as follows: 2 = often; 1 = sometimes; 0 = never. Items 7 and

8 regard specific symptoms, and are scored on a 5-point scale with 4 = to full awareness and

correct attribution and 0 = no awareness or bizarre/delusional attribution. Item 9 assesses the

client’s response to a hypothetical contradiction of his or her delusions or hallucinatory

experiences. This item is rated on a 5-point scale with the following ratings: 0 = They’re lying; 1

= I’m still sure despite what others say; 2 = I’m confused and don’t know what to think; 3 = I

wonder whether something’s wrong with me; and 4 = That’s when I know I’m sick. The SAI-E

also contains three items regarding treatment compliance that are completed by the patient’s

primary nurse. The maximum score is 24 points. The SAI-E has been demonstrated to have

high concurrent validity with other measures of insight, namely the insight question of the

PANSS (Kay et al., 1987) (r = .895), the Insight and Treatment Attitudes Questionnaire (ITAQ,

McEvoy et al., 1989a) (r = .845), and the Schedule for the Assessment of Insight (SAI) (r = .977)

(Sanz et al., 1998).

Additionally, patient insight was measured with the Birchwood’s Insight Scale, a self-

report insight scale for psychosis (Birchwood et al., 1994). The Birchwood’s Insight Scale

consists of eight uncomplicated and direct statements that patients rate on a 3-point scale (agree,

21

disagree, unsure) and can be completed quickly by even seriously disturbed patients. The test

items were designed to assess each of the three dimensions of insight advocated by David

(1990): awareness of illness, ability to relabel symptoms, and awareness of need for treatment.

Research on the reliability and validity of the Birchwood’s Insight Scale found the test to

have a high internal consistency (Cronbach’s alpha = .75) and high test-retest reliabilities (.90)

over a 1-week interval (Birchwood et al., 1994). Birchwood et al., (1994) also demonstrated the

Insight Scale to have adequate construct, criterion, and concurrent validities and sensitivity to

individual difference and change.

Procedure Patient charts and hospital staff were consulted to ascertain which patients were

diagnosed with a psychotic disorder and did not meet any of the exclusion criteria. These

patients were then approached in a public area of the unit or contacted by phone (in the case of

outpatients), told about the study, and asked if they would be interested in participating. If they

agreed to participate, testing times were arranged that were convenient for the patients. Testing

typically began immediately following completion of the informed consent.

All participation was on a voluntary basis. Informed consent was obtained from each

participant after the procedures involved were explained. Volunteers at TSH and the NTVHCS

were reimbursed for their participation. Patients at TSH were compensated with $5.00 voucher

books that could be used at the hospital canteen, and NTVHCS were paid $20.00 cash.

The symptom and insight assessments utilized for the present study were conducted as

part of a larger assessment battery also including neurocognitive and social knowledge

assessments. Testing was conducted in one or two sessions and lasted between 1 ½ to 3 hours.

22

Data Analysis

Structural equation modeling (SEM). SEM was utilized to test the proposed insight

model (see figure 1). SEM allows the researcher to determine the plausibility of the proposed

theoretical model within a particular population by testing the specified causal structure among a

set of variables. The EQS Structural Equations Program published by Multivariate Software,

Inc. was utilized to perform the SEM analyses for this study (Bentler, 1995). The EQS program,

developed by Bentler (1989), uses generalized least squares (GLS) procedures or maximum

likelihood (ML) procedures to examine the patterns of relationships among latent variables. It

incorporates a mathematical and statistical approach to the analysis of linear structural

relationships using matrix algebra and includes parameter estimates and several goodness of fit

indices. EQS generates a matrix specification and designates independent and dependent

variables in accordance with the representation system of Bentler and Weeks (1980).

EQS provides several goodness of fit indices indicating the degree of “fit” for a proposed

model to the sample data. The chi-square test serves as a test of the null hypothesis. A small

chi-square and large p value would indicate that the null hypothesis is correct and the model is a

good representation of the causal structure for the observed data. Use of chi-square can be

problematic, though, due to its sensitivity to sample size and multinormality. With large

samples, small non-meaningful discrepancies between the model and data can result in the

rejection of a model that is a good representation of the data. In addition, chi-square assumes

multi-normality and thus can result in the rejection of a model due to the data distribution not

being multi-normal rather than due to conceptual inaccuracies. As a result of these limitations,

the use of additional goodness of fit indices is recommended (Bentler, 1989).

23

Three additional indices are available through the EQS program will also be used to

address goodness of fit. These indices are the normed fit index (NFI; Bentler & Bonnett, 1980),

the nonnormed fit index (NNFI; Bentler & Bonnett, 1980) and the comparative fit index (CFI;

Bentler, 1980). These fit indices tend to provide an accurate assessment of how well the model

fits the data, regardless of sample size. Large values such as .9 and higher indicate a good fit,

while values of .7 and less indicate the model is a poor fit for the data.

For those instances when multinormality may not be met, EQS can compute robust

standard errors and test statistics. Chou, Bentler, and Satorra (1989) suggest that these robust

statistics may offer more reliable construct equations, variances, and covariances when the

distributional assumptions of normality are not fully met. This EQS feature provides the option

of an additional fit index, the robust comparative fit index (RCFI), based on the calculation of the

CFI fit index utilizing robust parameter values. Additional options available to address concerns

regarding assumptions of normality are the Satorra-Bentler chi-square (a modification of the

standard goodness-of-fit test yielding distributional behavior that more closely approximates the

chi-square) and Mardia’s (1974) kurtosis test of multivariate normality.

The root mean square residual (RMR) and the root mean square error of approximation

(RMSEA) are also available estimates of model fit. The RMR is the square root of the average

squared difference between each sample and estimated covariance divided by all of the variables

in the covariance matrix (Ullman, 1996). A small RMR value is associated with better fit. A

benefit of the RMSEA is that it provides an estimation of the lack of fit of the model to the

estimated population covariance matrix (Browne & Cudeck, 1993). Thus, a value of zero would

indicate an exact fit of the model while values of .05 or less would indicate a close fit of the

model. The overall fit of the proposed models in this research will be assessed using the Satorra-

24

Bentler chi-square, CFI, RCFI, standardized root mean squared residual (SRMR), and the

RMSEA.

25

CHAPTER 3

RESULTS

Sample Characteristics

Data was collected from 110 participants of whom 106 were included in the final

analyses due to completeness of their data and appropriateness of diagnosis. The sample

consisted of 96 adult inpatients and 10 adult outpatients with a diagnosis of schizophrenia,

schizoaffective disorder, psychotic disorder not otherwise specified (NOS), or bipolar disorder

with psychotic features between the ages of 19 and 66 (M = 41.1, SD = 11.17). The study

population was 84% male with 52% of the population being Caucasian, 45% African-American,

and 3% other. Participants were recruited from Vernon State Hospital (VSH) (36%), Terrell

State Hospital (TSH) (38%), and North Texas Veterans Health Care System (NTVHCS) (26%).

Descriptive Analyses

Descriptive statistics from the present study are provided, including mean and standard

deviations, for the manifest variables used for modeling (see Table 2). In addition, the

distributional properties of each measured variable were examined for normality for both

univariate and multivariate analysis. Univariate analysis indicated relatively mild skew (range

.00 to -1.85) and kurtosis (range -.20 to 2.36). Multivariate kurtosis was in the modest range for

Model 1 and very low for Model 2. Due to the presence of mild to modest skew and kurtosis,

robust parameters will be presented in the discussion of the modeling.

Insight was measured using the Schedule for the Assessment of Insight – Expanded

Version (SAI-E). The mean SAI-E total score for the sample was 12.25 (SD = 6.51) with 51%

demonstrating poor insight. The mean scores for depression (M = 3.0, SD = 1.61) and

grandiosity (M = 3.0, SD = 1.97) fell in the mild range, while the sample mean for unusual

26

thought content (UTC) (M = 3.8, SD = 1.77) was closer to the moderate range of severity. This

sample demonstrated a very low degree of guilt with the mean score (M = 2.3, SD = 1.51) falling

in the very mild range.

