Classification of schizoaffective disorder


Material Information

Classification of schizoaffective disorder
Physical Description:
vii, 226 leaves : ill. ; 29 cm.
Sprock, June, 1955-
Publication Date:


Subjects / Keywords:
Schizophrenia   ( mesh )
Clinical and Health Psychology thesis Ph.D   ( mesh )
Dissertations, Academic -- Clinical and Health Psychology -- UF   ( mesh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1986.
Bibliography: leaves 217-224.
Statement of Responsibility:
by June Sprock.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000578656
oclc - 16849963
notis - ADA6436
System ID:

This item is only available as the following downloads:

Full Text








I especially would like to thank Roger K. Blashfield, my chairman,

for his guidance and assistance throughout this project. Not only did

he provide support and thoughtful suggestions during all phases of this

research, but he also spent many hours assigning symptom ratings to

cases and compiling a glossary to consolidate the individual subject

feature lists. His ideas about classification have had a major

influence on my thinking about psychopathology.

I also would like to extend my thanks to Warren J. Rice for serving

as a symptom rater for the case histories and in assisting in the

feature list consolidation process and to Russell M. Bauer for his work

in rating the symptoms in the schizoaffective cases.

Finally, I would like to thank Rick Koenig for his encouragement

and patience, without which this project could not have been completed.



ACKNOWLEDGMENTS....................................................... ii


I SCHIZOAFFECTIVE DISORDER............................ 1

Statement of Problem ........................ .........1
History of Concept.................................. 2
Classification...................................... 5
Hypothetical Classifications of Schizoaffective

II PROTOTYPE MODEL.....................................52

Introduction....................................... 52
Standard Prototype Methodology......................55
Related Research................................... 62
Study to Test Alternative Models of
Schizoaffective Disorder..........................69

III SCHIZOAFFECTIVE EXEMPLARS: I........................73

Method............................................. 75
Results .............................................89
Discussion......................................... 99

IV SCHIZOAFFECTIVE EXEMPLARS: II ......................105

Method............................................ 107
Results........................................... 108

V FEATURE LISTS......................................137

Method............................................ 139

VI SIMILARITY RATINGS..................................165

Method............................................ 167
Results ............................................ 169

VII SUMMARY AND CONCLUSIONS............................184

History of the Classification of Schizoaffective
The Prototype Perspective.......................... 188
Summary of Studies................................ 189
Implications for Models of Schizoaffective
Comments, Applications and Directions for
Future Research..................................195


TWO RATERS FOR 17 CASES..........................212
SYMPTOMS FOR 17 CASES............................213
FROM 0 TO 2 FOR TWO RATERS .......................214
AND THE AFFECTIVE DISORDERS......................216
CASES ............................................217
CRITERIA SYMPTOMS................................218
CASES .............................................219
THREE RATERS AND MEANS...........................220
FOR ASSIGNMENT TO THE 30 CASES...................222


BIOGRAPHICAL SKETCH..................................................231

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



June Sprock

August 1986

Chairman: Roger K. Blashfield
Major Department: Clinical Psychology

This series of studies examined the viability of the major

alternative models of the classification of schizoaffective disorder:

that it is a subtype of schizophrenia, a type of affective disorder, on

a continuum between schizophrenia and affective disorder, a totally

independent disorder, or a heterogeneous category. The methodologies,

taken from the prototype view of classification in cognitive psychology,

study the behavior of the classifiers, in this case, clinicians

assigning psychiatric diagnoses. These paradigms are especially useful

in the study of schizoaffective disorder since different definitions of

this diagnosis have resulted in widely discrepant findings. Therefore,

finding the concept of schizoaffective disorder with the highest

consensus is a reasonable first step before conducting empirical studies

of patients.

The first study attempted to locate exemplars or typical

schizoaffective cases based on high agreement and fast decision time

when 20 clinicians assigned diagnoses to cases, including 15

schizoaffective cases from the literature. When none were found, it


was hypothesized that the large number of categories may have decreased

diagnostic consensus. When the study was repeated using fewer

categories, exemplars were found for other diagnoses, but not for

schizoaffective disorder. Schizoaffective cases were most often

diagnosed as schizophrenia. In a second study, 20 clinicians listed the

essential clinical features of schizoaffective disorder and other major

psychoses. The schizoaffective list shared as many features with

schizophrenia as did the schizophrenic subtypes, but shared only one

feature with the affective disorders. In the final study, 20 clinicians

assigned similarity ratings to a series of pairs of diagnoses.

Multidimensional scaling analysis showed schizoaffective disorder was

rated most similar to the schizophrenic spectrum of disorders and to

major depression with psychotic features. Students viewed

schizoaffective disorder as more similar to affective disorder while

faculty saw it as related to schizophrenia.

The overall results suggest that these clinicians perceived

schizoaffective disorder as a variant of schizophrenia. However, this

view of schizoaffective disorder may be changing in newer clinicians.

Implications for defining schizoaffective disorder, suggestions for

diagnosis, the generalizability of these findings and directions for

future research are discussed.


Statement of Problem

Schizoaffective disorder has been a controversial diagnostic

category since its introduction in 1933 by Kasanin. Even today, there is

still little consensus on whether it is an independent diagnostic

disorder, a subtype of schizophrenia or affective disorder, or a

borderline condition that falls somewhere along a continuum from

schizophrenia to affective disorder (Welner, Croughan & Robins, 1974;

Procci, 1976; Spitzer, Endicott & Robins, 1978; A.P.A., 1980). The

purpose of this study is to attempt to explicate the concept of

schizoaffective disorder and determine where it fits into the

classification of functional disorders.

First, the history of the development of the concept,

schizoaffective disorder, is reviewed, followed by a discussion of

diagnostic criteria. Alternative models of schizoaffective disorder in

the classification of functional disorders are then hypothesized based

upon a review of past research. In Chapter II, a group of methodologies

used to test the various conceptualizations of schizoaffective disorder,

the prototype model, is discussed. Prototype studies with findings

relevant to the classification of schizoaffective disorder are also

described. In chapters III through VI a series of studies utilizing

several different prototype methodologies to test the hypothetical

alternative models of schizoaffective disorder are presented. Chapter

VII pulls together the results from the four studies to arrive at some

conclusions about the classification of schizoaffective disorder based

on convergent findings.

History of Concept

Historically, since the term was coined by Kasanin in 1933,

schizoaffective disorder has been considered a subtype of schizophrenia

(Lehman, 1975). Kasanin used it to describe a group of young

schizophrenic patients with good premorbid adjustment, a sudden onset

and favorable prognosis. In his article, he provided a thorough account

of 5 patients initially diagnosed dementia praecox, whom he felt

represented a distinct subgroup, schizoaffective. All 5 were young,

intelligent, well-integrated individuals who experienced a sudden

psychosis, often ushered in by depression, and seemingly in response to

a difficult environmental situation. They exhibited very little bizarre

behavior during the episode and had a complete recovery, despite the

tendency for later repeat episodes. Kasanin believed that

differentiating homogeneous diagnostic subgroups within the major

diagnostic concepts was necessary in order for psychiatric research to

yield consistent results. Kasanin cited parallels to his

schizoaffective concept in earlier diagnoses discussed by Bleuler,

Dunton and Claude.

Bleuler (1920) recognized that there existed a group of patients in

which differentiation between schizophrenia and affective disorder is

extremely difficult. However, he viewed such cases with mixed affective

and schizophrenic symptomatology as schizophrenic, stating that

affective symptoms can occur in schizophrenia but not the reverse. No

matter how much affective symptomatology was present, the patient was

diagnosed as schizophrenic if what he considered "fundamental" symptoms

of schizophrenia were present (Tsuang and Simpson, 1984). Dunton's

(1910a) cyclic dementia praecox patients had manic symptomatology with

the addition of stereotypy, and despite normal functioning between

episodes, an eventual deterioration. In his intermittent dementia

praecox (Dunton,1910b), depression or excitement preceded development of

psychosis and there was a return to normal functioning afterwards.

Claude, Borel and Robins (1924) hypothesized that there were a series of

stages of schizophrenia. The first, schizomania, was characterized by

affective ideation and an extremely rich fantasy life which was seen in

the delusions. Patients may remain at this stage or progress to the

second and third stages, schizophrenia and dementia, which have marked

thought disturbance and deterioration.

At the time, Polozker (1933) and Von Domarus (1933) responded quite

favorably to Kasanin's concept, although somewhat skeptical at first

whether introducing another new term may simply add confusion to the

classification of psychoses. Both agreed that the term was a

significant improvement to the nosology by describing a homogeneous

subgroup of schizophrenics with onset, premorbid adjustment and

prognosis opposite to that usually seen in schizophrenics. Recently,

Stone (1980) has criticized Kasanin's cases for being quite

heterogeneous with respect to symptomatology. According to Stone, two

exhibited clear manic-depressive symptoms, another pure schizophrenic

symptoms, and only two showed a mixture of schizophrenic and affective

symptoms which would be called schizoaffective today.

Since the use of the term schizoaffective by Kasanin, a number of

similar syndromes have been described utilizing other names. In a

relatively recent review, Procci (1976) compared 11 different historical

definitions of schizoaffective disorder. While he found considerable

consistency concerning remission, symptomatology, onset and premorbid

adjustment, he found a plethora of names for this disorder. For example,

Hoch and Polatin (1949) discussed a borderline category between

schizophrenia and neurosis, often accompanied by affective

symptomatology, which they labeled pseudoneurotic schizophrenia. Only

Cobb (1943) and Astrup and Noreik (1966) labeled the disorder

schizoaffective. While Procci drew a parallel to earlier syndromes

described by Kirby (1913) and Hoch (1921), actually these syndromes were

not that similar and consisted of acute onset catatonia with full

recovery. Langfeldt's (1937) schizophreniform patients more closely

resembled Kasanin,s concept, having an acute onset psychosis, affective
symptoms, especially depression, a family history of depression and good

outcome. Kant (1941) also wrote about a subgroup of young

schizophrenics with affective symptoms, acute onset and a family history

of manic-depressive disorder.

In the encyclopaedic Comprehensive Textbook of Psychiatry (second

edition) (Lehman,1975), the clinical features of schizoaffective

disorder were described as "strong depression or euphoria in an

otherwise schizophrenic symptomatology", with a prognosis between

schizophrenia and manic-depression. Lehmann wrote that the

schizophrenic symptoms are the decisive factor in making this diagnosis.

However, he noted that several researchers had raised the question of

whether this disorder should be considered an affective disorder. In

the same volume, Linn (1975) wrote that a diagnosis of schizoaffective

disorder is given when differential diagnosis between schizophrenia and

affective disorder is not possible, emphasizing the borderline quality

of the concept.


Because of the influence of Bleuler in the classification of

psychopathology, schizoaffective came to to be regarded as a

schizophrenic disorder. In 1953, Coble introduced the compound term

"schizoaffective schizophrenia" which he described as having the

classical symptoms of schizophrenia with strong euphoria or depression.

The historical view that schizoaffective disorder is a subtype of

schizophrenia was still represented in the DSM-II (APA, 1968) and ICD-9

(WHO,1969). The DSM-II described schizophrenia, schizo-affective

subtype as a category "for patients showing a mixture of schizophrenic

symptoms and pronounced elation or depression." The DSM-II further

suggested distinguishing "excited" and "depressed" subtypes.

Differentiation from major affective illness was determined by the

domination of thought disorder rather than mood disorder.

In reaction to criticism of the reliability and validity of the

vague prose definitions used in the DSM-II, several classifications were

developed that attempted to provide operational definitions and explicit

criteria for diagnoses. The Feighner criteria (Feighner, Robins, Guze,

Woodruff, Winokur & Munoz, 1972), did not attempt to provide criteria

for this disorder at all, stating that only the 14 disorders with

sufficient reliability and validity were included in the nosology. In

the Research Diagnostic Criteria (RDC) (Spitzer, Endicott & Robins,

1975), schizoaffective disorder was for the first time viewed as a

distinct category and placed at the same level as schizophrenia. The

RDC required 4 of 7 manic or 5 of 8 depressive symptoms plus one of 5

schizophrenic symptoms for at least one week for a schizoaffective

diagnosis. RDC schizo-affective disorder was divided into manic and

depressive subtypes and further subclassified as mainly schizophrenic or

mainly affective, based on which type of symptoms had an earlier onset,

and as acute, subacute, chronic and subchronic. The acute subtype was

assigned if the duration of symptoms was less than 3 months and there

was a full recovery, chronic was assigned when the symptoms were

continuously present for at least 2 years, while the remaining subtypes

were in between and not as clearly defined.

Despite the very specific criteria in the RDC, the reliability for

schizoaffective disorder was at best moderate. Using coefficient kappa,

Spitzer et al. (1975) found an interrater reliability of only .48 for

the manic subtype, but as high as .86 for the depressive subtype,

compared to .80, .93 and .88 for schizophrenia, bipolar mania and major
depression. In this study 68 newly admitted inpatients were interviewed

by pairs of clinicians. While one clinician conducted the interview,

both made RDC diagnoses afterwards. In a second article, Spitzer,

Endicott and Robins (1978) reported the results of two interrater and

one test-retest reliability study. The interrater reliability studies

used 68 and 150 inpatients, respectively, who were jointly interviewed

by two psychiatrists. Kappa coefficients of agreement for

schizoaffective, depressed were very consistent and quite good across

studies (.86 and .85). No values were reported for schizoaffective,

manic because this diagnosis was used with a frequency of less than 5%

by either clinician. In the test-retest study, a second group of 60

inpatients were interviewed and then re-interviewed within 2 days using

a structured interview. Quite high kappa coefficients were found for

both the depressive (k=.73) and manic (k=.79) subtypes. The ROC has

been widely used in research on schizoaffective disorder because of the

good reliability of its operational definitions and because the

divisions into depressive versus manic, mainly schizophrenic versus

mainly affective and acute versus chronic subtypes have been found to be

meaningful distinctions and useful in research (Procci, 1976; Mattes &

Nayak, 1984; Maj, 1984).

The initial draft (1977) of the DSM-III returned schizoaffective

disorder as a subtype of schizophrenia. Both schizoaffective, depressed

and manic required meeting the criteria for schizophrenia and then

meeting the criteria for mood disorder, and temporal overlap of

psychotic and affective symptoms. A second draft published in 1978 then

moved it as an independent disorder at the same hierarchical level as

schizophrenia and affective disorder based on feedback from clinicians

and researchers. The Task Force justified the change from being a
schizophrenic subtype to being an independent category as a response to

accumulated evidence that patients with mixtures of schizophrenic and

affective symptoms have better prognosis, more rapid onset and increased

probability of returning to premorbid levels of functioning than other

schizophrenics. In addition, they did not show the increased familial

incidence of schizophrenia seen in other schizophrenics. Yet, the Task

Force was still somewhat noncommittal in their position about where this

category should fall in the classification of functional disorders.