A correlation matrix was run to examine the relationships between measured variables

(see Table 3). Correlations were inspected for unusual or unexpected relationships between

demographic and measured variables. One significant inverse correlation was noted between age

and guilt (r = -.202, p = .05) with a largely nonsignificant trend of the psychopathological

symptoms of interest being negatively correlated with age. Data regarding level of educational

attainment and chlorpromazine equivalence was only available for the Terrell subset of the

sample. Correlations previously run with this group noted a fairly consistent pattern of

nonsignificant positive relationships (mean r = .20) between insight dimensions and education

and a significant correlation between chlorpromazine equivalence and the SAI-E treatment

composite. This latter relationship may be artificially inflated, to some extent, due to the

composites inclusion of nurses’ reports of medication compliance.

A review of the correlations between the variables proposed for inclusion in structural

equation modeling revealed that SAI-E item 6 (assessing patient’s opinion of the need for

treatment) did not correlate as expected with the other items proposed for inclusion in the

treatment domain (nurses’ observations of treatment compliance). While item 6 was

significantly correlated with all three compliance items (range .37 - .23), this range was

markedly lower than the significant correlations between item six and the three items assessing

awareness of mental illness (range .67 - .49). Due to the conceptual differences from the other

treatment items and the lower than expected correlations, item six was not included in the

structural equation modeling.

27

Analyses of the Combined Sample

Correlations, a multivariate analysis of variance (MANOVA), and the Box’s test of

equality of covariance matrices (Box’s M) were performed to determine if combining data from

three sites would violate the assumptions of the statistical procedures to be used. SAI-E items

included in the models were combined to form subscale scores and correlations were run using

these subscale scores and the symptoms items for each site. Visual inspection suggested no

substantial differences in the correlation patterns between sites.

A one-way MANOVA was calculated examining the effect of site (Vernon, Terrell, or

Dallas) on SAI-E total score, depression, guilt, grandiosity, and unusual thought content. A

significant effect was found (Lambda (10, 200) = 2.955, p < .002). Follow-up univariate

ANOVAs indicated significant differences between the sites for SAI-E total score (F (2, 107) =

4.407, p = .014) and unusual thought content (F (2, 105) = 8.336, p = .000). Tukey’s HSD was

used to determine the nature of the differences among the sites on these two variables. These

analyses revealed that participants at the NTVHCS had a significantly greater degree of insight

(M= 15.05, SD = 5.99) than participants at VSH (M = 10.52, SD = 5.71). Participants at TSH

(M = 11.62, SD = 7.19) were not significantly different from either of the other two groups.

Patients at TSH had significantly higher ratings of unusual thought content (M = 4.63, SD =

1.43) than the patients at VSH (M = 3.16, SD = 1.91) and patients at the NTVHCS (M = 3.55,

SD = 1.62) (the latter samples did not significantly differ from each other). The Box’s M was

not significant (Box’s M = 40.207, p = .168).

28

Structural Equation Modeling (SEM)

SEM was utilized with the data of 106 participants to test the models proposed in

hypotheses 1 and 2. The demographics of the 106 participants did not differ from the

demographics of the 110 total study population.

Model 1. The proposed three-factor model of insight (Model 1, see figure 1) discussed in

hypothesis one was examined first. SAI-E items 1, 2, and 3 were manifest variables (MVs) for

the construct of awareness. SAI-E items 4 and 8 were MVs for the construct of relabeling.

Finally, SAI-E items A, B, and C were MVs for the construct of treatment compliance. All of

the measured variables loaded significantly on the identified latent factor (z-values and robust z-

values > 1.96). A review of the average absolute standardized residuals (.0395) and the average

off-diagonal absolute standardized residuals (.0508) supports the present model by indicating

that the covariances observed are due to the underlying factors of the proposed model.

Moreover, using the difference between model chi squares (and degrees of freedom), the three-

factor insight model fit significantly better than a one-factor model of insight (X²(3)diff = 85.49,

p < .001). In an effort to be comprehensive, the three-factor model was also tested against a two-

factor model in which awareness of illness and the ability to relabel symptoms was collapsed

into the first factor and treatment compliance served as the second. The two-factor model did

not result in a significant improvement in fit (X²(2)diff = 3.1, p > .20) and produced a slight

decrease in good model fit and slight increase in standardized residual sizes.

With respect to the first model hypothesis, a three-factor insight model was initially run

with SAI-E item 7 included (rating awareness of prominent individual symptoms, up to a

maximum of four). A review of the average absolute standardized residuals (.0510) and the

average off-diagonal absolute standardized residuals (.0638) suggested some model mis-

29

specification. The twenty largest standardized residuals revealed that 25% involved item 7,

including large residuals with the two other items from the relabeling factor, and that 65%

involved one of the three nurse rated compliance items, including one large residual between

items A and B. It should be noted that there is overlap in these percentages. While all of the

factors loaded significantly, a review of the standardized solution demonstrates that items 7

(.375/.927), 8 (.586/.810), and B (.598/.801) had the lowest factor loadings and the highest error,

respectively. Based on theoretical considerations and concern for the generalizability of the

model, the decision was made to remove item 7 from the model.

The standardized parameter latent variable and error loadings for hypothesized Model 1

are presented in table 5. Review of this table reveals that all loadings are relatively high and

statistically significant. The confirmatory factor analysis latent variable intercorrelations are

presented in Table 8 and included in Figure 1. Overall, the intercorrelations are consistent with

previous findings.

Model 2. The second model hypothesis stating that insight domains could be modeled in

conjunction with two separate psychopathological symptom factors reflecting depression and

delusions was also examined. The appropriate Brief Psychiatric Rating Scale (BPRS) or Positive

and Negative Syndrome Scale (PANSS) depression and guilt scores were MVs for the

depression construct. The appropriate BPRS or PANSS unusual thought content (UTC) and

grandiosity scores were MVs for the delusions construct. The MVs and latent variables (LVs)

for insight used to test hypothesis one were utilized again for this model (see figure 2). All of the

measured variables loaded significantly on the identified latent factor (z-values and robust z-

values > 1.96). A review of the average absolute standardized residuals (.0472) and the average

30

off-diagonal absolute standardized residuals (.0558) supports the present model by indicating

that the covariances observed are due to the underlying factors of the proposed model.

The standardized parameter latent variable and error loadings for Model 2 are presented

in table 6. Review of this table reveals that all loadings are relatively high and statistically

significant. The loading for guilt (Model 2 = .58) may be lower due to low endorsement of this

symptom by the participants in this sample.

The confirmatory factor analysis (CFA) LV intercorrelations for model 2 are presented in

Table 9 and are also included in Figure 2. The most interesting insight-to-symptom factor

intercorrelations were as follows: delusions were most strongly related with the ability to relabel

symptoms (-.425, p < .05) while the relationship between delusions and treatment compliance

was not significant; depression was only modestly correlated with treatment compliance (.30, p <

.05) compared to robust relationships between depression and awareness of illness (.79, p < .05)

and relabeling of symptoms (.62, p < .05); and finally, the relationship between depression and

delusions was not significant.

Model 3: Additional Analyses. Due to the current findings that SAI-E item 6 (participant

awareness of the need for treatment) did not correlate as expected with the other SAI-E treatment

items, an additional SEM was used to determine if a need for treatment factor could be modeled

in conjunction with the three initially proposed insight factors and two separate

psychopathological factors. SAI-E item 6 and the mean score for Birchwood’s Insight Scale

treatment items were MV’s for the construct of need for treatment. The MVs and LVs for

insight and psychopathology used to test hypothesis two were utilized for this model as well. All

of the measured variables loaded significantly on the identified latent factor (z-values and robust

z-values > 1.96). A review of the average absolute standardized residuals (.0459) and the

31

average off-diagonal absolute standardized residuals (.0530) supports the present model by

indicating that the covariances observed are due to the underlying factors of the proposed model.

The standardized parameter latent variable and error loadings are presented in table 7.

Review of this table reveals that all loadings are relatively high and statistically significant.

The CFA LV intercorrelations for Model 3 are presented in Table 10. The most

interesting intercorrelations were as follows: the correlation between the patient’s awareness of

need for treatment and the delusion factor (-.12, p > .05) and treatment compliance and the

delusion factor (-.15, p > .05) are quite similar while patient’s awareness of need for treatment

and the depression factor (.50, p < .05) is nearly twice the size of the correlation between

treatment compliance and the depression factor (.29, p < .05); awareness of the need for

treatment was more strongly related to the other insight factors (.89 & .76, p < .05) than

treatment compliance (.78 & .62, p < .05); while, the moderate correlation between awareness of

the need for treatment and treatment compliance (.45, p < .05) represented the strongest

relationship between treatment compliance and any of the other factors.

Goodness of Fit

The overall fit of the models was assessed using the Satorra-Bentler chi-square,

comparative fit index (CFI), robust comparative fit index (RCFI), standardized root mean square

residual (SRMR), and the root mean square error of application (RMSEA). In general, the fit

indices indicate good model fit for all three models with all but one of the fit indices within

desirable ranges. See Table 4 for model fit results. Good model fit indicates that the specified

relations between the MVs and LVs (see figures 2 & 3) are adequate representations of the

observed data. The Satorra-Bentler chi-square was significant for Model 2 (x² (44) = 74.59, p =

.003) and Model 3 (x² (62) = 84.99, p = .028). As mentioned previously, there are problems

32

associated with the use of chi-square statistics as a goodness of fit measure. Though the Satorra-

Bentler is more robust than the traditional chi-square, it is still hazard to the same issues as the

chi-square. It is likely in this instance that the sample size (n = 106) has resulted in a minor

discrepancy between the observed and proposed covariance matrixes achieving statistical, but

not clinical, significance. In support of this, it is important to note that the average and average

off-diagonal absolute standardized residuals, CFI, RCFI, SRMR, and RMSEA are all within

good ranges.

33

CHAPTER 4

DISCUSSION

Poor awareness of illness, often referred to as lack of insight, is a distinctive feature of

severe psychotic disorders, which has recently gained more attention in the literature. Research

demonstrating associations between insight and less time spent in the hospital (Van Os et al.,

1996), better treatment compliance (Bartko, Herczeg, & Zador, 1988; Sanz et al., 1998; Van

Putten, 1974), more contact with mental health services (Sanz et al., 1998), and better vocational

rehabilitation (Lysaker and Bell, 1995) helps to underscore the importance of this construct and

the need to understand it better. Insight has been defined as a multi-dimensional phenomenon

that is a separable component from pathological symptoms and neurocognitive deficits of

psychosis. So far, the use of exploratory factor analysis to determine if insight is multi-

dimensional has been mixed. The current research examined whether there was evidence for a

three factor model of the insight construct using a confirmatory factor analysis (CFA) which can

provide stringent statistical tests of hypothetical factor models while accounting for measurement

error. The three-factor model of insight was also examined along with two separate

psychopathological symptom factors (depression, delusions), which have been shown to be

significantly associated with insight. Principal findings from the current study were that the

three-factor model of insight was supported and that the insight factors were meaningfully

correlated to the two symptom factors. Moreover, the three-factor insight model provided

significantly better fit than a single factor model of insight. Thus, the present results support and

extend the findings for a three-factor model of insight, consistent with the results reported by

Kim et al. (1997).

34

The discussion that follows is organized into six sections: (a) analyses of the combined

sample; (b) Model 1 results; (c) Model 2 results; (d) clinical implications of the findings; (e)

limitations of the current study; and (f) suggestions for future research.

Analyses of the Combined Sample

The sample for the current study was created by combining the data from samples

collected at three separate sites. While doing so increased the total N for the structural equation

modeling (SEM) analyses and variability on the manifest variables, it created the potential

confound of combining significantly different covariance matrices. The significant findings of

the multivariate analysis of variance (MANOVA) and subsequent related analyses demonstrated

that the samples did differ significantly from each other in terms of mean level of total insight

and unusual thought content (UTC). Veterans from the North Texas Veterans Health Care

System (NTVHCS) were found to have significantly more insight that the patients at Vernon

State Hospital (VSH), a forensic facility. This difference is not surprising given the findings of

Arango, Calcedo-Barba, Gonzalez-Salvador, and Calcedo-Ordonez (1999) demonstrating that

insight into symptoms helped accurately predict future violence. An additional finding was that

the sample from Terrell State Hospital (TSH) had a significantly higher UTC mean that either of

the other samples. This finding makes sense due to a large portion of this sample having been

civilly committed compared to none of the NTVHCS sample and VSH sample all having been

hospitalized (and treated) for a sufficient period of time in order to earn enough privileges to be

placed on the least restrictive unit in the adult section of the facility.

Variability in the covariance matrices was explored by visual examination of the pattern

of correlations for the three subsamples and the Box’s test of equality of covariance matrices

(Box’s M). All sites showed similar trends in their pattern of correlations and the Box’s M

35

statistic was not significant. In sum, these results suggest that combining the data across these

sites provided the variance on traits helpful in factor analysis, and did not involve a combination

of substantially different covariance matrices that could have hindered the modeling process.

Moreover, if the subsamples had differed substantially in terms of variable relations, then the

model results would not have resulted in such good fit.

Model 1

The results indicate good model fit to the data in terms of large manifest variable

loadings onto their respective latent variables along with small residual error. Thus, it can be

concluded that the specified manifest variables are excellent indicators for their respective latent

constructs. Specifically, Schedule for the Assessment of Insight – Expanded Version (SAI-E)

items 1, 2, and 3 represent awareness of illness; and SAI-E items 4 and 8 represent ability to

relabel symptoms as signs of psychopathology. The SAI-E items A, B, and C represent nurses’

ratings of patients’ treatment compliance, a key indicator of patients’ awareness of need for

treatment (see figure 1).

Two modifications were made from the originally proposed model to achieve the

goodness of fit reported for the current model. Based upon the results of the initial correlation

matrix, item 6, the SAI-Es best measure of patient’s awareness of the need for treatment,

demonstrated higher correlations with the items assessing awareness of illness than with the

items assessing treatment compliance (though item 6 did correlate significantly with the other

treatment items). It has been suggested that awareness of the need for treatment and treatment

compliance are not equivalent concepts, particularly in settings where important privileges, such

as greater freedom within the hospital or release to the community, are dependent upon treatment

36

compliance. Thus, the decision was made, a priori, to exclude item 6 from the modeling

procedures.

The deletion of item 7 was part of the model modification process. When Model 1 was

initially run nearly all of the goodness of fit indices were approaching but fell just below what is

traditionally considered good fit. When the 20 largest standardized residuals were reviewed,

problems were identified with items 7, A, and B. From a theoretical standpoint, deleting item B

(which had a lower factor loading and higher error variance than item A), while potentially

justifiable with this sample, would have risked maximizing fit to a sample while reducing

generalizability to the population.

Meanwhile there were some theoretical problems with item 7 that would likely apply to

the population. In particular, item 7 is potentially a poor indicator of the capacity to relabel

symptoms due to the potential for averaging across diverse symptoms, such as hallucinations and

delusions, which previous research has suggested are differentially related to insight (David et

al., 1992; Nayani & David, 1996; Neumann et al., 1996). Additionally, different symptoms were

evaluated for item 7 with different patients resulting in a lack of consistency regarding what was

measured. These problems also exist with item 8 to some extent, however, but item 7 requires a

higher level of abstractive reasoning and capacity for personal responsibility for symptoms

which increased the potential for confounding factors such as intelligence and locus of control

that were beyond the scope of the current study. Thus, the present modifications of model 1

were based on theoretical grounds and resulted in better model fit and parsimony.

After the deletion of item 7, item 8 showed the lowest loading of the examined variables.

This result is likely due to the initial problems discussed in relation to item 7 (i.e., inconsistency

in what was measured across participants and less meaningful scores due to averaging across

37

symptoms with a differential relationship to insight). However, the loading was adequate, and it

can be concluded that all manifest variables represent their specified constructs.

Model 2

Overall, the results for Model 2 also indicate good model fit, as well as, high latent

variable loadings, especially for the insight items. A problem was noted with the Satorra-Bentler

chi-square for this model, which indicated that there was a significant difference between the

proposed model and what was observed in the sample. While the Satorra-Bentler was created to

provide a more robust alternative to the standard chi square goodness-of-fit indices, it is still

susceptible to the same problems. Specifically, when sample sizes become large enough,

statistically but not theoretically significant differences between the proposed model and

observed data can result. In support of this explanation for the present results, a review of the

other fit indices suggests a good fit of the model with only minimal differences between the

observed covariance matrix and the model-generated covariance matrix (e.g., small standardized

residuals). Thus, it can be concluded that the specified manifest symptoms variables were also

representative of their specific latent constructs. Model 2 differs from Model 1 in that the three-

factor model of insight was again supported, but this time in the presence of symptoms of

psychopathology that have been previously demonstrated to be associated with insight or the

lack thereof. These findings are consistent with the literature suggesting that insight manifest

variables (MVs) load onto their own latent factor, and are not simply alternative manifestations

of latent variable (LV) symptom factors (David, 1998; Cuesta & Peralta, 1994; Cuesta et al.,

2000; Peralta & Cuesta 1994). The current results also extend the past findings through use of

confirmatory procedures demonstrating that three dimensions of insight can be modeled

separately from psychopathology.