While they placed it as an independent disorder, they described the

clinical features as an admixture of affective and schizophrenic

symptomatology and admitted that it is still a controversial category.

Interestingly, the diagnostic code (295.7x) still reflected a

schizophrenic subtype (295.0-295.9) in order to be in line with the ICD-


The criteria in this second version also showed a radical shift

from the earlier draft. First, the patient had to meet criteria for a

manic or depressive episode except for the duration criterion. Then,

only one of 9 possible schizophrenic symptoms were necessary. A

duration of one week and temporal overlap of psychotic with affective

symptoms were also required. Subtyping was done according to

phenomenology (manic, depressed, mixed) and course (episodic: single,

recurrent, chronic, in remission). This draft also provided

supplementary text about the disorders it contained. The text described

schizoaffective disorder (note the elimination of the hyphen) as a

depressive or manic syndrome that precedes or develops concurrently with

certain psychotic symptoms. These initial criteria resulted in low
interrater and test-retest reliability (Spitzer, Forman & Nee, 1979).

Using joint interviews of 150 inpatients and outpatients, a kappa

coefficient of .56 was found for schizoaffective disorder (not

differentiated according to subtype). Test-retest kappa values were

similar (k=.53). Of the 131 patients interviewed, 40% of the second

interviews were conducted within one day of each other, the remaining

conducted greater than three days apart.

The final draft of the DSM-III placed it under the category
"psychotic disorders not elsewhere classified". No specific criteria

were listed because of the low reliability and lack of consensus about

the concept. A prose definition reminiscent of the DSM-II was provided,

however. Schizoaffective disorder was defined as a "depressive or manic

syndrome concurrent with psychotic symptoms incompatible with affective

disorder". The DSM-III also gave several examples of syndromes that

would be classified as schizoaffective, but suggested that clinicians

try to use other diagnostic categories with criteria, such as major

depression with psychotic features or schizophrenia with superimposed

atypical affective disorder. Their description of the syndrome closely

paralleled the 1978 draft and emphasized the temporal relationship of

onset of schizophrenic-like and affective-like symptoms. In DSM-III

schizoaffective disorder, affective symptoms must appear first or the

onset of symptoms must coincide. If psychotic symptoms develop first, a

diagnosis of schizophrenia with secondary depression may be more

appropriate. Tsuang and Simpson (1984) noted how the final version of

the DSM-III only resulted in increased confusion about this concept.

Two other sets of operational criteria to define schizoaffective

disorder have been used in a number of research projects. The first are

the criteria of Welner, Croughan and Robins (1974) developed as part

of a comprehensive series of studies to examine the classification of

schizoaffective disorder. Four polythetic criteria were required.

First, was a sufficient number of severe affective manifestations to

make the diagnosis of schizophrenia unlikely. In addition to having

depressed mood for at least one month, the patient must have at least 4

of 8 depressive symptoms or, instead, the patient may have 2 weeks of

euphoria or irritability and at least 3 of 6 manic symptoms. However,

they note that the criteria are experimental and may be modified so that

a patient may not meet either criteria but instead have 6 symptoms of

affective disorder. The second criterion is a sufficient number of

severe thought and behavioral disorders to make a diagnosis of affective

disorder unlikely (i.e., at least 2 of 5 schizophrenic symptoms). In

addition, the patient must have either an acute onset, episodic course

or pronounced confusion or perplexity. Finally, the disorder must not

be associated with alcohol or drug abuse or organic brain disorder.

While the reliability of these criteria have not been reported, the

individual symptoms are operationally defined and many closely resemble

the Feighner and the RDC criteria which appeared shortly afterwards.

The second set of operational criteria for schizoaffective disorder

were developed by Brockington, Kendell, Kellett, Curry and Wainwright

(1978) in England. While Welner et al. (1974) were forced to develop

their own experimental criteria since the Feighner criteria did not

include schizoaffective disorder and the RDC did not exist at that time,

Brockington et al. chose to develop a different set of criteria based on

work in England, particularly the Present State Exam (PSE) of Wing

(Wing, Cooper, Sartorius, 1974). In order to receive a schizoaffective

diagnosis a patient must fulfill the criteria for schizophrenia or a
paranoid psychosis and also for depression or mania. The criteria were

less clearly specified than in the previous classifications but users

were referred to the PSE for more detailed definitions. A schizophrenic

diagnosis required one nuclear symptom of 7 (e.g., thought insertion)

and two symptoms from each of 3 symptom groups, behavioral (e.g.,

mannerisms), affect (e.g., blunted affect) and speech (e.g.,

neologisms). The criteria for paranoid psychosis required a persistent,

preoccupying delusion of one of 8 types (e.g., influence, reference,

persecution). The criteria for depression required 4 of 16 symptoms and

3 signs, and for mania, 3 of 5 symptoms and 7 signs. Schizoaffectives

could be subdivided into "schizomanics" and "schizodepressives"

according to whether they fulfilled the criteria for mania or

depression. These criteria contrasted with those of Welner and the RDC

for schizoaffective disorder by including the possibility of a paranoid

psychosis and not requiring a specific duration of symptoms.

A number of studies have examined the concordance between the

various definitions of schizoaffective disorder. Brockington and Leff

(1979) found a very low rate of concordance using kappa (range: k = -.03

to .50, mean: k = .19) between 8 definitions of schizoaffective disorder

including the 3 discussed above, the combined Feighner criteria for

schizophrenia and affective disorder, Kasanin's (1933) original

definition and Perris' (1974) definition of cycloid psychosis. While

cycloid psychosis has often been included with schizoaffective disorder,

for example, in the ICD-9, studies by Brockington, Perris, Kendell,

Hillier and Wainwright (1982), Maj (1984) and Vogl and Zaudig (1985)

have found little concordance between the two concepts and suggest that

they represent substantially different disorders and should not be


Although Maj (1984) did not specifically study diagnostic

concordance, he applied 3 operational definitions of schizoaffective

disorder and Perris' definition of cycloid psychosis to a group of 38

inpatient schizoaffectives who fulfilled the ICD-9 definition for

schizophrenia, schizoaffective type. While 30 met Kendell's criteria

(12 manic and 18 depressive type) and 25 met the RDC for schizoaffective

disorder (10 manic and 15 depressive subtype), only 13 met the Welner

criteria and 15, Perris' cycloid psychosis. Maj did not directly

examine concordance between the definitions. Nonetheless, the results

suggest that there is little overlap between schizoaffective and the
cycloid psychoses and that the Welner criteria may be too narrow.

Vogl and Zaudig (1985) compared the ward diagnosis and their own
diagnosis of 84 functional psychotic inpatients to 4 operational

definitions of schizoaffective disorder, (i) Welner's, (ii) Brockington

and Kendell's, (iii) the RDC and (iv) the combined Feighner criteria for

schizophrenia and affective disorder, as well as Perris' definition of

cycloid psychosis. Schizoaffectives were divided into "schizomanic" and
"schizodepressive" subtypes. The authors' diagnoses were made

independently based on clinical interview, then disagreements were

resolved through discussion. The agreement between their final

diagnoses and the ward diagnoses was 95%. The 4 sets of operational
criteria were applied by the authors after the interviews using

checklists containing all of the criteria. They found a high level of

concordance between the Welner and Brockington criteria and between

these two definitions and the ward and author's diagnoses. Looking at

the coverage of each definition showed that 20 patients had ward

diagnoses of schizoaffective disorder, 15 were diagnosed schizoaffective
by the authors, 15 met the Welner definition, 14, the Brockington
definition and 11, the RDC. In contrast, only 6 met the Perris criteria

for cycloid psychosis and none met the combined schizophrenia and

affective disorder Feighner criteria. When the operational definitions

were compared to the ward diagnoses, 9 of the 26 paranoid schizophrenics

(35%) fulfilled the Brockington definition, 8 met the RDC (31%) and 6
met the Welner criteria (23%) for schizoaffective disorder. Between 33

and 50% of the remaining schizophrenic subtypes met the Brockington and

RDC definitions of schizoaffective disorder, while only 4 depressives

(26%) and no manics met these two definitions of schizoaffective
disorder. The results were almost identical when comparing the authors'

diagnoses to the operational definitions of schizoaffective and cycloid


Overall, the Brockington criteria had the highest concordance with

the authors' diagnoses (73%). While these criteria had the highest

sensitivity, they also diagnosed the largest number of schizophrenics as

schizoaffective, probably because of the broad conceptualization of

schizoaffective and lack of duration criteria. The Welner criteria also

had high sensitivity possibly because mood-congruent and mood-

incongruent delusions are permitted. The authors concluded that these

two sets of criteria are best suited for diagnosing schizoaffective

disorder. The Feighner and ROC definitions were considered unsuitable

because of very high specificity primarily due to very restricted

conceptualizations of schizophrenia. These results contrast with those

of Maj (1984) who found that the Welner definition was too restrictive

and the RDC had broad coverage, though both found the Brockington

criteria to have broad coverage. Vogl and Zaudig also concluded that

the low concordance of the cycloid psychoses with schizoaffective

disorder argues against including it with schizoaffective disorder.

Hypothetical Classifications of Schizoaffective Disorder

A review of the literature suggests several plausible alternative

conceptualizations of schizoaffective disorder in the classification of

functional disorders. Schizoaffective disorder may be viewed as a

distinct disorder or a borderline condition along a continuum from

schizophrenia to affective disorder. Other possibilities are that it is

a subtype of schizophrenia or affective disorder. While these are the

major historical conceptualizations of schizoaffective disorder, a newer

view is that schizoaffective disorder may be heterogeneous and may

consist of a subtype related to schizophrenia and another related to

affective disorder. Two variables may also be influential in diagnosing
schizoaffective disorder. The first is the temporal relationship of the

onset of symptoms. The DSM-III, for example, differentiates

schizoaffective disorder from residual schizophrenia with secondary

depression according to the temporal onset of symptoms. Another

possibility is that treatment response plays a role in the decision

to assign a schizoaffective, schizophrenic or affective disorder

diagnosis. Patients who respond to tricyclics are called major

depression with psychotic features, while those who improve with

antipsychotics are labeled schizophrenic or schizoaffective. Evidence

concerning each of these possible models is presented below.

Subtype of Schizophrenia or Affective Disorder

Traditionally, schizoaffective disorder has been viewed as a subtype

of schizophrenia. In fact, both the DSM-II and the initial draft of the

DSM-III retained this classification. Hoch and Polatin (1949) and more

recently, Kolb (1973) wrote that schizoaffective disorder is a variation

of schizophrenia. According to this position, positive clinical
characteristics such as high intelligence and good premorbid adjustment

in conjunction with a predisposition for schizophrenia result in an

attenuated form of schizophrenia. Indeed Stone (1980) questioned

whether schizoaffective disorder is actually a dilute form of

schizophrenia in which there is a genetic predisposition, though perhaps

milder, so that normal affectivity is intact. Therefore, the affective

symptoms in schizoaffectives are seen as normal mood and affect in the

context of schizophrenic thought disorder. They may appear as affective
disorder because these symptoms contrast with the usually flat, blunted

affect seen in schizophrenics.

Another view is that the term schizoaffective is superfluous

because depression is a common features of schizophrenia (Bleuler, 1920;

Vaillant, 1963; Batchelor, 1969; Procci, 1976). Both Vaillant and

Procci argued that depression is merely a positive prognostic sign in

schizophrenia. Further, they claimed that the term "schizoaffective" is

a misnomer because it emphasizes the affective symptom which is only one

of a number of symptoms.

Welner has completed a series of studies of schizoaffective

disorder that provided some support for it being a subtype of

schizophrenia. An early study by Croughan, Welner and Robins (1974)

applied the Welner criteria in a record review of 266 discharged

inpatients with a diagnosis of schizoaffective or related a psychosis,

either schizophreniform or acute schizophrenia. Only patients with at

least one of the 5 schizophrenic symptoms in the Welner criteria were

included. Charts were reviewed for 225 psychiatric symptoms and 137

variables concerning demographic information, premorbid functioning, age

of onset, number of hospitalizations and past diagnoses, and family

history of psychopathology. Subjects were divided into two groups,

those that met criteria for depression or mania (group A) and those with

insufficient affective symptoms (group B). Groups were then reduced by

using only subjects with at least 2 schizophrenic symptoms; the

resulting groups were compared on symptoms and course. Both groups had

a chronic, nonremitting course (A: 79%, B: 76%). There was, however, a

significant difference in the type of chronic symptoms in the groups

with group A patients having chronic affective symptoms and group B

patients having permanent thought and behavior disorders. They

concluded there is a definite distinction between the 2 groups and that

the group B patients were best diagnosed as paranoid schizophrenia. It

is not clear whether these patients represented a schizophrenic subgroup

of schizoaffectives or schizophrenic patients who were included because

of the broad sampling method. The authors did not reach any

conclusions, however, about whether the A group was a variant of

schizophrenia because of the chronic course, affective disorder because

of meeting the affective disorder criteria or a distinct disorder

because of differences in symptoms from schizophrenia and difference in

the course from affective disorder. They recommended doing a followup

study to determine if differences in outcome or family history were

associated with these groups of patients.

In a 5 year followup study of these same 114 schizoaffectives,

Welner, Croughan, Fishman and Robins (1977) found a chronic course in

71% and occupational or marital deterioration in 80%. All patients met

Welner's criteria for schizoaffective disorder to insure that no

schizophrenics were included. Followup was done with a personal

interview using a comprehensive interview form. They argued that the

deteriorating course of the patients suggests that their symptoms are

part of a schizophrenic process. In addition, when they divided the

subjects into 3 groups according to course, they found that affective

symptoms had no predictive value for either the course of the illness or

prognosis. Using the same group of probands, Welner, Welner and Fishman

(1979) examined the family histories of psychopathology. In first

degree relatives with psychopathology, more than 5/6 (27 of 30) had a

chronic, deteriorating course with mixed schizophrenic and affective

symptoms. The ultimate deterioration in function in patients and their

relatives was used to argue for schizoaffective disorder being a form of
schizophrenia in both of these studies.