38

The CFA LV intercorrelations for Model 2 produced some interesting results that may

help to further our understanding of the relationships between insight and symptoms of

psychopathology. The ability to relabel symptoms as pathological was more strongly (inversely)

correlated to the delusion factor compared to the correlation between the awareness and delusion

factors. In addition, the relationship between the delusion and treatment compliance factors was

not significant. This is somewhat consistent with findings on a related self-report measure of

insight (Birchwood’s Insight Scale) that relabeling of symptoms was the only insight domain

significantly correlated with delusions (-.45, p < .05) (Gonterman 2002). It is also interesting

that the relationship between delusions and treatment compliance was not significant since

delusional explanations for discontinuing medication are not uncommon in clinical settings.

Perhaps this finding is more a reflection of the subset of patients with psychosis who were

sampled (i.e., hospitalized patients), unfortunately the number of outpatients in this sample was

too small to provide a clear indication regarding whether this relationship would change with

different conditions.

With respect to the associations between the depression and insight factors, the strongest

correlation occurred between the awareness of illness and depression factors. The association

between the depression factor and the relabeling factor was still quite robust, though the

correlation was only modest between the depression factor and treatment compliance. These

findings are fairly consistent with the previous literature (Amador et al., 1994; Carroll et al.,

1999; David et al., 1995; Dickerson et al., 1997; Gonterman, 2002; Kemp et al., 1995;

Michalakeas et al., 1994; Moore et al., 1999; Neumann et al., 1996; Peralta et al., 1994; Sanz et

al., 1998; Smith et al., 1998; Smith et al., 2000). Taken together, the current findings highlight

39

the negative impact that awareness of having a mental disorder can have on individuals’ well

being.

Another important finding from the Model 2 results involves the non-significant

relationship between depression and delusions in psychotic individuals. This finding is

inconsistent with the protective hypothesis that delusions develop as a way of warding off the

depression associated with recognizing one has a serious illness that is highly stigmatized by

society. Furthermore, the finding that delusions are more modestly intercorrelated with

awareness suggests that delusions likely arise for other reasons. Perhaps a more plausible

explanation is that delusions arise as a way to make sense of the disjointed and confusing pieces

of information these persons perceive due to the neurocognitive deficits associated with

psychotic (particularly schizophrenia spectrum) illnesses.

Clinical Implications

The findings of the current research have implications for clinical treatment; a summary

of which follows. First, awareness of illness and the ability to relabel factors were only modestly

to moderately related to the observable treatment compliance factor. Thus, programs designed to

increase treatment compliance should include components beyond increasing the awareness of

one’s illness and include components focusing on concrete positive reinforcements for ongoing

compliance with psychotropic medication.

This type of a component would seem especially important when working with

hospitalized patients since the data from this sample suggested that patients’ awareness of the

need for treatment was more moderately correlated with treatment compliance than initially

predicted. Supplemental SEM analyses demonstrated that awareness of the need for treatment

could be modeled as a factor of insight in conjunction with a treatment compliance factor with

40

these two independent variables being moderately correlated. Nurses’ observations of treatment

compliance were more strongly correlated with patients’ awareness of the need for treatment

compared to the correlations between treatment compliance and the awareness and relabeling

factors. While it is remarkable that two completely separate sources of measurement parallel to

this degree, the current findings that the need for treatment factor is much more strongly related

to the other insight factors and depression while being similarly associated with delusions,

further supports the SEM findings that, at least in populations where treatment compliance

results in secondary gain, awareness of the need for treatment and treatment compliance are

separate, but related factors. In addition, the possibility that the treatment compliance factor may

reflect more about the relationship between nurses and specific patients suggests that programs

designed to increase treatment compliance might benefit from placing a strong emphasis on

developing therapeutic relationships with their clients.

Second, clinicians involved in programs designed to increase patients’ insight should be

aware of the risk of increased depression in patients’ whose level of insight improves. It would

appear beneficial for such programs to closely monitor the degree of participant’s depression,

particularly since research has suggested that increased insight into illness, fewer years of

treatment, and more severe depressive symptoms each significantly increase patients’ risk of

suicidality (Schwartz & Smith, 2004). Ongoing monitoring might allow for the implementation

of steps to reduce resulting impairment in daily functioning or suicide risk while not introducing

interventions that might reduce depression to the point that it could interfere with the mission of

the program.

Third, the current findings do not support the protective hypothesis that delusions

develop to protect against depression and anxiety. While increased awareness of illness is

41

associated with increased depression, at least cross-sectionally, it does not appear to follow that

delusions will increase as a way to protect one self. To the extent that delusions arise from

neurocognitive deficits in perception, memory, and attention, neurocognitive rehabilitations

programs specially designed for psychiatric patients could result in a decrease in delusions over

time.

Limitations

The generalizability of these findings is hindered by the following limitations. First, the

majority of this sample consisted of hospitalized patients and 64% of the sample had experienced

a significant exacerbation in psychotic symptoms within two weeks of testing. Thus, these

results are most reflective of acute hospitalized patients. Second, the same rater conducted the

symptom and insight interviews, and this may have inflated the relationships between the

symptom ratings and the SAI-E. However, it is notable that results available for the TSH subset

demonstrated significant associations between a self-report measure of insight and symptoms.

Third, the study population consisted of a higher percentage of minorities than found in the

general population.

Future Directions

The current study suggests several new directions for research on insight. Research

should include additional samples beyond acute populations and hospitalized patients. Included

in this recommendation are inpatients ready for discharge to the community, patients with

exacerbations managed on an outpatient basis, and stabilized outpatients.

Furthermore, longitudinal studies exploring changes in symptom and insight levels over

time would be useful in explaining the role of insight in psychotic disorders. It would prove

particularly valuable for future research to include data collected during multiple acute, recovery,

42

and stabilized phases. Such research would help to determine if the current model of insight

could be supported across different phases of the illness. Additionally, it may further elucidate

whether insight fluctuations within the individual are the sole product of illness status or

additional factors since the few longitudinal studies published so far have suggested that the

relationship between insight and symptoms changes between acute and stable phases (Cuesta et

al., 2000; Mintz, Addington, & Addington, 2004).

The relationship between awareness of the need for treatment and observable treatment

compliance deserves closer consideration. The current study found that awareness of the need

for treatment and observable treatment compliance were only modestly to moderately correlated.

A more complete exploration of these factors and whether the relationship changes in different

settings is warranted.

43

Table 1

Studies Showing Significant Associations across Insight Dimensions with Symptoms

Study Dep Del Dis Pos Neg Sym Scale Dimension____

Amador et al. 1993 I - - I NS - SUMD Aware (P, C) Amador et al. 1994 NS NS I I I NS SUMD Aware Illness NS I I I I NS Consequences, Treatment P I I I I NS Attributions

David et al. 1992 - - - - - I SAI Relabel, Treatment

David et al. 1995 P - - - - NS SAI Awareness, Relabel, Treatment

Debowska et al. 1998 - I NS I I I PANSS, Aware Illness, SALI Treatment - I I I I I Relapse Aware

Gonterman 2002 P I I - NS - SAI-E Awareness, Relabel, Treatment NS I NS - NS - IS Relabel NS NS NS - P - Treatment

Kemp & Lambert 1995 P I - I NS I SUMD Aware (C) P I - I I I Aware (P), Attribute (P) P I - NS NS I Attribute (C)

Kim et al. 1997 NS I I I NS - SAI-E Awareness, Relabel

Peralta & Cuesta 1994 P - - - - - AMDP Awareness

Moore et al. 1999 P - - - - - SUMD Awareness (C), Conseq. (C & P)

(table continues)

44

Table 1 (continued)

Studies Showing Significant Associations across Insight Dimensions with Symptoms

Study Dep Del Dis Pos Neg Sym Scale Dimension____

Schwartz 1998 - - - I NS NS SUMD Awareness, Consequences

Smith et al. 1998 P - I NS NS - SUMD Aware (C), Attribute (C) Smith et al. 2000 P - - NS NS - SUMD Aware (C), Attribute (C) NS - - P NS - Attribute (P)

Young et al. 1998 - - - - - I SUMD Awareness

Note. To simplify interpretation, all scales are reported so that a positive association (P) means increased (better) insight was associated with increased correlate, and a negative association (I) means increased (better) insight was associated with decreased correlate. (NS = not significant). Dep = depression, Del = delusions, Dis = disorganized, Pos = positive symptoms, Neg = negative symptoms, Sym = total symptoms, C = current, P = past.