Apparently, since the coining of the term, the concept of

schizoaffective has slowly come to be seen as more closely related to

affective disorder than to schizophrenia, primarily because of recent

studies of treatment response and family history. For example, Pope and

Lapinski (1978) argued that many patients with schizophrenic symptoms

and good prognosis and/or affective symptoms are actually manic-

depressive. Stone (1980) attributed this change in conceptualization to

the advent of lithium therapy. The finding that schizoaffective
patients will respond to lithium resulted in the consideration of this

disorder as a dilute or atypical form of manic-depression. In addition,

Stone (1980) concluded that no matter what definition of schizoaffective

disorder is used, there is an increased occurence of affective

disorders, rather than schizophrenia, in relatives. A number of

different research findings have also supported this notion. A study by

Clayton, Rodin and Winokur (1968) found a strong family history of

affective disorder in 39 schizoaffective patients. Pure affective

disorder was found in 9 first degree relatives and 54% had some family

history of affective disorder. In addition, 85% of the patients no

longer had schizophrenic symptoms on a 1 to 2 year followup, which is

more consistent with an affective disorder course. They concluded that

schizoaffective disorder may be an inherited variant of affective


In a large record review of 420 pairs of veteran twins, Cohen,

Allen, Pollin and Hrubec (1972) examined concordance of schizoaffective,

schizophrenic and affective disorder. They found a MZ concordance rate

for schizoaffective disorder more than twice as high as for

schizophrenia, but not significantly different than for manic-

depression. Concordance rates were 50%, 38.5% and 23.5% for

schizoaffective, manic-depression and schizophrenia, respectively.

Charts were also rated for affective and schizophrenic symptoms. The

authors hesitated to use the similar concordance rate alone to conclude

that schizoaffective disorder is a form of manic-depression. They found

nearly an identical mean number of affective symptoms for

schizoaffective and manic-depressive twins and a slightly higher rate of

suicide for the schizoaffectives. Together with the concordance data,

they concluded that the results supported Clayton et al.'s (1968)

finding that schizoaffective disorder is more closely related to manic-

depression than to schizophrenia.

There are a number of problems with this study, however. First,

the criteria to define schizophrenia and mania were not operationally

defined. While they listed the symptoms used in diagnosis, no minimum

number of symptoms was required. Winokur's criteria for primary

affective disorder were used to diagnose depressive episodes. The major

diagnostic problem was the criterion used to diagnose schizoaffective

disorder. This label was applied to patients who could not be

classified as definitely having schizophrenia or affective disorder.

Additional criteria were recurrent psychotic affective reactions along

with symptoms or episodes characteristic of schizophrenia. Chronicity

and deterioration, including chronic alcoholism, were called

"nonessential criteria" providing supporting evidence for a

schizoaffective disorder. Another major problem was the logic used to

interpret the results. While a nonsignificant difference in concordance

between schizoaffective and manic-depressive disorders was used to argue

that they are related, the difference between schizophrenia and manic-

depression was also nonsignificant, and therefore could have been used

to argue that these two disorders are also related. In addition,

accepting the null hypothesis is usually considered statistically

inappropriate. The high rate of schizoaffective concordance could have

instead been used to argue for schizoaffective being an independently

inherited disorder.

A study by Taylor and Abrams (1975) revealed no differences between

schizoaffective, manic and bipolar manic patients according to clinical

presentation, demographic features, family history and treatment

response. Further, there was no family history of schizophrenia in

their schizoaffective patients. In a second study Abrams and Taylor

(1976) compared 88 patients diagnosed either manic or schizoaffective on

clinical symptoms, demographic variables, family history and treatment

response. Patients included 78 patients who met the RDC for mania and

10 who met the RDC schizoaffective, manic subtype criteria. Treatment

response was rated on a 4 point scale by a rater blind to diagnosis.

There were no significant differences between the two groups on any

variable. The percentages of clinical variables and symptoms present

were nearly identical for both groups as was the family history data.

No patients had a family history of schizophrenia while 14% had a family

history of affective disorder and 10% a family history of alcoholism in

both groups. Both groups were also similar on demographic variables,

past and current response to treatment and mean number of episodes.

Therefore, the manic-depressive and schizoaffective, manic patients were


Dempsey, Tsuang, Struss and Dvoredsky-Wortsman (1975) presented 5

case histories of schizoaffective disorder in which treatments

traditionally used for affective disorder, ECT and/or lithium, proved

more effective than antipsychotics. In each case the patient presented
with an acute, early onset of schizophrenic and manic symptoms, an

episodic course, and a family history of depression and alcoholism.

Reiser and Willett (1976) found that both members of a father-son

schizoaffective pair responded to lithium. Similarly, Pope (1976)

found a significant response to lithium in a patient who had been

unresponsive to phenothiazine therapy. The patient was initially

diagnosed schizophrenia but later developed manic symptoms accompanied

by delusions and hallucinations.

While Procci (1976) viewed schizoaffective as a heterogeneous group

of psychoses, he concluded that evidence from treatment response and

family history studies suggested that at least a subgroup is related to

affective disorder. Procci combined data from a number of uncontrolled

studies comparing the effectiveness of lithium in manics to

schizoaffectives. He found 87% of 149 manics and 77% of 92

schizoaffectives improved on lithium. However, he cautioned that

improvement in schizoaffectives may be limited to affective symptoms and

only excited patients.

Rosenthal, Rosenthal, Stallone, Dunner and Fieve's (1980) finding

that schizoaffective patients did not differ on clinical symptoms,

family history, drug response and lab studies (platelet MAO, RBC COMT &

ATPase) from bipolar patients provided further support for

schizoaffective being a variant of affective disorder. Rosenthal et al.

compared 46 RDC bipolars to 25 patients who met the RDC for

schizoaffective disorder based on a structured interview. While

schizoaffective patients had a significantly earlier age of onset, more

past diagnoses of schizophrenia and more psychotic symptoms, both groups

had a family history of affective disorder and no family history of

schizophrenia. Response to lithium and probability of relapse was

similar for both groups. Along with the biological marker data they

concluded that RDC schizoaffective disorder is part of the broad

spectrum of manic-depressive illness.

Baron, Gruen, Asnis and Kane (1982) found a similar rate of

affective disorder in the first degree relatives of 50 schizoaffectives

(19.5%) and 85 affective disorder patients (25.4%) which was

significantly more than for relatives of 50 schizophrenics (5.1%), all

diagnosed with the RDC. Family risk for schizophrenia was significantly

higher for the schizophrenics (7.9%) than for either the

schizoaffectives (2.0%) or affective disorder patients (.4%) suggesting

that genetically, schizoaffective disorder is related to affective


Gershon, Hamovit, Guroff, Dibble, Leckman, Sceery, Targum,

Nurnberger, Goldin and Bunney (1982) found evidence for a continuum of

affective disorders from normal to schizoaffective disorder based on

family prevalence of major affective disorder. They studied 1254

relatives of patients treated for affective disorder on an inpatient

unit and 30 normal controls admitted for medical procedures. Patients

were diagnosed using the RDC and included 11 schizoaffectives,

predominantly affective, 96 bipolar I, 34 bipolar II and 30 unipolar

depressive patients. The SADS-L (Spitzer and Endicott, 1975) was used

to interview the proband, spouse and first degree relatives when
possible. Attempts were made to directly interview more distant

relatives when psychopathology was suspected.
Unipolar depression was the most frequent illness in relatives of

all groups, suggesting a genetic overlap between these disorders.
Family incidences of major depression were 37%, 24%, 25%, 20% and 7% for

schizoaffective, bipolar I, bipolar II, unipolar depression and normal

controls, respectively. While schizophrenia was more common in

relatives of schizoaffectives than in relatives of bipolar or unipolar

patients, there was more affective disorder than schizophrenia in the

relatives of schizoaffectives. Based on their results they hypothesized

that the affective disorders represent a continuum of underlying genetic

vulnerability with schizoaffectives being most vulnerable, and a

multifactorial model of transmission. Further, schizoaffective disorder

is more similar to affective than schizophrenic disorder. They admitted

that a weakness was the small number of schizoaffective patients in the


In an earlier study using a similar methodology, Tsuang, Dempsey,

Dvoredsky and Struss (1977) also found a higher frequency of affective

disorder in 228 relatives of schizoaffective patients, 11.8%, than in

the 519 relatives of bipolar patients, 7.1%, or 1,413 relatives of
unipolar depressives, 8.7%. The differences between the groups were not

significant, however. They described four other similar studies that

failed to find a higher rate of affective disorder in relatives of

schizoaffectives than bipolar patients including one by Angst (1980) and

another by Angst, Frey, Lohmeyer and Zerbin-Rudin (1980) that will be

discussed later.

After a major review of a number of well-controlled treatment

response studies comparing schizoaffective with schizophrenic and

affective disorder patients, Goodnick and Meltzer (1984) concluded that

schizoaffective disorder is similar to affective disorder. They

reviewed 15 studies comparing the efficacy of lithium carbonate and

neuroleptics, primarily chlorpromazine, in the acute treatment of mania

and schizoaffective, manic subtype. Based on these results, they

concluded that lithium alone or with antipsychotics are equally

effective in the treatment of mania and schizoaffective manic disorder

in improving both mood and psychotic symptoms, though there is some

evidence that the schizoaffectives responded more slowly. They also

reviewed 12 studies comparing antipsychotics, antidepressants and a

combination of the two in the treatment of major depression and

schizoaffective, depressed type and concluded that both respond to all

three treatments, though the course for schizoaffective, depressed was

less predictable. In schizoaffective, depressed patients a combination

of antidepressants and antipsychotics was most beneficial and

antidepressants alone least. While lithium carbonate had only modest

efficacy for schizoaffective, depressed patients, ECT seemed to produce

a similar response as in major depressives. Finally, they reviewed 10

studies examining the use of lithium as a prophylaxis in preventing

recurrences of schizoaffective episodes. They found that lithium

reduced the frequency and duration of relapse in both subtypes and

prevented relapse in up to 90%. Based on this extensive review of the

treatment literature, they concluded that schizoaffectives have a

similar pharmacological response to that of affective disorder patients.

Often, good prognosis schizophrenia has been equated with

schizoaffective disorder since both exhibit good premorbid adjustment,

rapid onset at an early age, affective symptoms during psychosis and

good outcome. McCabe, Fowler, Cadoret and Winokur (1971) found a higher

incidence of affective disorder in relatives of good prognosis

schizophrenics and a higher incidence of schizophrenia in families of

poor prognosis schizophrenics. The criteria for good prognosis

schizophrenia used resembles those for schizoaffective disorder.

Winokur (1974) equated acute and good prognostic schizophrenia with

schizoaffective disorder. Schizophrenic patients with affective

symptoms, euphoria or dysphoria, agitation, confusion and acute onset

had better prognosis (61% versus 6% for other schizophrenics) one to two

years later and had a family history of depression or mania (Winokur,

1974). He used this evidence to conclude that schizoaffective is a

variant of affective disorder. Garver, Erikson, Casper, Pandley and

Davis (1977) found a much higher risk of affective disorder in first

degree relatives of good prognosis schizophrenics (10%) compared to poor

prognosis schizophrenics (1.5%) but only a 3.3% incidence of familial

schizophrenia. In Japan, Mitsuda (1967) found that good prognosis

schizophrenics had an increased family incidence for manic-depression,

not schizophrenia. Earlier, in an amazing 50 year followup study,

Vaillant (1963) found that schizophrenics with a remitting course had a

higher incidence of manic-depression, not schizophrenia, in relatives.

Therefore, even when diagnosed schizophrenics with affective symptoms

are examined, the evidence suggests that the disorder may be genetically

linked to manic-depression rather than schizophrenia.

Harrow and Grossman (1984), however, criticized equating good

prognosis schizophrenia with schizoaffective disorder because

characteristics associated with good prognostic schizophrenia have not

been shown to be reliably predictive and because prognosis can only be

known after long term followup. They suggested using criteria that

define schizoaffective as a mixture of schizophrenia and affective


The RDC (1975) implicitly defined schizoaffective disorder as

closer to affective disorder than schizophrenia by requiring a full

depressive or manic syndrome but only one of five possible schizophrenic

symptoms. In addition, the duration criterion was consistent with

affective disorder, not schizophrenia, which required 2 weeks of


Using the RDC criteria, Rosenthal et. al. (1980) found that 25 of 71

RDC bipolar patients also met the criteria for schizoaffective disorder,

suggesting it may be a subtype of affective disorder. Earlier, Abrams

and Taylor (1976) found that 10 of 88 patients who had met criteria for

RDC mania met criteria for RDC schizoaffective disorder. Quite in

contrast to the initial draft, the final version of the DSM-III

described schizoaffective disorder as a depressive or manic syndrome

with psychotic symptoms incompatible with affective disorder. In

addition, the criteria of the second draft of the DSM-III required first

meeting the criteria for affective disorder and then having only one of

9 possible schizophrenic symptoms. Therefore, the three stages in the

development of the DSM-III showed the changing trend from seeing

schizoaffective disorder as a schizophrenia to viewing it as more like

an affective disorder.

Distinct disorder or Continuum

The possibility that schizoaffective is a distinct diagnostic

entity was first recognized in 1933 by Kasanin. Prior to this such

patients were either diagnosed as schizophrenic or affective disorder

and assumed to be atypical, or remained undiagnosed. The relatively

late identification of this syndrome compared to other psychiatric

disorders suggests that it is either relatively rare or not very

obvious. Brockington et al. (1978) reported a rate of schizoaffective

disorder of only 3-4% of all inpatient admissions using their criteria.

Indeed, using an early draft of the DSM-III, only 4.5% of a group of

inpatients met criteria for schizoaffective disorder compared to 12.9%

for schizophrenia and 29.6% for the major affective disorders. However,

a recent review of inpatient and outpatient prevalence concluded that

schizoaffective disorder ranges from 6 to as high as 20% (Singh, 1984)

and Clower (1984) wrote that the rate is much higher than previously

believed. In an introduction to a special journal issue devoted to the

study of schizoaffective disorder, Meltzer (1984) wrote that samples of

psychotic patients often consist of between 10% and 30% schizoaffective


Whether schizoaffective disorder is actually a distinct diagnosis,

rather than a variant of or combination of schizophrenia and affective

disorder, has yet to be decided. Studies of symptomatology, family

history and treatment response have been used to support the idea that

schizoaffective disorder is a viable, distinct diagnostic category.

Endicott and Spitzer (1979) found evidence that it is a separate

disorder from major depression with psychotic features. RDC major

depressives with psychotic features had significantly more endogenous

symptoms while schizoaffective patients had a more rapid onset of


More recently, Grossman, Harrow, Fordala and Meltzer (1984)

conducted a 1 year followup study on 39 schizoaffectives, 47

schizophrenics, 33 manics and 48 major depressives, in order to

determine whether the outcome of schizoaffective disorder is more

similar to schizophrenia or affective disorder. Subjects were patients

discharged from an inpatient unit using the PSE or SADS. They found

that the outcome for schizoaffectives was significantly worse than

patients with affective disorder and nonsignificantly better than

schizophrenics on 2 scales of adjustment. Only 10% of the

schizoaffectives had a positive outcome. While occupational functioning

was significantly better for both affective disorder and schizoaffective

than for schizophrenic patients, social functioning was significantly

worse for both schizophrenic and schizoaffective than for affective

disorder patients. They concluded the results disprove the notion that

schizoaffectives have a similar favorable outcome to affective

disorders. While schizoaffectives show an outcome between affective and

schizophrenic patients, they have a different pattern with good

functioning in some areas and poor functioning in others, suggesting

that it may represent an independent disorder. They found no difference

in outcome between manic and depressive, mostly affective and mostly

schizophrenic schizoaffective subtypes.