45

Table 2

Descriptive Statistics for the Manifest Variables for Models 1 and 2

Variable___________________M__________SD_________N_____________________

SAI-E 1 .99 .90 109

SAI-E 2 .94 .89 109

SAI-E 3 1.08 .90 109

SAI-E 4 .75 .80 109

SAI-E 8 .87 1.13 109

SAI-E A 1.72 .56 108

SAI-E B .90 .88 108

SAI-E C 5.2 1.39 108

Depression 2.9 1.61 107

Guilt 2.3 1.51 109

Grandiosity 2.7 1.97 109

Unusual Thought Content 3.8 1.77 108

46

Table 3

Spearman’s Rho Correlations between Model 1 & 2 Proposed Manifest Variables

S-1 S-2 S-3 S-4 S-7 S-8 S-6 S-A S-B S-C Dep Gui UTC Gran

S-1 ---

S-2 .56** ---

S-3 .56** .73** ---

S-4 .49** .62** .61** ---

S-7 .19* .31** .16 .19* ---

S-8 .32** .36** .31** .38** .53** ---

S-6 .49** .56** .67** .49** .23** .34** ---

S-A .31** .32** .29** .26** .18* .21* .23* ---

S-B .46** .35** .36** .24** -.03 -.03 .32** .35** ---

S-C .41** .33** .27** .25** .09 .11 .37** .57** .68** ---

Dep .40** .45** .36** .30** .13 .25** .22* .20* .16 .18* ---

Gui .46** .27** .38** .32** .12 .18* .31** .21* .20* .17 .33** ---

UTC -.11 -.29** -.12 -.20* -.44** -.35** -.03 .01 .12 -.04 -.12 .01 ---

Gran -.05 -.23** -.17* -.20* -.32** -.19* -.17* -.16 -.10 -.15 -.21* .01 .43** ---

Note.

*denotes p < .05 **denotes p < .01

47

Table 4

Model Fit Results

Model df S-B X² CFI RCFI SRMR RMSEA_____

1 17 21.2 .970 .989 .074 .079 p=.22

2 44 74.6 .929 .935 .072 .083 p=.00 3 62 84.9 .943 .963 .069 .072 p=.03 Note. S-B X² = Satorra-Bentler chi-square; CFI = comparative fit index; RCFI = robust comparative fit index; SRMR = standardized root mean squared residual; RMSEA = root mean square error of application Table 5 Standardized Parameter Latent Variable and Error Loadings: Model 1 Latent Variables Manifest Variables Awareness Relabeling TX Compliance Error SAI-E 1 .69 .73

SAI-E 2 .86 .51

SAI-E 3 .84 .54

SAI-E 4 .84 .55

SAI-E 8 .53 .85

SAI-E A .74 .67

SAI-E B .60 .80

SAI-E C 1.00 .00

Note. TX = treatment

48

Table 6 Standardized Parameter Latent Variable and Error Loadings: Model 2 Latent Variables Manifest Variables 1 2 3 4 5 Error 1. SAI-E 1 .70 .72

1. SAI-E 2 .86 .52

1. SAI-E 3 .83 .57

2. SAI-E 4 .80 .60

2. SAI-E 8 .55 .83

3. SAI-E A .72 .69

3. SAI-E B .61 .79

3. SAI-E C .99 .11

4. DEP. .63 .78

4. GUILT .58 .82

5. UTC .66 .76

5. GRAND. .69 .73

Note: 1 = Awareness; 2 = Relabeling; 3 = Treatment Compliance; 4 = Depression; 5 = Delusions; Dep. = Depression; UTC = Unusual Thought Content; Grand. = Grandiosity

49

Table 7 Standardized Parameter Latent Variable and Error Loadings: Model 3 Latent Variables Manifest Variables 1 2 3 4 5 6 Error 1. SAI-E 1 .69 .72

1. SAI-E 2 .83 .56

1. SAI-E 3 .85 .52

2. SAI-E 4 .78 .63

2. SAI-E 8 .57 .83

3. SAI-E 6 .81 .58

3. IS TX .76 .65

4. SAI-E A .72 .70

4. SAI-E B .61 .79

4. SAI-E C 1.0 .00

5. DEP. .63 .77

5. GUILT .57 .82

6. UTC .75 .66

6. GRAND. .60 .80

Note: 1 = Awareness; 2 = Relabeling; 3 = Need for Treatment; 4 = Treatment Compliance; 5 = Depression; 6 = Delusions; TX = Treatment; Dep. = Depression; UTC = Unusual Thought Content; Grand. = Grandiosity

50

Table 8

Model 1 Confirmatory Factor Analysis (CFA) Latent Variable Intercorrelations

Total Sample

Latent Variables 1 2 3

1. Awareness -- .86* .43*

2. Relabeling -- .33*

3. TX Compliance --

Table 9

Model 2 Confirmatory Factor Analysis (CFA) Latent Variable Intercorrelations

Total Sample

Latent Variables 1 2 3 4 5

1. Awareness -- .87* .41* .79* -.32*

2. Relabeling -- .32* .62* -.43*

3. TX Compliance -- .30* -.19

4. Depression -- -.17

5. Delusions --

*p<.05

51

Table 10

Model 3 Confirmatory Factor Analysis (CFA) Latent Variable Intercorrelation

Total Sample Latent Variables 1 2 3 4 5 6

1. Awareness -- .87* .40* .78* -.29* .89*

2. Relabeling -- .32* .62* -.43* .76*

3. TX Compliance -- .29* -.15 .45*

4. Depression -- -.16 .50*

5. Delusions -- -.12

6. Need for TX -- *p<.05

52

Legend E = Measurement Error; Aware = Awareness of Illness; Relabel = Relabeling of Psychotic Symptoms; Treatment = Treatment Compliance Figure 1. Model 1: Three Factor Model of Insight using SAI-E Items as Manifest Variables

AWARE

SAI -E 1

SAI-E4

SAI -E 8

-

SAI -E A

SAI -E B

SAI -E C

SAI -E 2

SAI -E 3

E1

E2

E3

E5

E6

E7

E8

RELABEL

TREATMENT

E4

.73

.51

.54

.55

.85

.67

.80

.00

.69

.86

.84

.84

.53

.74

.60

1.0

.33

.86

.43

53

Figure 2. Model 2: Three Factor Model of Insight and Psychopathology using SAI-E, BPRS,

and PANSS Items as Manifest Variables

AWARE

SAI-E 1

SAI-E A

SAI-E 8

SAI-E 4

SAI-E 3

SAI-E C

E

DEL

DEP

RELAB

TREAT

.79

.62

.30

-.32

-.43

-.19

.86

.43

.70

.86

.83

.80

.55

.72

.61

SAI-E B

.99

.63

.58

.66

.69

.72

.52SAI-E 2

.57

.80

.83

.69

.79

.11

.73 GRAND E

.76 UTC

E

E

E

E

E

E

E

.82 GUILT

.78 DEP

E

E

E

Legend E = Error; Aware = Awareness; Relab = Relabeling; Treat = Treatment Compliance; Del = Delusions; Dep = Depression; UTC = Unusual Thought Content; Grand = Grandiosity

54

REFERENCES

Addington, J., & Addington, D. (1998). Facial affect recognition and information

processing in schizophrenia and bipolar disorder. Schizophrenia Research, 32,

171-181.

Almeida, O., Howard, R., Levy, R., & David, A. (1996). Insight and paranoid disorders

of late life (late paraphrenia). International Journal of Geriatric Psychiatry, 11, 653-658.

Amador, X. F. & David, A. S. (1998). Insight and psychosis. New York: Oxford

University Press.

Amador, X. F., & Kronengold, H. (1998). The description and meaning of insight. In X.

F. Amador & A. S. David (Eds.), Insight and psychosis (pp. 15-32). New York:

Oxford University Press.

Amador, X.F., Strauss, D.H., Yale, S.A., & Gorman, J.M. (1991). Awareness of illness in

schizophrenia. Schizophrenia Bulletin, 17, 113-132.

Amador, X. F., Strauss, D. H., Yale, S., Gorman, J. M., & Endicott, J. (1993). The

assessment of insight in psychosis. American Journal of Psychiatry, 150, 873-879.

Amador, X. F., Flaum, M., Andreasen, N. C., Strauss, D. H., Yale, S. A., Clark, S. C., &

Gorman, J. M. (1994). Awareness of illness in schizophrenia and schizoaffective

and mood disorders. Archives of General Psychiatry, 51, 826–836.

Andersen, J., Korner, A., Larsen, J. K., Schultz, V. Nielsen, B. M., Behnke, K, Munk-

Andersen, E., & Bjorum, N. (1993). Agreement in psychiatric assessment. Acta

Psychiatrica Scandinavica, 87, 128-132.

Andreason, N. C. (1984a). Scale for the Assessment of Negative Symptoms (SANS).

Iowa City: University of Iowa College of Medicine.

55

Andreason, N. C. (1984b). Scale for the Assessment of Positive Symptoms (SAPS). Iowa

City: University of Iowa College of Medicine.

Arango, C., Calcedo-Barba, A., Gonzalez-Salvador, T., & Calcedo-Ordonez, A. (1999).

Violence in inpatients with schizophrenia: A prospective study. Schizophrenia

Bulletin, 25, 493-503.

Asarnow, R. F., & MacCrimmon, D. J. (1981). Span of apprehension deficits during

postpsychotic stages of schizophrenia. Archives of General Psychiatry, 38, 1006-

1011.

Bartko, G., Herczeg, I., & Zador, G. (1988). Clinical symptomatology and drug

compliance in schizophrenic patients. Acta Psychiatrica Scandinavica, 77, 74-76.

Basso, M. R., Nasrallah, H. A., Olson, S. C., & Bornstein, R. A. (1998).

Neuropsychological correlates of negative, disorganized and psychotic symptoms

in schizophrenia. Schizophrenia Research, 31, 99-111.

Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory

for measuring depression. Archives of General Psychiatry, 4, 561-571.

Bell, M., Milstein, R., Beam-Goulet, J., Lysaker, P., & Cicchetti, D. (1992). The Positive

and Negative Syndrome Scale and the Brief Psychiatric Rating Scale: Reliability,

comparability, and predictive validity. Journal of Nervous and Mental Disease,

180, 723-728.

Berman, I., Viegner, B., Merson, A., Allan, E., Pappas, D., & Green, A. I. (1997).

Differential relationships between positive and negative symptoms and

neuropsychological deficits in schizophrenia. Schizophrenia Research, 25 (1), 1–

10.

56

Berrios, G. E., & Markova, I. S. (1998). Insight in the psychoses: A conceptual history. In

X. F. Amador & A. S. David (Eds.), Insight and psychosis (pp. 33-46). New

York: Oxford University Press.

Birchwood, M., Smith, J., Drury, V., Healy, J., Macmillan, F., & Slade, M. (1994). A

self-report Insight Scale for psychosis: reliability, validity and sensitivity to

change. Acta Psychiatrica Scandinavica, 89, 62-67.

Braff, D. (1993). Information processing and attention dysfunctions in schizophrenia.

Schizophrenia Bulletin, 19, 233-259.

Brenner, H., Roder, V., Hodel, B., Kienzle, N. Reed, D., & Liberman, R. (1994).

Integrated Psychological Therapy for Schizophrenic Patients (IPT). Seattle, WA:

Hogrefe and Huber.

Carroll, A., Fattah, S., Clyde, Z., Coffey, I., Owens, D., & Johnstone, E. C. (1999).

Correlates of insight and insight change in schizophrenia. Schizophrenia

Research, 35, 247-253.

Collins, A. A., Remington, G. J., Coulter, K., & Birkett, K. (1997). Insight,

neurocognitive function and symptom clusters in chronic schizophrenia.

Schizophrenia Research, 27, 37-44.

Cornblatt, B. A., & Erlenmeyer-Kimling, L. (1985). Global attentional deviance as a

marker of risk for schizophrenia: Specificity and predictive validity. Journal of

Abnormal Psychology, 94, 470-486.

Cornblatt, B. A., & Keilp, J. G. (1994). Impaired attention, genetics, and the

pathophysiology of schizophrenia. Schizophrenia Bulletin, 20, 31-46.

57

Cornblatt, B. A., Risch, N. J., Faris, G., Friedman, D., & Erlenmeyer-Kimling, L. (1988).

The Continuous Performance Test, Identical Pairs Version (CPT-IP): I. New

findings about sustained attention in normal families. Psychiatry Research, 26,

223-238.

Cornblatt, B. A., Obuchowski, M., Schnur, D. B., & O’Brien, J. D. (1997). Attention and

clinical symptoms in schizophrenia. Psychiatry Quarterly, 68, 343-359.

Cornblatt, B. A., Winters, L., & Erlenmeyer-Kimling, L. (1989). Attentional markers of

schizophrenia: Evidence from the New York High Risk Study. In S. C. Schulz, &

C. A. Tamminga (Eds.), Schizophrenia: Scientific Progress (pp. 83-92). New

York: Oxford University Press.

Cuesta, M. J., & Peralta, V. (1994). Lack of insight in schizophrenia. Schizophrenia

Bulletin, 20, 359-366.

Cutting, J., & Murphy, D. (1988). Schizophrenic thought disorder. British Journal of

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. British Journal

of Psychiatry, 157, 355-358.

Czobor, P., Bitter, I., & Volavka, J. (1991). Relationship between the Brief Psychiatric

Rating Scale and the Scale for the Assessment of Negative Symptoms: A study of

their correlation and redundancy. Psychiatry Research, 36, 129-139.

David, A. S. (1990). Insight and psychosis. British Journal of Psychiatry, 156, 798-808.

David, A. S. (1998). The clinical importance of insight. In X. F. Amador & A. S. David

(Eds.), Insight and psychosis (pp.332-351). New York: Oxford University Press.

58

David, A. S., Buchanon, A., Reeds, A., & Almeida, (1992). The assessment of insight.

British Journal of Psychiatry, 161, 599-602.

David, A. S., Van Os, J., Jones, R., Harvey, I., Forster, A., & Fahy, T. (1995). Insight and

psychotic illness: Cross-sectional and longitudinal associations. British Journal of

Psychiatry, 167, 621-628.

Derogatis, L. R., Lipman, R. S., Rickels, K., Uhlenhuth, E. H., & Covi, L. (1974). The

Hopkins Symptom Checklist (HSCL): A self-report symptom inventory.

Behavioral Science, 19, 1-15.

Deutsch, G., Papanicolaou, A. C., Bourbon, W. T., & Eisenberg, H. M. (1987). Cerebral

blood flow evidence of right frontal activation in attention demanding tasks.

International Journal of Neuroscience, 36, 23-28.

Dickerson, F. B., Boronow, J. J., Ringel, N., & Parente, F. (1997). Lack of insight among

outpatients with schizophrenia. Psychiatric Services, 48, 195-199.

Earle-Boyer, E. A., Serper, M. R., Davidson, M., & Harvey, P. D. (1991). Continuous

performance tests in schizophrenic patients: Stimulus and medication effects on

performance. Psychiatry Research, 37, 47-56.

Erlenmeyer-Kimling, L., & Cornblatt, B. A. (1987). High-risk research in schizophrenia:

A summary of what has been learned. Journal of Psychiatric Research, 21, 401-

411.

Erlenmeyer-Kimling, L., & Cornblatt, B. A. (1992). A summary of attentional findings in

the New York High-Risk Project. Journal of Psychiatric Research, 26, 405-426.

59

Flemming, K., & Green, M. F. (1995). Backward masking performance during and after

manic episodes. Journal of Abnormal Psychology, 104, 63-68.

Freedman, L. R., Rock, D., Roberts, S. A., Cornblatt, B. A., & Erlenmeyer-Kimling, L.

(1998). The New York High-Risk Project: Attention, anhedonia, and social

outcome. Schizophrenia Research, 30, 1-9.

Garety, P. A. (1998). Insight and delusions. In X. F. Amador & A. S. David (Eds.),

Insight and psychosis (pp. 66-77). New York: Oxford University Press.