Harrow and Grossman (1984) reviewed 6 major prospective and 6 major

retrospective studies of outcome in schizoaffectives. While the results

of the studies were mixed, overall, they suggested that the outcome in
schizoaffective disorder is worse than in affective disorder but better

than in schizophrenia. Of the 6 retrospective studies reviewed,

schizoaffective outcome was found to be equal to that of schizophrenia

in one study and to be equal to affective disorder in another; the

remaining 4 studies had equivocal results. Five of the 6 prospective

studies, however, found poorer outcome in schizoaffective than in

affective disorder patients, and one found similar outcomes. While

similar data has been used to conclude that schizoaffective is a

separate disorder from schizophrenia or the affective disorders, Harrow

and Grossman did not reach any conclusions regarding its classification.

Instead, they focused on deficiencies in the research designs and

criteria used to define schizoaffective disorder. They concluded that

prospective studies yield the most useful and valid data. They also

criticized the use of good prognosis schizophrenics in research on

schizoaffective disorder since prognosis can only be known after long-

term followup and suggested the use of specific criteria such as the RDC

that define schizoaffective as a mixture of affective and schizophrenic


While the Vogl and Zaudig (1985) study focused on evaluating

different definitions of schizoaffective disorder by examining

diagnostic concordance, they did demonsL.-'ate that schizoaffective

disorder is distinct from the cycloid psychoses which have often been

considered equivalent disorders. They advocated a more stringent

definition of schizoaffective disorder to obtain a more homogeneous

population and felt it could be differentiated from schizophrenia and

the major affective disorders.

A number of studies of family psychopathology have provided

evidence that schizoaffective is a distinct disorder. The Cohen, Allen,

Pollin and Hrubec (1972) study already discussed found evidence that

schizoaffective disorder is a distinct entity using data from a large

scale diagnostic concordance study of monozygotic twins. They found a

50% concordance for schizoaffective disorder in MZ twins compared to

only 39% and 24% concordant diagnoses for manic-depressive and

schizophrenic twins. While they concluded their study suggested that
schizoaffective disorder was related to affective disorder, in his

review of research on schizoaffective, Procci (1976) noted that the

Cohen study and other family history and genetic data suggest that it is

a distinct category at least from schizophrenia. Earlier, Mitsuda

(1967) also concluded that it is a distinct disorder from schizophrenia

and probably from affective disorder after reviewing evidence from
genetic studies.

Angst, Frey, Lohmeyer, and Zerbin-Rudin (1980) and Angst (1981)

also examined family history of psychopathology in relatives of

schizoaffective, bipolar and unipolar affective disorder patients over a

15 year period. Like Gershon et al. (1982), they found a higher rate of

affective disorder in relatives of schizoaffectives than in relatives of

unipolar, but not bipolar depressives. They also found more incidences

of schizophrenia in relatives of schizoaffectives than in relatives of

bipolars, but schizoaffectives still had a higher family rate of

affective than schizophrenic disorder. Angst used this data to conclude

that schizoaffective disorder is a single, distinct disorder but with

genetic relationships to both schizophrenia and affective disorder.

Frequently in this literature, similar findings are used to draw very

different conclusions.

Studies of treatment response have also been used to argue that

schizoaffective is a distinct disorder. Goode and Manning (1983)

examined treatment response in 20 RDC schizoaffective, depressed

patients treated with bupropion (an antidepressant) alone or with the

addition of haldol. Improvement for affective symptoms was measured on

a depression rating scale and global ratings, while other symptoms were

rated on the Brief Psychiatric Rating Scale (BPRS) (Overall and Gorham,

1962). One-third of the patients in the bupropion only group showed

increased psychotic symptoms, while no specific symptoms or features

seemed associated with this increased risk. Schizophrenics administered
bupropion have been shown to have an increased risk of exacerbation of

psychosis. In the combined treatment group, half were rated improved

and the remainder moderately improved for both schizophrenic and

affective symptoms on the BPRS. In a letter to the editor defending his

research, Goode (1984) claimed that the results show that

schizoaffective disorder is distinct from major depression with

psychotic features. It has been shown that major depressives with

psychotic features rarely show an increased level of psychosis and

respond well to this drug. Further, he noted that schizoaffectives,

like schizophrenics, show a lowered Hoffman reflex recovery curve,

suggestive of increased dopaminergic metabolism while major depressives

with psychotic features do not. Therefore, he concluded that

schizoaffective, depressed patients are clinically and physiologically

distinct from psychotic depressives. However, one could use his data to

conclude that schizoaffective, depressed subtype patients represent a

schizophrenic disorder.

In another letter to the editor commenting on the Goode and Manning

study, Clower (1984) wrote that a combination of antidepressants and
antipsychotics is best suited for treating schizoaffective disorder.

When schizoaffectives are diagnosed as schizophrenic they are treated

with only antipsychotics, yet when they are diagnosed schizoaffective,

they are often only treated with lithium. Both approaches are

inadequate. He also criticized the lack of differentiation between

major depression with psychotic features and schizoaffective, depressed

disorder in the diagnostic criteria and also stated that the presence of

schizoaffective disorder is much higher than is often recognized.

Schizoaffective may instead represent a borderline condition in

which a patient has partial characteristics of two syndromes. In fact,

the lack of criteria for schizoaffective disorder in the current

official nosology suggests that its validity as a distinct diagnosis is

questionable. Welner et al. (1974) distinguished between viewing

schizoaffective as a borderline condition between schizophrenia and

affective disorder versus a combination of both disorders. While Beck

(1967) hypothesized that there is a spectrum of disorders from manic-

depression to schizophrenia with schizoaffective between, Vaillant

(1963) and more recently, Stone (1980) considered it a mixture of both


Kendell (1968) has proposed that depressive subtypes may fall along

a continuum, rather than being distinct diagnostic entities. In order

to demonstrate that there is more than one population of depressives, he

wrote, one must demonstrate a bimodal distribution or discontinuity

between populations. Boundaries of disorders need to be placed at

points of rarity so that there are more members in either population

than in between. Applied to schizoaffective disorder, one would need to

show that the population from affective disorder to schizoaffective to

schizophrenia is discontinuous (i.e., forms 3 separate populations with

few cases between). Instead, depression and schizophrenia may be two

populations that somewhat overlap, resulting in a small population of

patients with varying amounts of symptoms from both syndromes.
Depending where the patient fell along the continuum, schizophrenic or

affective symptoms would dominate the picture. The overlapping region

would contain both schizophrenics and affective disorder patients that

are labeled schizoaffective.

Kendell (1975) argued that it is because syndromes merge into each

other without natural boundaries and few patients actually fit clearly

into a category that borderline concepts such as schizoaffective

disorder were created. Kendell and Gourlay (1970) performed a

discriminant function analysis of extensive questionnaire data on a

large group of patients. Based on their findings they argued for

symptom overlap between schizophrenia and affective disorder with

heterogeneous inheritance of schizoaffective disorder.

More recently, Stone (1980) proposed a continuum model of

schizoaffective disorder. In general, Stone advocated utilizing

dimensional classification rather than typological, which results in the

loss of valuable information. His Sz-MDP phenotypic classification had

a continuum from undisputed schizophrenia to undisputed affective

disorder with a host of mixed conditions between. The continuum was

divided into five regions according to a 20% increase in schizophrenic

or affective symptoms. The mid-region contained patients with 50%

schizophrenia and 50% affective symptoms. Since Stone conceptualized

schizoaffective disorder as a mixture between the two types of symptoms,

he considered only patients in this region to be schizoaffective. A

vertical dimension was used to represent severity of pathology. Stone
modified a weighted rating scale of Cohen et al. (1972) to allow

calculation of the numerical ratio of schizophrenic to affective

symptoms and placement of patients in one of the 5 horizontal cells. He

pointed to the need to conduct family history studies to examine

genotypes related to the phenotypic classification. He hypothesized

that schizoaffective disorder is the result of a vulnerability from a

genetic loading for both schizophrenia and manic-depression with the

interaction of additional independent factors.

Mendlewicz, Linkowski and Wilmotte (1980) found genetic evidence

that schizoaffective disorder represents an overlap of family risk for
schizophrenia and affective disorder. They compared the family history

of schizoaffectives with schizophrenics, bipolar and unipolar affective

disorder patients (N = 55 each) matched for age and sex. Schizophrenia

and schizoaffective disorder was diagnosed using the Feighner criteria

while affective disorder patients were selected according to Winokur's

notions of affective disorder. They found a similar risk of affective

disorder in relatives of schizoaffectives (34.6%), bipolar (39.4%) and

unipolar (28.5%) affective disorders, which was significantly higher
than in relatives of schizophrenics (8.6%). In contrast to Baron et al.

(1982), they found a family risk of schizophrenia in first degree

relatives of schizoaffectives (10.8%) that was similar to but lower than

for schizophrenics (16.9%) and significantly higher than for bipolar

(1.8%) and unipolar (3.2%) affective patients.

Singh (1984) examined the efficacy of flupenthixol decanoate in 20

inpatients diagnosed as schizoaffective according to the RDC. Patients

were rated on the BPRS before and at 2 week intervals during the study.

Flupenthixol, a thioxanthene derivative with both antipsychotic and

antidepressant properties, was selected because of the hypothesis that

schizoaffective disorder represents a mixed schizophrenic and affective

condition. Of the 20 patients, 15 were significantly improved, 2

unchanged and 3 worsened. These three patients became excited and

unmanageable and were dropped from the study. Based on the results,

Singh concluded combined antipsychotics and antidepressants are most

appropriate for this intermediate disorder. Note that the Goode and

Manning (1983) results could have been used to argue this position as

Meltzer, Arora and Metz (1984) reviewed nearly 200 biological

studies comparing schizoaffective disorder with schizophrenia and

affective disorder. The studies found evidence of similar biological

abnormalities in schizoaffective and affective disorder patients not

found in schizophrenics, including decreased platelet serotonin (5-HT)

uptake, reduced clonodine induced increase in serum growth hormone,

decreased latency of REM sleep and increased REM density. However,

several studies found abnormalities in affective psychosis patients

(e.g., increased plasma norepinephrine NE) that were not found in

schizoaffectives. In addition, a number of studies have found similar

biological abnormalities in schizophrenics and schizoaffectives not

found in affective disorder patients including increased cerebrospinal

fluid (CSF) NE, increased platelet 5-HT, and reduced prostaglandin El-

stimulated adenylate cyclase activity. Further, similar biological

defects, dysfunction in eye tracking, increased CSF concentration of

gabba aminobutyric acid and neuromuscular abnormalities, have been found

in all three types of disorders. While the results were not entirely

consistent across all studies, they do suggest that a simple

reductionist model of schizoaffective disorder as a subtype of affective
disorder cannot be supported. One possibility is that there are cases

of schizoaffective disorder related to affective disorder and others

related to schizophrenia. However, the authors concluded that the data

are consistent with a continuum model of the psychoses with polygenetic

inheritance and a number of genetic factors common to all psychoses.

Winokur (1984) wrote that both schizoaffective disorder and

psychotic affective disorder have a remitting course, a mixture of

affective and psychotic symptoms and neither responds to

antidepressants. In order to show that schizoaffective is a viable

diagnosis, one would need to demonstrate a different family history

between schizoaffective and psychotic affective disorder patients.

Winokur used 122 bipolar and 203 unipolar affective disorder patients

who were part of the large Iowa 500 series of studies and were diagnosed

with the Feighner criteria. He divided each group according to the

presence or absence of psychotic symptoms and then examined the

morbidity risk of affective disorder in relatives. While no

schizoaffectives were included, he claimed that many of the psychotic

affective disorder patients would qualify. He found no difference in

family history of affective disorder between psychotic and nonpsychotic,

unipolar or bipolar affective patients and concluded there is no reason

to separate out schizoaffective disorder based on family history. He

also used the data to dispute Gershon's (1982) model of a continuum of
affective disorders. Instead, he hypothesized that the propensity

towards affective disorder and psychosis are transmitted independently.

Therefore, Winokur, like Welner et al. (1974) distinguished between

viewing schizoaffective disorder as a borderline condition between

schizophrenia and affective disorder and his own view that it is a

combination of inheriting the propensity for both disorders.

Braden (1984) also hypothesized a two factor vulnerability model of

schizoaffective disorder. Because of the remitting course of the

affective disorders, schizoaffective, and to a lesser extent,

schizophrenia, he proposed that a vulnerability model is most

appropriate. In his model, the first factor was a vulnerability to

psychotic illness characterized by psychomotor and vegetative agitation.
After reviewing studies on the role of life events and response to

neuroleptics in acute schizophrenia, he hypothesized that there was a

relationship between stress and increased activation that may be

mediated by dopamine. Stress itself was nonspecific and resulted in

psychosis only if there was a pre-existing vulnerability. Those with a

low vulnerability would require an unusually strong psychosocial or

biological stressor to result in the activated, psychotic state.

Specific challengers likely to induce psychosis in the vulnerable

included drugs (e.g., hallucinogens, amphetamines), childbirth, severe

weight loss and environmental events. Because dopamine receptor

blockers prevent the effects of stress, dopaminergic systems may mediate

the aspects of response to stress that result in an acute psychotic

episode. While he hypothesized that inheritance of bipolar affective

disorder was the factor most often associated with increased

vulnerability to episodes of increased activation, he cautioned that the

activated state was etiologically and diagnostically nonspecific.

The second factor postulated by Braden was vulnerability to

cognitive disturbance in response to the increased activation from the

first factor. Therefore, while a similar process of activated state

occurs in schizoaffective, schizophreniform and affective disorders, in

schizoaffective and schizophreniform, a second process was triggered by
the first resulting in cognitive symptoms. This model allowed for

similar prognosis in schizoaffective and affective disorders and similar

family histories since it assumed that factor 1 and 2 were independently

inherited. Treatment response may also be similar since the second

process, the development of cognitive symptoms, was dependent upon the

first. Braden cited evidence from findings of similar family histories

and treatment response in schizoaffective and affective disorder

patients. He did not make any conclusions about the most appropriate

classification of schizoaffective disorder but did suggest several

research designs comparing specific clinical features in subgroups of

patients differing in level of activation, and cognitive symptoms.

After reviewing 5 genetic studies (Tsuang, 1979; Mendlewicz et al.,

1980; Abrams and Taylor, 1980; Baron et al., 1982; Gershon et al.,

1982) on schizoaffective disorder that met stringent methodological

requirements, Abrams (1984) also hypothesized a multifactorial,

multithreshold model of inheritance for schizoaffective disorder. There

was no evidence for genetic transmission of schizoaffective disorder as

an independent entity (Tsuang, 1979; Baron et al., 1982). For example,

Baron et al. (1982) found a low, similar rate of schizoaffective

disorder in first degree relatives of schizoaffectives, schizophrenics

and affective disorder patients (2.3%, 2.3%, 3.0%, respectively).

Instead, the results of the review suggested a genetic overlap between

schizophrenia and affective disorder with the larger contribution coming

from affective disorder. Abrams concluded that the presence of

schizophrenic symptoms reflects a severity dimension with

schizoaffective disorder being most severe.

Meltzer (1984) also hypothesized that overlapping biological

vulnerability for psychosis and affective disorder interacting with

environmental stress results in schizoaffective disorder and a spectrum

of psychoses. He suggested the use of a dimensional rather than

categorical classification to investigate these disorders.

Heterogeneous Category

Kost recently, a number of researchers have come to the conclusion

that all of the above models are too simplistic and that many of the

results are stretched in order to fit one of the models. The numerous

contradictory findings can be explained by postulating that

schizoaffective disorder is a heterogeneous group composed of one

subgroup related to affective disorder and another related to

schizophrenia. The results found in past studies depend upon which

definition of schizoaffective disorder was used and which subtype was

studied; mixed results simply reflected a mixture of subtypes.

Kendell (1968) hypothesized that schizophrenia and affective

disorder were two separate but overlapping populations and that patients

in the overlapping region were diagnosed schizoaffective. According to

this model, schizoaffective disorder is a heterogeneous category

consisting of a subgroup of schizophrenics and a subgroup of affective

disorder patients. Kendell and Gourlay (1970) found overlap between

schizophrenic and affective disorder symptoms and concluded

heterogeneous inheritance of schizoaffective disorder.
The first study by Welner and associates (Croughan et al., 1974)

found two distinct subgroups, one resembling paranoid schizophrenia and

another, which they hesitated to categorize as schizophrenic, affective

or a distinct disorder. This second subgroup consisted of patients with

an affective-like presentation. However, because of the sampling, it is

not clear if these were really two subgroups of schizoaffective disorder
or if the first group included the acute schizophrenics and


After reviewing the literature on schizoaffective disorder, Procci

(1976) concluded that schizoaffective disorder is a heterogeneous

category with a subgroup related to schizophrenia and another related to

manic-depression, based on response to medications, either

antipsychotics or lithium.

In a study of diagnostic concordance in pairs of siblings, Tsuang

(1979) found a significantly lower number of schizoaffective-

schizoaffective pairs (4) than expected (10.5) but an expected number of

schizophrenic-schizophrenic (5) and affective-affective pairs (11).

Diagnoses were based on case records of 71 pairs of siblings who had

both been hospitalized for psychiatric disorder. Patients were

diagnosed according to the Feighner criteria; for schizoaffective

disorder, a combination of the schizophrenic and affective disorder

criteria was used. Of the 70 patients, there were 17 diagnoses of

schizophrenia, 21 of schizoaffective and 32 of affective disorder.

There were 6.5 less pairs concordant for schizoaffective disorder than

expected suggesting that there was evidence for schizophrenia and

affective disorder, but not schizoaffective disorder, being genetically

distinct. The number of schizophrenic-schizoaffective and affective

disorder-schizoaffective pairs was not significantly higher than

expected, as would have been the case if schizoaffective disorder was

genetically linked to either. Tsuang concluded that schizoaffective

disorder is a heterogeneous category consisting of a subgroup related to

schizophrenia, another related to affective disorder and an

undifferentiated subtype. However, the method used was a record review

rather than clinical interview and the decreased number of

schizoaffective pairs may reflect the simultaneous application of the

Feighner criteria for schizophrenia and affective disorder to diagnose

schizoaffective disorder which Vogl and Zaudig (1985) demonstrated

resulted in too restrictive a definition.

After reviewing the history of the concept of schizoaffective

disorder, Tsuang and Simpson (1984) reviewed the evidence from followup

and family history studies of schizoaffective disorder. Based on their

review, they concluded that schizoaffective is a distinct, but

heterogeneous disorder with a subgroup similar to schizophrenia and

another similar to affective disorder. They also described their

current research program of long-term followup (30 to 40 years) and

family history studies of schizoaffective probands and first degree

relatives using the 85 subjects from the earlier Iowa 500 series of

studies. They recommended that the most fruitful way to study

schizoaffective disorder is to attempt to locate homogeneous subgroups

and then analyze their characteristics in order to refine diagnostic

criteria and improve differential diagnosis. They plan to compare

followup and family history data of schizoaffectives with "typical"

psychotics with affective disorder and schizophrenia. One strategy will

be to discover which variables discriminate typical schizophrenics and

affective disorders and then use these variables to determine if they

can help divide schizoaffective patients into homogeneous subgroups.

The RDC divided schizoaffective disorder into depressive versus

manic types and then further classified the disorder as mainly

schizophrenic versus mainly affective. The division into manic and

depressive subtypes has been used most frequently, perhaps because the

difference in clinical symptoms makes the distinction obvious and
logical. In his book on depression, Singh (1980) differentiated

schizoaffective manic from depressed subtypes based on clinical

characteristics. Brockington and Kendell's criteria also distinguished

between so-called "schizomanic" and "schizodepressive" patients. The

majority of the results suggest that schizoaffective, manic may be

related to affective disorder while schizoaffective, depressed is more

related to schizophrenia.

Brockington et al. (1978) conducted two month long drug trials of

schizoaffectives divided into manic and depressed subtypes.

"Schizomanic" (N = 19) patients were treated with either chlorpromazine

or lithium while 41 "schizodepressives" were treated with either

amitriptyline, chlorpromazine or a combination of both. Patients were

diagnosed used the Brockington and Kendell criteria after a clinical

interview using the PSE. "Schizomanics" were administered either 100

mg. chlorpromazine or 250 mg. lithium carbonate daily, while
"schizodepressives" were given 25 mg. amitriptyline, 75 mg.

chlorpromazine or both. Patients were rated on the PSE and BPRS at the

beginning and end of the study. "Schizomanic" patients responded well
to both lithium and chlorpromazine with 2 patients in each group showing

complete recovery. In the lithium group only one patient showed an

increase of symptoms, which were depressive. Surprisingly, lithium

seemed superior to chlorpromazine in treating schizophrenic symptoms

(e.g., auditory hallucinations, delusions of persecution, catatonia)

except for delusions of reference, which improved more with

chlorpromazine, as did total number of symptoms. The response rate to

lithium (59%) was similar to the improvement seen in manic-depression.

In contrast, the "schizodepressives" responded poorly to all treatment

with only 20% showing recovery. This rate is less than the rate for

either schizophrenics treated with antipsychotics or depressives treated

with antidepressants. Only one recovered in the amitriptyline group

(N = 13), 4 in the chlorpromazine group (N = 11) and 2 in the combined

treatment group (N = 12). Seven in the antidepressant group, 5 in the

antipsychotic group and 3 in the combined treatment group showed no

change. Patients in the antipsychotic and combined treatment groups,

which were nearly equivalent, did best. Based on their results they

concluded that "schizodepressives" are basically schizophrenic while
"schizomania" is a variant of manic-depressive disorder.

Clayton (1982) also suggested dividing schizoaffective disorder

into depressive and manic subtypes. In a review of the research on

schizoaffective disorder, she concluded that data on symptomatology,
course, family history and treatment response suggested that

schizoaffective, manic is probably related to bipolar affective

disorder, but that the data on schizoaffective, depressed is less clear.

For example, the schizoaffectives in the Rosenthal et al. (1980) study,

did not differ on clinical, family history, lab data or treatment

response from manic patients. Abrams and Taylor (1976) found no

difference between their schizoaffectives, who were also manic subtype,

and manics on clinical, demographic, family history or treatment

variables. The 5 cases presented by Dempsey et al. (1975) who all

displayed manic symptomatology also responded to lithium. The course of

schizoaffective, manic has also been more clearly described (Abrams and

Taylor, 1976; Brockington et al., 1978; Rosenthal et al., 1980) and

seems to be similar to bipolar disorder. The course of schizoaffective,

depressed disorder has been less well studied, but Brockington et al.

(1978) found a poorer outcome similar to schizophrenia. More recently,
the Goode and Manning (1983) study found that schizoaffective, depressed

patients had a similar response to an antidepressant as schizophrenics.

Coryell, Lavori, Endicott, Keller and Van Eerdewegh (1984)

conducted a 6 month followup of 24 patients with schizoaffective,

depressed type, 56 psychotic depressives and 274 nonpsychotic major

depressives. They examined outcome at 12, 18 and 24 months and family

psychopathology in a portion of the groups. They found a group of

schizoaffective, depressed patients with a long term course like

schizophrenia. While the presence of delusions in depressives predicts

poor response to some antidepressants and poorer short term outcome,
there was no difference in long term outcome and family history between

delusional and nondelusional depressives. On all measures of outcome,

the schizoaffective, depressed were significantly worse than the

psychotic depressives. Other studies have shown higher rates of

schizophrenia in relatives of schizoaffectives and depressives with

mood-incongruent psychotic features than in other major depressives.

Together with the results from this study they concluded there is a

subgroup of schizoaffective, depressed type patients that have a

schizophrenic disorder.

The Goodnick and Meltzer (1984) review of treatment studies

discussed earlier also provides evidence for the heterogeneity of

schizoaffective disorder, particularly the differentiation of depression

and manic subtypes. While the studies they reviewed showed that lithium

was only moderately effective for treating acute schizoaffective

depressive episodes, it was highly effective for acute schizoaffective

manic episodes, suggesting this subtype resembled bipolar disorder in

treatment response. In contrast, the acute schizoaffective depressed

episode was best treated with a combination of neuroleptics and

antidepressants; antipsychotics alone were next most effective followed

by antidepressants alone. Schizoaffective, depressed were also similar

to major depressives in their response to ECT. In general the course of

the depressed subtype tended to vary and be unpredictable. Therefore,

while Goodnick and Meltzer concluded that schizoaffective is a form of

affective disorder, the results also suggested that there are at least

two subtypes, one that resembles bipolar disorder and another that

resembles major depression based on differences in treatment response

and course. The authors also hypothesized that schizoaffectives who are

nonresponders may represent a homogeneous subgroup of psychoses,

suggesting the possibility of a third subtype. While their conclusion

that the depressed subtype is similar to major depression differs from

the conclusions of others, the increased response to neuroleptics in

this group could be used to argue that they resemble schizophrenics.

The RDC also made a distinction between mainly affective and

mainly schizophrenic schizoaffective disorder. The Baron et al. (1982)

study already discussed also attempted to find evidence for the validity

of the RODC divisions into mainly schizophrenic and mainly affective

subtypes of schizoaffective disorder by examining differences in. family

risk of psychopathology in first degree relatives. They compared the
family morbidity of 50 schizoaffectives divided into subtypes with the

family risk of matched groups of 50 schizophrenics, 45 bipolar and 45

unipolar affective disorder patients all diagnosed with the RDC.

Personal interviews were conducted with 72% of first degree relatives.
They found the schizoaffective mainly schizophrenic subtype to have a

family risk similar to the schizophrenics while the mainly affective

subtype had family histories similar to affective disorder patients.

When the mainly affective subgroup was further divided into manic and

depressed subtypes they found a nonsignificant trend for relatives of

the manic subtype to have an increased risk of either bipolar or

unipolar affective disorder compared to the depressed subtype,

suggesting that this division may also be meaningful.

Mattes and Nayak (1984) compared the prophylactic effect of lithium

to fluphenazine in mainly schizophrenic schizoaffectives. They

hypothesized that studies showing lithium to be prophylactic for

schizoaffective disorder (Abrams and Taylor, 1976; Rosenthal et al.,

1980) used schizoaffective patients with primarily affective

symptomatology. Studies using schizoaffectives with more schizophrenic

symptoms, especially thought disorder, found lithium to be less

effective. They compared the effectiveness of the neuroleptic,

fluphenazine with lithium carbonate in preventing relapse in 14

schizoaffective, depressed patients diagnosed with the RDC. Patients

were further subclassified as chronic (5 patients), subchronic (8

patients) and subacute (1 patient) because of the suggestion that

lithium is more effective with acute schizoaffective disorder (Pope and

Lapinski, 1978). They found that lithium was not an adequate

prophylaxis for this group; 6 relapsed on lithium, 3 displaying

schizophrenic and 3 schizoaffective symptoms, while only one remained

stable. Even this one patient relapsed one month after the study was

terminated. In contrast, there was only one relapse with the

neuroleptic, that took the form of a manic episode, and 4 patients

remained stable. Patients on lithium relapsed an average of 3.6 months

after starting treatment, the same time as relapse occurs for

schizophrenics on placebo. The authors concluded that the distinction

between mainly schizophrenic and mainly affective schizoaffective

disorder is meaningful and important in selecting treatment.

Maj (1984) examined the effectiveness of lithium as a prophylaxis

in 4 groups of schizoaffectives, patients who met the RDC criteria, the

Brockington and Kendell criteria, Welner's criteria and Perris' criteria

for a cycloid psychosis. His review of the literature had revealed

inconsistencies in the usefulness of lithium for treating the acute

phase and for prophylaxis. He used a polydiagnostic approach in which

different diagnostic criteria are applied to the same subjects because

of the varying definitions of this disorder and in order to help single

out which schizoaffectives respond to lithium. He used 38

schizoaffectives who resembled the ICD-9 description of schizophrenia,

schizoaffective type. When rediagnosed, 25 met the RDC, 10 manic and 15

depressed types, 30 met the Brockington and Kendell definition, 12 manic

and 18 depressed, while only 13 met the Welner criteria and 15, Perris'

cycloid psychosis criteria. All subjects were treated with lithium

carbonate, 600 to 1500 mg. per day, for two years. Maj compared the

number of morbid episodes for the two years during treatment with the

two prior years for the total group and for each subgroup. He found a

highly significant difference in number of episodes for the total group

and for all subgroups but the Welner criteria group. Within the RDC and

Brockington criteria groups, manic subtypes responded better than the

depressed subtypes.