Ghaemi, S. N. (1997). Insight and psychiatric disorders: A review of the literature, with a

focus on its clinical relevance for bipolar disorder. Psychiatric Annals, 27, 782-

790.

Goldman, R. S., Tandon, R., Liberzon, I., Goodson, J., & Greden, J. F. (1991). Stability

of positive and negative symptom constructs during neuroleptic treatment in

schizophrenia. Psychopathology, 24, 247-252.

Green, M. (1996). What are the functional consequences of neurocognitive deficits in

schizophrenia? American Journal of Psychiatry, 153, 321-330.

Green, M. F., Nuechterlein, K. H., & Mintz, J. (1994). Backward masking in

schizophrenia and mania. II. Specifying the visual channels. Archives of General

Psychiatry, 51, 945-951.

Green, M. F, & Walker, E. F. (1986). Symptom correlates of vulnerability to backward

masking in schizophrenia. American Journal of Psychiatry, 143, 181-186.

Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery,

and Psychiatry, 23, 56-62.

60

Harvey, P. D., Keefe, R. S. E., Moskowitz, J., Putnam, K. M., Mohs, R. C., & Davis, K.

L. (1990). Attentional markers of vulnerability to schizophrenia: Performance of

medicated and unmedicated patients and normals. Psychiatry Research, 33, 179-

188.

Heaton, P. K., Baade, L. E., & Johnson, K. L. (1978). Neuropsychological test results

associated with psychiatric disorders in adults. Psychology Bulletin, 85, 141-162.

Heinrichs, D. W., Cohen, B. P., & Carpenter, W. T. (1986). Early insight and the

management of schizophrenic decompensation. Journal of Nervous and Mental

Disease, 173, 133-138.

Ito, M., Kanno, M., Mori, Y., Niwa, S. (1997). Attention deficits assessed by Continuous

Performance Test and Span of Apprehension Test in Japanese schizophrenic

patients. Schizophrenia Research, 23, 205-211.

Judd-Finkelstein, J. R., Cannon, T. D., Gur, R. E., Gur, R. C., & Moberg, P. (1997).

Attentional dysfunctions in neuroleptic-naïve and neuroleptic-withdrawn

schizophrenic patients and their siblings. Journal of Abnormal Psychology, 106,

203-212.

Kay, S. R., Fiszbein, A., & Opler, L. A. (1987). The Positive and Negative Syndrome

Scale (PANSS) for schizophrenia. Schizophrenia Bulletin, 13, 261-276.

Keefe, R. S. (1998). The neurobiology of disturbances of the self: Autonoetic agnosia in

schizophrenia. In E. F. Amador & A. S. David (Eds.), Insight and psychosis (pp.

142-173). New York: Oxford University Press.

Kemp, R., & David, A. (1996). Psychological predictors of insight and compliance in

psychotic patients. British Journal of Psychiatry, 169, 444-450.

61

Kemp, R., & David, A. S. (1997). Insight and compliance. In B. Blackwell (Ed.),

Treatment compliance and the therapeutic alliance (pp. 61-84). Newark, NJ:

Gordon and Breach Publishing Group.

Kemp, R., David, A. S., & Hayward, P. (1996). Compliance therapy: An intervention

targeting insight and treatment adherence in psychotic patients. Behavioural and

Cognitive Psychotherapy, 24, 331-350.

Kemp, R., & Lambert, T. (1995). Insight in schizophrenia and its relationship to

psychopathology. Schizophrenia Research, 18, 21-28.

Kim, Y., Sakamoto, K., Kamo, T., Sakamura, Y., & Miyaoka, H. (1997). Insight and

clinical correlates in schizophrenia. Comprehensive Psychiatry, 38, 117–123.

Kinsbourne, M. (1998). Representations in consciousness and the neuropsychology of

insight. In X. F. Amador & A. S. David (Eds.), Insight and psychosis (pp. 174-

190). New York: Oxford University Press.

Kraepelin, E. (1902). Clinical psychiatry: A textbook for physicians (A. Diffendorf,

Trans.). New York: Macmillan.

Kraimer, H. C. & Thiemann, S. (1987). How many subjects?: Statistical power analysis

in research. Newbury Park, CA: Sage.

Lewis, A. (1934). The psychopathology of insight. British Journal of Medicine and

Psychology, 14, 332-348.

Lezak, M. D. (1995). Neuropsychological assessment (3rd ed.). New York: Oxford

University Press.

Lieh-Mak, F., & Lee, P. W. H. (1997). Cognitive deficit measures in schizophrenia:

Factor structure and clinical correlates. American Journal of Psychiatry, 154, 39-46.

62

Lysaker, P. H., & Bell, M. D. (1994). Insight and cognitive impairment in schizophrenia.

Journal of Nervous and Mental Disease, 182, 656-660.

Lysaker, P. H., & Bell, M. D. (1995). Work rehabilitation and improvements in insight in

schizophrenia. Journal of Nervous and Mental Disease, 183, 103-106.

Lysaker, P. H., & Bell, M. D. (1998). Impaired insight in schizophrenia: Advances from

psychosocial treatment research. In X. F. Amador & A. S. David (Eds.), Insight

and psychosis (pp. 307-316). New York: Oxford University Press.

Lysaker, P. H., Bell, M. D., & Bioty, S. M. (1995). Cognitive deficits in schizophrenia:

prediction of symptom change for participants in work rehabilitation. Journal of

Nervous and Mental Disease, 183, 332-336.

Lysaker, P. H., Bell, M. D., Milstein, R., Bryson, G., & Beam-Goulet, J. (1994). Insight

and psychosocial treatment compliance in schizophrenia. Psychiatry, 57, 307-315.

Markova, I. S., & Berrios, G. E. (1992). The assessment of insight in clinical psychiatry:

A new scale. Acta Psychiatrica Scandinavia, 86, 159-164.

McEvoy, J. P. (1998). The relationship between insight in psychosis and compliance with

medications. In X. F. Amador & A. S. David (Eds.), Insight and psychosis (pp.

289-306). New York: Oxford University Press.

McEvoy, J. P., Apperson, L. J., Applebaum, P. S., Ortlip, P., Brecosky, J., & Hammill, K.

(1989a). Insight into schizophrenia: its relationship to acute psychopathology.

Journal of Nervous and Mental Disease, 177, 43-47.

McEvoy, J. P., Freter, S., Everett, G., Geller, J. L., Appelbaum, P., Apperson, L. J., and

Roth, L. (1989b). Insight and the clinical outcome of schizophrenics. Journal of

Nervous and Mental Disease, 177, 48-51.

63

McEvoy, J. P., Applebaum, L, Apperson, J., Geller, J. L., & Freter, S. (1989c). Why must

some schizophrenic patients be involuntarily committed? The role of insight.

Comprehensive Psychiatry, 30, 13-17.

McEvoy, J. P., Freter, S., Merrit, M., & Apperson, L. J. (1993). Insight about psychosis

among outpatients with schizophrenia. Hospital and Community Psychiatry, 44,

883-884.

McEvoy, J. P., Hartman, M., Gottlieb, D., Godwin, S., Apperson, L. J., Wilson, W.

(1996). Common sense, insight, and neuropsychological test performance in

schizophrenia patients. Schizophrenia Bulletin, 22, 635-641.

McEvoy, J. P., Schooler, N. R., Friedman, E., Steingard, S., & Allen, M. (1993a). Use of

psychopathology vignettes by patients with schizophrenia or schizoaffective

disorder and by mental health professional to judge patients insight. American

Journal of Psychiatry, 150, 1649-1653.

McGlynn, S. M., & Schacter, D. L. (1989). Unawareness of deficits in

neuropsychological syndromes. Journal of Experimental and Clinical

Neuropsychology, 11, 143-205.

Michalakeas, A., Skoutas, C., Charalambous, A., Peristeriis, A., Marinos, V., Keramri, E.

et al. (1994). Insight in schizophrenia and mood disorders and its relation to

psychopathology. Acta Psychiatrica Scandinavica, 90, 46-49.

Mintz, A. R., Addington, J., & Addington, D. (2004). Insight in early psychosis: A 1-

year follow-up. Schizophrenia Research, 67, 213-217.

64

Moore, O., Cassidy, E., Carr, A., & O’Callaghan. (1999). Unawareness of illness and its

relationship with depression and self-deception. European Psychiatry, 14, 264-

269.