Maj also divided the population into responders and nonresponders

(both N = 18) and then compared the two groups on clinical/history

variables (age of onset, number of episodes per year, history of pure

affective or schizophrenic disorder, family history) and biological

variables (plasma lithium, red blood cell plasma to lithium ratio,

platelet MAO activity, HLA antigens). All patients with a history of

pure affective episodes responded to lithium and there was a

nonsignificant trend for those with family histories of affective

disorder to be in the responder group and those with family histories of

schizophrenia to be in the nonresponder group. There were no

differences on the biological variables. Maj concluded that lithium

carbonate is effective only in schizoaffective patients with a full

affective syndrome. Welner's criteria tends to select schizoaffectives

with more prominent schizophrenic symptoms. Maj concluded that

schizoaffective disorder is a heterogeneous category and stressed the

role of the diagnostic criteria and whether they emphasize schizophrenic

or affective symptoms in determining the results of studies. When manic

symptoms are stressed, the patients look like bipolars on a number of

variables. Finally, he concluded that clinical/history variables,

especially course and family history, but not biological variables, are

predictive of schizoaffective response to lithium.

Other Variables

The models described above represent the major historical and

current perspectives on the classification of schizoaffective disorder

in the major functional psychoses. However, two variables may be

influential in the decision to assign a schizoaffective diagnosis, the

temporal order of the onset of schizophrenic and affective symptoms and

the patient's response to medications. While neither of these variables

are in themselves hypotheses about where schizoaffective disorder should

be classified, they may effect actual usage of the diagnosis and also

provide some support for the different models of classification.

Temporal characteristics

While schizoaffective disorder seems to share characteristics of

both schizophrenia and affective disorder to some degree, it is possible

that the temporal relationship of the onset of symptoms, rather than any

particular symptoms or combinations of symptoms, is most important in

diagnosing schizoaffective disorder. For example, both the RDC and DSM-

III emphasize temporal relationships in their definitions. In the RDC

it is the onset or the order that symptoms appear, not the number of

symptoms that predominate, which determine the classification into
"mainly schizophrenic" and "mainly affective" subtypes. The mainly

schizophrenic subtype requires the presence of the "core" schizophrenic

symptoms in the schizoaffective definition for at least a week along

with social withdrawal, reduced occupational functioning, eccentric

behavior or unusual thoughts or perceptions before the onset of

affective symptoms. The diagnosis of mainly affective subtype requires

only that affective symptoms occur prior to any schizophrenic symptoms

plus good premorbid social and occupational functioning. The DSM-III

also focuses on the temporal relationship of symptoms in diagnosing

schizoaffective disorder. This diagnosis should be made when affective

symptoms develop first or the onset of schizophrenic and affective

symptoms is concurrent, not when psychotic symptoms develop first. When

schizophrenic symptoms develop first it is suggested that the diagnosis

of schizophrenia or some other psychotic disorder (e.g., schizophreni-

form, brief reactive psychosis) with secondary depression be considered.

Earlier, Spitzer et al. (1978) suggested that temporal factors may

be used to divide the category into valid subtypes. They hypothesized

that the acute form is more related to mood disorders while the chronic

form is more closely related to schizophrenia. Endicott and Spitzer

(1979) found that rapid onset of symptoms differentiated

schizoaffectives from psychotic major depressives. Overall, these

findings suggest that temporal factors, not the number of schizophrenic

or affective symptoms, may be very important in diagnosing

schizoaffective disorder.

Treatment factors

Another variable which is potentially influential in diagnosing

schizoaffective disorder is response to treatment. None of the current

diagnostic systems utilize treatment response as diagnostic criteria.

Murray and Murphy (1978), however, argued for using treatment response
to help refine diagnostic categories. On the other hand, Kendell (1968)

criticized the use of treatment response in constructing

classifications, pointing out that both headaches and rheumatism respond

to aspirin. Murray and Murphy, however, were not advocating combining

already validly distinct categories, but rather in using treatment

response to find homogeneous subgroups of treatment responders within

categories. While Stone (1980) agreed with Kendell that drug response

should not be incorporated into the diagnostic process, he did recommend

using it for external validation of diagnoses.
As previously discussed, recent findings suggest that

schizoaffective disorder patients may be more like affective disorder

patients than schizophrenics in their response to medication. Stone

(1980) noted that it was the response of schizoaffectives to lithium

that resulted in schizoaffective disorder being viewed as an affective


Because a number of authors had reported greater effectiveness of

lithium in bipolar compared to unipolar depressives (Goodwin, Murphy,

Dunner & Bunney, 1972; Stone, 1980), others have suggested using lithium

response as a diagnostic aid (Forssman & Walinder, 1970; Stone, 1980).

Forssman and Walinder described two cases with confusing symptoms who

received several inconsistent diagnoses. Since both patients eventually

responded to a trial of lithium, the authors suggested that both were

actually variants of manic-depressive states. Further, they proposed

that positive lithium response can be used to make more accurate

diagnoses from the standpoint of etiology. In supporting the use of

treatment response in differential diagnosis, they pointed to the use of

treatment response in medicine to define subcategories of diabetes

(insulin and non-insulin dependent diabetes mellitus, ADH sensitive and

ADH resistant forms of diabetes insipidus). A continuum of lithium

effectiveness from bipolar to schizoaffective to unipolar depression has

also been proposed (Johnson, Gershon, Burdock, Floyd & Hekiman, 1971;

Goodwin, et al., 1972; Stone, 1980), with schizoaffectives receiving a

moderate amount of benefit from lithium. The Rosenthal et al. (1980)

and Pope and Lapinski (1978) studies, however, suggested that lithium

may be equally effective for schizoaffective disorder and mania.

However, these studies used only schizoaffective, manic patients and

Procci (1976) noted that lithium may only be effective for this group.

Recently, the idea that treatment response is a valid means of

making diagnoses is becoming more accepted. Lithium response has been

advocated to help diagnose schizoaffective disorder which provides some

support for the notion that it is actually a variant of affective

disorder. Because lithium is the treatment of choice for bipolar mania,

lithium response may be useful for differentiating schizoaffective

disorder from schizophrenia, but not from bipolar mania. However, it

may be less effective with depressed schizoaffectives. Response to
tricyclics or other antidepressants may allow differentiation from

psychotic major depression. Response to antipsychotics may suggest

either a schizophrenic or schizoaffective diagnosis but can be used to

distinguish it from affective disorders. While response to treatment

may not yet have the specificity that Forssman and Walinder hoped for,

it may be useful in narrowing down diagnostic possibilities. However,

it is also possible that schizoaffective disorder is a heterogeneous

category, as proposed by Procci (1976) made up of subgroups that respond

to antipsychotics or to lithium. Therefore, caution is necessary in

using treatment response to diagnose and classify schizoaffective

Summary and Comment
Based on this review of the literature, there seems to be most

empirical support for schizoaffective disorder being a type of affective

disorder or a borderline condition, either on a continuum between

affective disorder and schizophrenia or a combination of the two

disorders. There is also much evidence supporting the hypothesis that

it is actually a heterogeneous category with one subgroup related to

schizophrenia and another one related to affective disorder. This most

recent proposal also has the advantage of integrating and explaining the

many contradictory findings in the empirical studies.

Which definition of schizoaffective disorder was used had a major

effect on the findings, pointing out the importance of careful selection

of the diagnostic criteria. Early studies frequently used clinician

impressions or idiosyncratic definitions that made it difficult to

compare results and draw conclusions across studies. The methodology in

recent studies has greatly improved, particularly in the use of

operational definitions of schizoaffective disorder. There also appears

to be increased interest in studying this diagnosis recently. For

example, 25 of the citations in this review were published since 1980.

One continuing problem is in the interpretation of the results. Many of

the authors reached opposite conclusions from similar findings. One

reason was that the design of the studies frequently did not allow

testing of alternative models of schizoaffective disorder. Researchers

often started with a strong belief in a particular model and then simply

examined whether their results could be consistent with that model.

Further, many investigators appeared to be strongly devoted to a

particular view and frequently ignored more parsimonious interpretations

of their data that argued against their position.


The prototype model has been used extensively in cognitive

psychology to study natural language and the structure of natural

categories (Rosch, 1975; Rosch & Merviss, 1975), pattern recognition

(Posner, Goldsmith & Welton, 1967) and semantic memory (Lang, 1978;

Wessels, 1980). Recently, the concepts derived from this research have

been utilized to examine topics outside the realm of cognitive

psychology including psychopathology (Cantor, Smith, French and Mezzich,

1980; Horowitz, Post, French, Siegelman and Wallis, 1981).

According to the prototype view, categories are first learned by

learning exemplars or typical members of categories rather than learning

a list of features of the category. Instances are classified by

comparing them to the category exemplars so that instances that are

highly similar to an exemplar are quickly and accurately classified. It

is only after a number of members of a category are learned that the

features associated with a category are abstracted. In contrast to the

classical view, a category is represented by a group of associated

features, rather than definitive characteristics which must be present.

Members are more or less prototypic based on the number of features of

that category they possess. Category members with fewer prototypic

features or more shared with other categories are identified slower and

with less accuracy than those with more of the associated features.

Incomplete rather than perfect nesting of features in hierarchically

arranged categories is permissible.

The origin of the notion of prototypes has been attributed to

Wittgenstein, a philosopher. Wittgenstein (1953) wrote that formal

criteria or defining features are not a logical prerequisite for

categorization; instead, categorization can be understood according to

family resemblence. No real categories have attributes shared by all

members; members bear a family resemblence according to the number of

attributes of that category that they possess. Further, he suggested

studying clear cases since category judgments are problematic only when

boundaries between categories are examined. Wertheimer's (1938) earlier

description of "ideal types", perceptually salient stimuli that act as

anchoring points for perception, also resembles more recent notions of

prototypes. Looking back at ancient history, Socrates' theory of

universals is the forerunner of prototype theory. According to

Socrates, individual examples of categories are recognized by comparison

to a universal ideal or perfect member (Robinson, 1981).

The prototype model may be especially useful in the study of

schizoaffective disorder, because of the controversial nature of this

concept. Empirical studies of patients have demonstrated that

schizoaffectives look like schizophrenics, affective disorders, an

intermediate group or a heterogeneous category depending upon the

definition of schizoaffective that is selected. For example, using the

Welner et al. (1974) criteria, that emphasize the psychotic features, to

define schizoaffective disorder will result in the finding that

schizoaffectives resemble schizophrenics, while using the RDC mainly

affective, manic subtype definition of schizoaffective disorder will

likely define a group of patients that do not differ from bipolar

patients. One method that has been suggested to resolve this problem

with definition is to use the polydiagnostic method proposed by Maj

(1984). In this approach, all of the multiple definitions of

schizoaffective disorder are used to define different groups of

schizoaffectives to study. While this method allows for comparability

of studies, it still retains the dilemma about which group of patients

to select as the actual schizoaffectives.

Another approach that may be useful is selecting the definition of
schizoaffective disorder with the highest consensus among clinicians

based on their experience with these disorders. The prototype model may

be most appropriate for this investigation since clinicians and their

concepts about diagnosis, not patients, are the objects of study.

Because of the highly uncertain status and changing definitions of

schizoaffective disorder, the accumulated knowledge of clinicians may

yield valuable information about this concept. While empirical studies

of patients that meet operational criteria may be useful in the future,

in order to select a set of definitional criteria, it should first be

determined which model or view of schizoaffective disorder makes sense

to clinicians. The different prototype methodologies attempt to explore

their knowledge using direct and indirect methods.

First, various methodologies used within the prototype model are

presented. Then, findings relevant to the classification of

schizoaffective disorder from studies utilizing the prototype view are

discussed. Finally, a series of studies utilizing the prototype

methodology to study the classification of schizoaffective disorder and

test the alternative models is discussed.

Standard Prototype Methodology
A number of methodologies have commonly been used in the prototype

research. Several methodologies have developed out of the early work

utilizing prototypes to study pattern recognition. First and most

frequent is having subjects learn to classify patterns that differ

according to degree of prototypicality. The effect of typicality on

rate of learning (number of errors, trials to criterion) and transfer of

learning to a new set of patterns (classification rate or reaction time

(RT), number of errors in classification) are examined.

Using artificial categories of numbers and letters, Rosch and

Merviss (1975) found that items typical of a category were learned

faster and identified more rapidly and accurately once learned. Rosch,

Simpson and Miller (1976) also found that highly typical instances of

dot patterns were learned more quickly than atypical ones and later were

classified faster and with more accuracy. Earlier, Posner et al. (1967)

had found that the ease (speed and accuracy) with which subjects learned

to classify dot patterns correlated with the degree of prototypicality

of the pattern and that once the classification was learned, typical

patterns were classified faster and with more accuracy.

This situation is analagous to the situation in psychopathology.

When first learning, clinicians make diagnoses of cases and receive

feedback until a criterion level of accuracy is reached. Later, the

clinician is presented with a set of new patterns of symptoms and is

expected to make diagnoses based on transfer of learning.

A related paradigm is the recognition paradigm which assumes that

the classification is already known by the subjects. In a recognition

paradigm, subjects are asked to identify members of categories that

differ according to degree of prototypicality. Accuracy of

categorization (diagnosis) and speed in identification (reaction time)

are the usual dependent measures. Often prototypicality or typicality
ratings are also obtained. Prototypic members are those identified

rapidly and with a high degree of accuracy. Usually a criterion level

of speed and accuracy is set beforehand. For example, Horowitz, French,

Lapid and Siegelman (1981) found that cases with a high level of

diagnostic consensus had significantly more prototypic features than

those with low diagnostic agreement. Within this paradigm it is also

useful to look at category confusions since increased confusion between

two categories suggests that subjects perceive them to be very similar.

A multidimensional scaling analysis of the confusions can be performed

in order to construct a dimensional representation of the categories

based on perceived similarity.

Often subjects are asked to rate the distance between an instance

and the category prototype. This paradigm is similar to the use of

prototypicality ratings in Rosch's work in studying natural language

structure. Using common noun categories, Rosch (1975) found that

subjects could rate typicality of a category member with a high degree

of reliability. Rosch and Merviss (1975) found that prototypicality

ratings cui elated with the degree of family resemblence or number of

features of that category that an instance possessed. Horowitz, Post,

et al. (1981) found that case histories were judged to be more depressed

or typical of depression when they contained more features of the

prototype for depression. Rosch, Simpson and Miller (1976) used a

prototypicality rather than a distance rating in their pattern

recognition study. They also asked subjects to draw all patterns they

could remember afterwards, hypothesizing that prototypic instances are

more memorable. While they found that instances rated as prototypic

were more often recalled, subjects may have remembered prototypic items

better because they had just rated them as most typical.

A final paradigm used in the pattern recognition research is having

subjects make paired similarity ratings between all instances of a

category. A multidimensional scaling analysis is then used to produce a

conceptual map of the structure of the categories. This paradigm has

also been used in Rosch's work in semantic memory using common nouns.

Rosch and Merviss (1975) utilized the paired similarity rating approach

to examine prototypicality. Earlier, a group of subjects listed 21

instances for 6 general categories of English nouns (e.g., fruit). Then

an additional 15 subjects per category rated all possible pairs of the

21 items and the general category itself on a 9-point scale of

similarity. The resulting similarity matrices were then analyzed for

each category separately using multidimensional scaling to examine the

structure of the semantic space. They found that instances rated as

most prototypic of the category, as well as the general category name,

were most central in the solutions regardless of the number of

dimensions or types of rotations used. Prototypic members were central

and atypical members peripheral because the prototypic members were

rated as similar to all category instances while less typical members

were rated as similar to only some other members. Posner et al. (1967)

also used paired distance ratings between all pairs of dot patterns in a

category and found that these ratings correlated with the amount of

distortion from the central or prototypic pattern for the category

suggesting that the ratings were reliable.

One problem with this paradigm concerns the reliability of the

similarity judgments. Recently, it has been questioned whether subjects

can reliably reduce multidimensional differences to a single scale of

perceived distance (Green, 1975). For example, Summers and McKay (1983)

found that random error accounted for the majority of the variance in

subjects' similarity judgments about automobiles. They recommended that

researchers test the reliability and validity of individual subjects'

maps before accepting the results of group data. When they had subjects

select their own cognitive maps they were not able to distinguish their

own from those of other subjects or the group data solution.

In the features list approach, subjects list the characteristic

features of a number of categories. Judges then consolidate the

multiple lists into a summary consensual list, usually based on a

criterion number of subjects who listed that feature. The resultant

feature list prototypes are summary or abstract representations of what

is known about the category (Cantor & Genero, in press). The lists of

features are then compared for nesting of hierarchically ordered

categories (subordinate categories containing features of superordinate

categories) or overlapping (sharing) of features between categories at

the same level. Richness and distinctiveness are used to select "basic

level" categories, those categories that provide maximal information and

act as anchoring points for other concepts. Richness refers to the total

number of features associated with a category, while distinctiveness

refers to the number of features unique to a category (i.e., not listed

for other categories). Basic categories are those that are relatively

rich and distinct from other categories. The lists of features may also

be used for construction of instances meant to represent various levels

of prototypicality of a concept. The features list approach has been

used widely in Rosch's studies of natural categories as well as in

recent studies of psychopathology (Cantor, Smith, French & Mezzich,

1980; Blashfield & Sprock, unpublished manuscript). There are a number

of problems with the features list methodology, however (Blashfield &

Sprock, unpublished manuscript). Most significant is that considerable

subjectivity enters into the consolidation of the individual subject

lists into a summary list because many linguistic decisions must be


Murphy and Merviss (1985) criticized the feature list methodology

as being inadequate to understand classification. One problem with the

methodology itself is that often it is necessary for the judges to

ammend the consolidated lists because subjects fail to list attributes

listed for superordinate categories as features of categories at lower

levels of abstraction. For example, "has wings" might be listed for the

category bird but not for robin. Which features are listed by subjects

is determined by the implicit contrast set (e.g., all animals versus

types of birds) selected for the category so that features listed for

higher level categories may be irrelevant for lower level categories.

While there are an infinite number of features that could be listed for

a category, certain attributes are listed by most subjects. According

to Rosch and the prototype view the listed features are those that are

perceptually salient or have maximal cue validity (increased probability

of being associated with a category). Instead, Murphy and Merviss

hypothesized that it is people's theories about a category, how they

interact with a category and how the category relates to their goals and

activities that determines which features are focused on. Further, a
list of features does not represent the subjects' understanding of a

category; discovering the subjects' knowledge about the relationship

between features and underlying theory is necessary to understand a


The notion of contrast categories has been utilized in prototype

research. Contrast sets are categories selected to assist

categorization by highlighting between category distinctions (Cantor &

Genero, in press). Rosch and Merviss (1975) obtained the most common

contrast sets for basic level categories by asking subjects to imagine

that they were asked to guess an object and the word on the stimulus

card was their response. Further, they were told that their guess was

very close and asked to make another choice. When features were listed

for the contrast categories, prototypic members of the target category

shared less features with the contrast set than less prototypic members.

More recently, Cantor and Genero proposed that there are two

processes in categorization, similarity matching and differentiation.

It is in the second process that a contrast set is selected. They

differentiated two stages in selecting a contrast set. First is

selection of a contrast category that is distinct and shares minimal

features with the basic category. For categorizing borderline

categories, however, they suggested using categories with feature

overlap as contrast sets. They hypothesized that in psychiatric

diagnosis, clinicians first perform the contrast with distinct

categories (e.g., rule out schizophrenia and organic disorder for an
affective disorder). Then, if the case is borderline, they make

comparisons with overlapping contrast sets (e.g., rule out bipolar

disorder for a major depressive). In Rosch and Merviss's study, it is

the second kind of contrast that subjects are being asked to make.

Murphy and Merviss (1985) criticized the prototype methodologies

including the feature list, exemplar and similarity rating approaches,

as being insufficient to explain and understand categorization. While

according to these authors, these approaches to understanding categories

are not wrong, they do not provide enough information to understand the

concept; knowledge of the subjects' theories about the category is

neccesary. Subjects' theories about categories include their knowledge

of the relationship between the features of the category, relationships

between categories and implicit contrast sets and the meaning of the

category to the subject, how it relates to their goals and activities.

However, it could also be argued that if people's theories about a

category guide which features are focused on in a feature list paradigm

or for making similarity judgements, it should also be possible to learn

something about people's theories about categories through examination

of which features are selected. Therefore, the results of these

different methodolgies can lend some insight into their theory about a

particular category.

Other paradigms used in semantic memory research by Rosch (1975)

include a priming paradigm, averaging of shapes and identifiability of

averaged shapes, hedges and substitutability into sentence frames,

listing of exemplars and listing superordinate categories. The last

four will be briefly discussed because they are relevent to this

proposal. Instances already rated for typicality are then rated for

substitutability for superordinate constructs in sentences by another

group of subjects. Prototypic members are highly substitutable for the
higher order term. The substitutability of members into a clause

expressing a linguistic gradient (linguistic hedge) has also been shown

to correlate with prototypicality ratings. In this paradigm, subjects

rate how well an example fits into a statement of the nature, "A

(instance) is virtually a (superordinate category)." Prototypic

members are rated as less suitable in the first slot than less

prototypic members. In another, subjects are asked to list members or

exemplars for a category. Reaction time and number of items listed are

the dependent variables. Fast RT and many exemplars or members suggest

that a clear prototype exists for a category. Subjects have also been

asked to list superordinate categories for instances. Intersubject

agreement is the dependent measure, with subjects showing a higher rate

of agreement for prototypic instances.

Related Research
Before discussing how the prototype model can be used to test the

alternative models of schizoaffective disorder and the methodology for
this study, the implications of several prior prototype studies for this

issue are presented.

Cantor et al. (1980) had 13 clinicians list the essential clinical

features of 9 functional disorders, functional psychosis, manic-

depressed, manic, manic-depressed, depressed, affective disorder,

schizophrenia, paranoid schizophrenia, chronic undifferentiated

schizophrenia and involutional melancholia. Judges combined the 13

individual feature lists using a minimum criterion of 3 subjects listing
a feature to arrive at one list of features per diagnosis. Examining

the features listed for schizophrenia, schizoaffective and affective

disorder showed a high overlap between schizoaffective disorder and

schizophrenia. In contrast, only 3 symptoms listed for affective

disorder were also listed for schizoaffective. No features of

schizophrenia and affective disorder overlapped. According to the

prototype view, if schizoaffective disorder is a subtype of

schizophrenia, it should possess many, but not necessarily all, of the

features of schizophrenia. The Cantor et al. results were consistent

with this notion. However, clinicians were asked to list the features

of the "schizophrenia, schizo-affective subtype", which may have biased

them towards listing schizophrenic features. While schizoaffective

disorder had a relatively rich prototype with 10 features (depression,

associational disturbance, elation, affect disturbance, delusions,

hallucinations, ambivalence, autism, inappropriate affect and

inappropriate behavior), none of these symptoms were distinctive or

unique to schizoaffective, suggesting it is not clearly differentiated

from other disorders, especially schizophrenia. Cantor et al. concluded

that schizophrenia and affective disorder were basic level categories

because their features lists were richest and most distinctive.

In the second part of their study Cantor et al. asked 9 clinicians

to diagnose 4 case histories selected to represent manic-depressed,

manic and depressed, and paranoid and chronic undifferentiated

schizophrenia. Clinicians selected from these four diagnostic choices

and also rated how well the patient fit the diagnosis on a 7 point

scale. Three case histories were written for each diagnosis to

represent either high, medium or low typicality according to the number

of features from the first part of the study that they contained. They

found that accuracy and confidence in diagnosis increased with number of

prototypic features.

In a second study that partially replicated Cantor et al.,

Blashfield and Sprock (unpublished manuscript) also used a features list

approach to study the classification of functional disorders. Twenty

clinicians were asked to list the essential features of the 9 categories

in the Cantor study minus involutional melancholia because it was no

longer an accepted diagnostic category. Categories were presented

without their superordinate classification (i.e., the term

schizoaffective was presented without reference to its superordinate

category). When 3 judges, all experienced clinicians, independently

combined the 20 individual lists to arrive at a single list of features

for each diagnosis, it was found that considerable subjectivity entered

into the consolidation process so that the three final lists were quite

different. A committee meeting between judges was used to resolve

differences and arrive at a single list of features for each diagnosis.

A criterion of at least 6 subjects listing a feature was used for a

feature to be included on the final list.

The final lists were considerably shorter than the Cantor et al.

lists, probably because subjects were personally tested and a 3 minute

time limit used for each diagnosis while the Cantor study was conducted

through the mail. A 3 minute time limit was used because pilot work

suggested that most clinicians could list the features in this time and

it forced subjects to list only the most essential features, not all

possible features they could think of given unlimited time. While the

schizoaffective list was much shorter than in the Cantor study (4

features), 3 of the 4 symptoms overlapped with those on the Cantor list.

Examining the overlapping features for evidence of nesting shows
that the features listed for schizoaffective overlapped approximately

equally with schizophrenia (1) and affective disorder (2), while

schizophrenic and affective disorder symptoms were mutually exclusive.

The results do not seem to support the view that schizoaffective

disorder is a subtype of either schizophrenia or affective disorder.

In order to support the notion that schizoaffective is viewed as an

independent diagnostic entity, rather than a borderline condition

between schizophrenia and affective disorder, it should be demonstrated

that clinicians have a clear prototype for schizoaffective disorder.
Two important concepts in the prototype research which have been used to

demonstrate prototypicality are richness and distinctiveness (Rosch,

1975). Returning to the Blashfield and Sprock study, it can be seen

that the prototype for schizoaffective disorder contained only 4

features that clinicians agreed upon (depression, both schizophrenic and

affective symptoms, elation and delusions) compared to 11 for

schizophrenia and 10 for affective disorder, and therefore was not very

rich. In addition, 3 of the features overlapped with schizophrenia or

affective disorder, leaving only one unique feature for schizoaffective

disorder. These results suggest that clinicians do not have a clear

prototype or stored cognitive representation for schizoaffective.

Instead, they make this diagnosis when mixed features of schizophrenia

and affective disorder are seen. The one unique feature was "both

schizophrenic and affective symptoms", providing support for the model

that it is a borderline condition.

When the criteria for inclusion in the final summary prototype is

reduced from 30% to 25% of the clinicians, 3 additional symptoms can be

added. Interestingly, one of these features is "responds to

antipsychotic medication", providing some support for the hypothesis

that treatment response may be influential in diagnosis. Patients with

mixed schizophrenic and affective symptoms who respond to antipsychotics

may be diagnosed schizoaffective, while those who respond to

antidepressants receive affective disorder diagnoses. Note that in

listing "responds to antipsychotics" clinicians are differentiating the

disorder from mood disorders, not schizophrenia. Based on the usual

interpretation of contrast sets, this group of clinicians viewed

schizoaffective disorder as more similar to affective disorder. It is

the response to antipsychotics rather than antidepressants that

differentiates schizoaffective from mood disorders. Treatment response

was not listed as a characteristic for any other diagnostic categories,

further suggesting that it may be an important variable only in

situations of diagnostic uncertainty.

In Part II of the Blashfield and Sprock study, the same 20

clinicians were asked to diagnose 8 case histories selected to represent

each of the diagnoses in Part I and which were rewritten to contain all

of the associated features of the categories found in Part I. They also

rated their level of confidence and the typicality of the case each on a

7 point scale. Because the cases contained all the features, they were

assumed to be highly prototypic and should have been diagnosed

accurately and with confidence. Compared to other categories, the

schizoaffective case was rated least typical and diagnosed with the

least confidence. In addition, only 40% of the clinicians correctly

diagnosed the case, the rest calling it paranoid schizophrenia. One

interpretation of the diagnostic confusion between the schizoaffective

case and the diagnosis of paranoid schizophrenia is that clinicians

perceive the two disorders as very similar. Therefore, the results

suggested that clinicians do not have a clear conceptualization of

schizoaffective disorder; the feature list was not rich or distinctive
and clinicians were not able to identify a representative

schizoaffective case with a high degree of accuracy. While Part I was

somewhat suggestive that schizoaffective disorder is seen as an

affective disorder with 2 affective features and affective disorder as a

contrast set, Part II was suggestive that it is seen as closer to

paranoid schizophrenia.

Horowitz, French, Lapid and Siegelman (1981) also examined if the

prototype approach could be applied to understanding how clinicians

diagnose psychopathology. They had 20 clinicians diagnose 26 case

histories selected at random from a large outpatient clinic using the

DSM-II. They divided the cases into high, medium and low consensus

groups based on the frequency of the same diagnosis being assigned.

Then 40 undergraduate psychology students rated each case for the

presence of symptoms abstracted by the authors from the DSM-II on a 5-

point scale. They compared the number of relevant and irrelevant

(features of other diagnoses not assigned) features for the consensus

diagnosis for cases in the 3 groups and found that high consensus cases

had the most prototypic features and fewest irrelevant features while

the reverse was true for the low consensus cases. The results support

the validity of applying the prototype model to diagnosing case

histories of psychopathology and its utility in understanding how

diagnoses are assigned.