Nayani, T., & David, A. S. (1996). The auditory hallucination: A phenomenological

survey. Psychological Medicine, 26, 29-38.

Neale, J. M. (1971). Perceptual span in schizophrenia. Journal of Abnormal Psychology,

77, 196-204.

Neale, J. M., McIntyre, C. W., Fox, R., & Cromwell, R. L. (1969). Span of Apprehension

in acute schizophrenics. Journal of Abnormal Psychology, 74, 593-596.

Nelson, E. B., Sax, K. W., & Strakowski, S. M. (1998). Attentional performance in

patients with psychotic and nonpsychotic major depression and schizophrenia.

American Journal of Psychiatry, 155, 137-139.

Neumann, C. S., Baum, K. M., Walker, E. F., & Lewine, R. J. (1996). Childhood

behavioral precursors of adult neuropsychological functioning in schizophrenia.

Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 9, 221-229.

Neumann, C. S., Bryan, N. C., Avery, & Cromwell, R. L. (1997). Schizotypal symptoms,

affect and handedness in college students from a rural southern state university.

Poster presentation presented at the 12th Annual Conference of the Society for

Research in Psychopathology, Palm Springs, CA.

Neumann, C. S., Walker, E. F., Weinstein, J., & Cutshaw, M. A. (1996). Psychotic

patients’ awareness of mental illness: implications for legal defense proceedings.

Journal of Psychiatry and Law, Fall, 421-442.

65

Nicholson, I. R., Chapman, J. E., & Neufeld, R. W. (1995). Variability in BPRS

definitions of positive and negative symptoms. Schizophrenia Research, 17, 177-185.

Nuechterlein, K. H., & Dawson, M. E. (1984). Information processing and attention in

the developmental course of schizophrenic disorders. Schizophrenia Bulletin, 18,

387-425.

Overall J. E., & Gorham, D. R. (1962). The Brief Psychiatric Rating Scale. Psychological

Reports, 10, 799-812.

Orzack, M. H., & Kornetsky, C. (1966). Attention dysfunction in chronic schizophrenia.

Archives of General Psychiatry, 14, 323-326.

Orzack, M. H., & Kornetsky, C. (1971). Environmental and familial predictors of

attention behavior in chronic schizophrenics. Journal of Psychiatric Research, 9,

21-35.

Peralta, V., & Cuesta, M. J. (1994). Lack of insight: Its status within schizophrenic

psychopathology. Biological Psychiatry, 36, 559-561.

Perry, W., & Braff, D. (1994). Information processing deficits and thought disorder in

schizophrenia. American Journal of Psychiatry, 151, 363-367.

Roitman, S. E., Keefe, R. S., Harvey, P. D., Siever, L. J., Mohs, R. C. (1997). Attentional

and eye tracking deficits correlate with negative symptoms in schizophrenia.

Schizophrenia Research, 26, 139-146.

Sanz, M., Constable, G., Lopez-Ibor, I., Kemp, R., & David, A. S. (1998). A comparative

study of insight scales and their relationship to psychopathological and clinical

variables. Psychological Medicine, 28, 437-446.

66

Schwartz, R. C., & Smith, S. D. (2004). Suicidality and psychosis: The predictive

potential of symptomatology and insight into illness. Journal of Psychiatric

Research, 38, 185-191.

Serper, M. R., Bergman, R. L., & Harvey, P. D. (1990). Medication may be required for

the development of automatic information processing in schizophrenia. Psychiatry

Research, 32, 281-288.

Silverstein, M. L., & Zerwic, M. J. (1985). Clinical psychopathologic symptoms in

neuropsychologically impaired and intact schizophrenics. Journal of Consulting

and Clinical Psychology, 53, 267-268.

Smith, T. E., Hull, J. W., Israel, L. M., & Wilson, D. F. (2000). Insight, symptoms and

neurocognitive functioning ins schizophrenia and schizoaffective disorder.

Schizophrenia Bulletin, 26, 193-200.

Smith, T. E., Hull, J. W., & Santos, L. (1998). The relationship between symptoms and

insight in schizophrenia: A longitudinal perspective. Schizophrenia Research, 33,

63-67.

Spielberger, C. D., Gorsuch, R. G., & Lushene, R. E. (1970). Manual for the State-Trait

Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press.

Steffy, R. A., & Oakman, J. M. (1997). Attentional functioning in individuals diagnosed

and at risk for schizophrenia. In J. A. Burack & J. T. Enns (Eds.), Attention,

development, and psychopathology (pp. 318-349). New York: Guilford Press.

Stratta, P., Daneluzzo, E., Prosperini, P., Bustini, M., Mattei, P., & Rossi, A. (1997). Is

Wisconsin Card Sorting Test performance related to ‘working memory’ capacity?

Schizophrenia Research, 27, 11-19.

67

Strauss, J. S. (1998). Epilogue. In X. F. Amador & A. S. David (Eds.), Insight in

psychosis (pp. 352-358). New York: Oxford University Press.

Strauss, M. E., Buchanan, R. W., & Hale, J. (1993). Relations between attentional deficits

and clinical symptoms in schizophrenic outpatients. Psychiatry Research, 47, 205-

213.

Stuss, D. T., & Benson, D. F. (1986). The frontal lobes. New York: Raven Press.

Stuss, D. T., Benson, D. F., & Kaplan, E. F., (1983). The involvement of orbitofrontal

cerebrum in cognitive tasks. Neuropsychologia, 21, 235-248.

Takai, A., Uematsu, M., Hirofumi, U., Sone, K., & Kaiya, H. (1992). Insight and its

related factors in chronic schizophrenic patients: A preliminary study. European

Journal of Psychiatry, 6, 159-170.

Taylor, K. E., & Perkins, R. E. (1991). Identity and coping with mental illness in long-

stay psychiatric rehabilitation. British Journal of Clinical Psychology, 30, 73-85.

Van Os, J., Fahny, T., Jones, P., Harvey, I., Sham, P.., Lewis, S., Toone, B., Williams,

M., & Murray, R. (1996). Psychopathological syndromes in the functional

psychosis: Associations with course and outcome. Psychological Medicine, 26, 161-176.

Van Putten, T. (1974). Why do schizophrenic patients refuse to take their medication?

Archives of General Psychiatry, 31, 67-72.

Van Putten, T., Crumpton, E., & Yale, C. (1976). Drug refusal in schizophrenia and the

wish to be crazy. Archives of General Psychiatry, 21, 1443-1446.

68

Ventura, J., Green, M. F., Shaner, A., & Liberman, R. P. (1993). Training and quality

assurance with the Brief Psychiatric Rating Scale: ‘The drift busters.’

International Journal of Methods in Psychiatric Research, 3, 221-244.

Voruganti, L. N., Heslegrave, R. J., & Awad, A. G. (1997). Neurocognitive correlates of

positive and negative syndromes in schizophrenia. Canadian Journal of

Psychiatry, 42, 1066-1071.

Walker, E., & Lewine, R. J. (1988). The positive/negative symptom distinction in

schizophrenia: Validity and etiological relevance. Schizophrenia Research, 1,

315-328.

Walker, E., & Rossiter, J. (1989). Schizophrenic patients’ self-perceptions: Legal and

clinical implications. Journal of Psychiatry and Law, Spring, 55-73.

Watson, D., Clark, A. L., & Tellegen, A. (1988). Development and validation of brief

measures of positive and negative affect: The PANAS Scales. Journal of

Personality and Social Psychology, 54, 1063-1070.

Wing, J. K., Cooper, J. E., & Sartorius, N. (1974). The measurement and classification of

psychiatric symptoms. Cambridge: Cambridge University Press.

Winters, L., Cornblatt, B. A., & Erlenmeyer-Kimling, L. (1991). The prediction of

psychiatric disorders in late adolescence. In E. Walker (Ed.), Schizophrenia: A

life-course developmental perspective (pp. 123-137). New York: Academic Press.

Young, D. A., Davila, R., Scher, I. I. (1993). Unawareness of illness and

neuropsychological performance in chronic schizophrenia. Schizophrenia

Research, 10, 117-124.

69

Young, D. A., Zakzanis, K. K., Bailey, C., Davila, R., Griese, J., Sartory, G., & Thom, A.

(1998). Further parameters of insight and neuropsychological deficit in

schizophrenia and other chronic mental disease. Journal of Nervous and Mental

Disease, 186, 44-50.