Another more recent study employed the recognition or exemplar

paradigm to identify prototypic personality disorder cases. While the

Blashfield, Sprock, Pinkston and Hodgin (1985) study dealt with the Axis

II personality disorders, not major psychoses, it is worth describing

because of its unique methodology. Thirty case histories selected to

represent the personality disorders and several nonpersonality disorders

were presented on a microcomputer screen. The clinician simply had to

type in the number corresponding with the best diagnosis on a list of

the personality disorders. Not only was the task more interesting to

the subjects, but it also allowed for accurate timing of diagnostic

decision time which was automatically recorded by the internal clock in

the microcomputer. Rosch advocated using both accuracy and speed of

identification to locate prototypes or exemplars in a recognition

paradigm. The recognition parts of the Cantor et al. (1980) and

Blashfield and Sprock studies utilized only accuracy or diagnostic

agreement. This methodology allowed use of both criteria to identify

exemplars. A minimum level of interclinician agreement of 80% and a

diagnostic decision time less than the mean for all cases were the

criteria for a prototypic case. Reaction times were converted to

normalized standard scores (z-scores) for each subject so that negative

z-scores were less than the mean. This resulted in locating 9

prototypes, one for each of the personality disorders except for

avoidant, borderline and narcissistic. Afterwards, the diagnostic

confusions, times one clinician assigned a particular diagnosis while

another clinician assigned a different diagnosis, were counted to form a

matrix. It was assumed that the more times clinicians confused

diagnoses the more similar they were perceived. The similarity matrix

was then analyzed using multidimensional scaling, a non-linear procedure

that provides a visual representative of the structure of the

similarities matrix. The analysis yielded a circumplex ordering of the

personality disorders that appeared quite meaningful. While the
implications for diagnosis and classification of the personality

disorders are not of concern in this paper, the study does show the

ability of the methodology to identify exemplars and of the

multidimensional scaling analysis of the diagnostic confusions to

produce meaningful results.

Study to Test Alternative Models of Schizoaffective Disorder

The prototype methodology has been proposed as particularly useful

to study this enigmatic diagnosis, schizoaffective disorder. Several of

the methodologies already discussed were used to explore clinicians'

conceptualizations of the classification of this disorder. Because of

the nature of the prototype methodology, in which completion of

subsequent studies is dependent upon findings in initial stages, the

alternative strategies that were considered are presented.

In Chapter III a study that attempted to identify prototypic

schizoaffective cases is described. Case histories were collected from

journals and textbooks on psychopathology to represent particular

disorders, including schizoaffective, schizophrenia, bipolar disorder

and major depression, and edited to fit on a microcomputer screen.

Clinicians were asked to diagnose the cases, which were presented on a

microcomputer screen, and reaction time (RT) was automatically recorded.

Cases with high interclinician agreement and relatively fast RT were

called prototypic. Earlier, each schizoaffective case was rated for the

percentage of schizophrenic and affective symptoms using the Stone scale

and the Feighner criteria. Diagnoses of the schizoaffective cases were

examined according to the proportion of schizophrenic and affective

symptoms the cases contained. The diagnostic confusion was also

analyzed using multidimensional scaling (MDS) to construct the

clinicians' conceptual map of the functional psychoses.

Initially, it was proposed that the prototypic schizoaffective

cases found using the recognition paradigm would have been manipulated
to explore the relationship of temporal factors and treatment response

to assigning a schizoaffective diagnosis. The temporal sequence of

schizophrenic and affective symptoms would have been manipulated, so

that schizophrenic symptoms appeared first, affective symptoms appeared

first or the onset of both was simultaneous. Case histories would

contain the features they did originally; only the temporal factor or

order of appearance of symptoms would have been altered. Depending on

the number of prototypes found for schizoaffective disorder, half would

have been used in this study of temporal factors and the rest would have

been manipulated by adding two sentences concerning response to

treatment (e.g., to antipsychotics, lithium, antidepressants or non-

pharmacological intervention). A different group of clinicians would

have been asked to diagnose the cases and RT would be automatically

recorded. The effects of the temporal onset of schizophrenic and

affective symptoms on agreement and RT could then be examined and a MDS

analysis performed using diagnostic confusions as a measure of

similarity to confirm the results in the first part of the study.

However, the above manipulations required finding at least a few

schizoaffective prototypes to alter. Since it was obvious early in the

study that not one prototype was going to be located for schizoaffective

disorder, a decision was made to alter the methodology slightly and

consider this study a pilot study. One potential difficulty with the

study was that subjects had 17 categories of diagnoses to select from.

Subjects in the Cantor study, for example, had only 4 diagnostic

categories. In addition, many of these were at the subordinate level.

According to the prototype model, the basic level is the hierarchical

level at which categories are most clearly and reliably differentiated.

Further, many studies have shown that reliability is higher when broader

categories are used (Katz, Cole and Lowery, 1969; Kendell, 1973).
Therefore, the study was continued using only categories found to be at

the basic level in the Cantor et al. (1980) and Blashfield and Sprock

studies. Chapter IV presents this study and its results. Because only

one exemplar was located for schizoaffective disorder even with the

revised methodology which found 11 other exemplars, the manipulations of

temporal order of symptoms and treatment response could not be

performed. However, these manipulations were not crucial to the

question of which model of classification is most appropriate for

schizoaffective disorder.

Chapter V presents a study that uses the features list approach to

obtain summary prototypes for schizoaffective disorder, schizophrenia

and affective disorder and their subordinate categories. Clinicians

were asked to list the features that they perceived as most

characteristic of each diagnosis. Three judges, the author and two

experienced clinicians consolidated the individual subject lists into

the consensual prototype for each diagnosis. Because there have been

concerns about this phase of the procedure, two different strategies in

combining lists were used, an independent consolidation of lists by

judges and a committee meeting between the three judges. Summary

prototypes were then compared for feature overlap. The higher the

number of shared features, the more similar the two categories.

Superordinate categories were included to look for evidence of feature

nesting. Richness and distinctiveness were examined to determine if

clinicians have a clear prototypic representation for schizoaffective


In order to examine the validity of the MDS solution found

using the exemplar paradigm, another study was conducted using paired

direct similarity ratings. Clinicians were asked to rate the similarity

of a number of pairs of diagnoses, including schizoaffective disorder,


schizophrenia, schizophrenic subtypes and affective disorder subtypes. A

MDS analysis was performed to examine spatially the perceived

similarity between schizoaffective disorder and other diagnoses.


The purpose of this study was to attempt to identify exemplars or

prototypic cases for schizoaffective disorder utilizing the exemplar

methodology taken from the prototype approach. Clinicians were asked to

read a series of 30 case histories, half which were selected from the

literature to represent schizoaffective disorder, that were presented on

a microcomputer screen. Exemplars for a disorder are exceptionally

clear cases that most clinicians can quickly diagnose with a high level

of agreement. A minimum criterion of agreement of 50% of the subjects

was selected and a fast diagnostic decision time was defined as a

reaction time (RT) of less than the mean time for the 30 cases. Using

the computer presentation of cases and recording of data like the

Blashfield et al. (1985) study allowed for accurate recording of

diagnostic decision time and the chance to use both agreement and RT

criteria to identify exemplars. Identification of clear exemplars for

schizoaffective disorder would suggest that clinicians do have a clear

concept of schizoaffective disorder as a distinct diagnosis. The

failure to find any clear exemplars for schizoaffective disorder would

suggest that clinicians do not have a differentiated prototype for

schizoaffective disorder. Instead, schizoaffective disorder may be seen

as a subtype of schizophrenia, a subtype of affective disorder or as not

existing at all as a viable diagnosis.

Afterwards, a debriefing questionnaire was given to subjects to

determine if the cases appeared realistic and believable and to discover

which cases were remembered as vivid. Because exemplars are cases that

have clear, obvious diagnoses that are quickly identified by most

clinicians, they are more likely to be vivid and subsequently recalled

by subjects.

The diagnostic confusions, or number of times clinicians assigned

different diagnoses to the same case, were also examined to help

determine if clinicians confused schizoaffective cases with

schizophrenic or affective disorder. It was assumed that diagnoses that

are frequently confused are perceived as more similar by the clinicians.

A multidimensional scaling analysis was performed to map the structure

of the perceived similarity between diagnoses.

Prior to the clinicians reading the cases, three experienced

clinicians rated the schizoaffective cases for number of schizophrenic

and affective symptoms in order to compare prototypic and nonprototypic

schizoaffective cases to help determine which symptoms are essential for

a case to be diagnosed as schizoaffective.

If schizoaffective disorder is seen as a distinct entity, it was

hypothesized that clear schizoaffective exemplars would be found, there

would be little diagnostic confusion between schizoaffective and other

disorders so that it would not be clustered with other disorders in the

multidimensional scaling solution and subjects would recall these

exemplars as particularly vivid.

If schizoaffective disorder is seen as a subtype of schizophrenia

it was hypothesized that few, if any, schizoaffective exemplars would be

identified, classical schizoaffective cases from the literature would

contain primarily schizophrenic symptomatology and be diagnosed as

schizophrenic disorders, and there would be much confusion between

schizoaffective and schizophrenic diagnoses so that they would be

geographically close in the multidimensional scaling solution.

If schizoaffective disorder is seen as a type of affective

disorder, it was hypothesized that few, if any, exemplars would be found

for it, cases representing schizoaffective disorder would contain

primarily affective symptoms and would be diagnosed as major depression

or bipolar disorder, and there would be much confusion between

schizoaffective and affective disorder diagnoses so that they cluster

together in the multidimensional scaling solution.

Finally, if schizoaffective disorder is not seen as a viable

diagnostic concept, no schizoaffective disorder exemplars would be

identified and in fact, clinicians would avoid using this diagnosis.

The multidimensional scaling analysis would show schizoaffective

disorder near the origin on each dimension, suggesting that it is not

well-represented by the dimensional structure characterizing the other

disorders or that it was confused with a number of different diagnoses.


Eight subjects, 7 psychologists and one psychiatrist, served as

pilot subjects to read the 30 cases and assign diagnoses from among 17

subcategories. Five of the psychologists were faculty members, one, a

post-doctoral student and the last, an intern. Three of the faculty had

psychology licenses in Florida and one in Michigan. The psychiatrist, a

third year resident, had a Florida license in medicine. All are at the

University of Florida except for one psychology faculty member and the

one intern who were at the Gainesville Veterans Administration Hospital.

There were 6 males and two females ages 28 to 56 years old (mean = 34.5,

SD = 9.0, median = 31.0). Years of clinical experience ranged from 2 to

25 years (mean = 7.9, SD = 7.4, median = 5.5). Examining the proportion
of their time spent in various activities, 0% to 85% was spent in

research (mean = 30.0, SD = 31.3, median = 17.5), 10% to 100% of their

time was devoted to clinical activities (mean = 43.8, SD = 31.1, median

= 45.0) and 0% to 80% to teaching (mean = 17.5, SD = 18.9, median =

12.5). Four primarily used the DSM-III diagnostic system, three, the

ICD-9 and one, both equally. Clinically, two primarily saw outpatients

while the remainder had a "mixed" practice of outpatient, inpatient and

consultation work. Most (6 clinicians) primarily saw patients between

18 and 45 while the remaining two also saw elderly patients. All but

one, who mainly treated females, saw an equal distribution of males and

females. Diagnostic categories most frequently represented in their

practice were nonpsychotic Axis I disorders (3 clinicians), Axis II

disorders (2 clinicians) and No Axis I or II diagnosis (1 clinician).

Two clinicians reported a mixture of psychotic and nonpsychotic Axis I

and II disorder patients.

The 3 judges who rated the schizoaffective case histories for

presence of schizophrenic and affective symptoms were all experienced
clinical psychologists. Two were faculty at the University of Florida

and one was at the Veterans Administration Medical Center in

Gainesville, Florida.
StLi uli

A total of 30 cases histories were selected from an initial pool of

47 cases. Schizoaffective cases were selected from 5 sources, Dempsey,
Tsuang, Struss and Dvoredsky-Wortsman (1975), Cohen, Allen, Pollin and

Hrubec (1972), Stone (1930), Kasanin (1933) and the DSM-III Case Book

(Spitzer, Skodol, Gibbon et al., 1981). The cases were abstracted from

these sources to fit on a microcomputer screen. Of the 17
schizoaffective cases that were examined, 15 were eventually selected.

In addition, 15 cases, two each for major depression, bipolar disorder,

schizophrenia, schizophreniform disorder and organic disorders plus five

representing other disorders were selected. Sources for these included

Akiskal, Djenderedjian, Rosenthal and Khani (1977), Stone (1980),

Goldenberg (1977), Rainy and Crowder (1975), Kolb (1973), the DSM-III

Training Guide (Webb, Diclemente, Johnstone et al., 1981) and the DSM-

III Case Book (Spitzer et al., 1981). Table 3-1 lists the source, the

intended diagnosis and the number of lines after editing for each of the

30 case history stimuli.

In order to help determine which characteristics of these cases are

associated with the case being an exemplar for schizoaffective disorder

(clinicians easily recognize it as schizoaffective based on agreement

and RT), each schizoaffective case was rated for the presence of

schizophrenic and affective symptoms.

Initially, Stone's scale (1980), a modification of a scale by Cohen

et al. (1972), was used to rate each schizoaffective case. This scale

contains a list of 12 affective symptoms and 8 schizophrenic symptoms

that are weighted. The weighted symptoms present in a case are then

summed to estimate the percentage of schizophrenic and affective

symptoms in that case. Two experienced clinicians rated the presence of

each of these 20 symptoms on a 3 point scale: 2 (definitely present), 1

(probably present) and 0 (absent) (See Appendix 1).
While there was a significant correlation between the ratings of

the two clinicians (r = .579, T = 13.3832, df = 355, p < .001) across

all symptoms across all cases, this accounted for only 33% of the

Table 3-1

Description of the 30 Case History Stimuli

Intended Diagnosis

Number of Lines

Dempsey et al. (1975)
Dempsey et al. (1975)
Dempsey et al. (1975)
Cohen et al. (1972)
Stone (1980)
Stone (1980)
Stone (1980)
Stone (1980)
Stone (1980)
Kasanin (1933)
Kasanin (1933)
Kasanin (1933)
Kasanin (1933)
Kasanin (1933)
DSM-III Case Book (1981)

DSM-III Case Book (1981)
DSM-III Training Guide (1981)
Akiskal et al. (1977)
Stone (1980)
Goldenberg (1977)
DSM-III Case Book (1981)
DSM-III Case Book (1981)
Rainy & Crowder (1975)
DSM-III Case Book (1981)
DSM-III Case Book (1981)
DSM-III Case Book (1981)
Kolb (1973)
Kolb (1973)
Kolb (1973)
DSM-III Case Book (1981)


Bipolar, manic
Paranoid schizophrenia
Major depression
Bipolar, manic
Unspecified psychosis
PCP psychosis
Atypical psychosis
Major depression
Catatonic schizophrenia
Organic senile psychosis
Paranoid psychosis
Schizotypal personality



variance. In addition individual correlations for symptoms or cases

were not as high, suggesting little consistency between the two

clinicians in rating these symptoms. For the original group of 17

cases, correlations between the two clinicians across symptoms ranged
from a high of .949 (case #18) to a low of .148 (case #11), with a

median value of .539. Other very low nonsignificant correlations

were .161 (case #10), and .276 (case #9) (See Appendix 2).

Correlations between the two raters across the 17 cases for the

individual symptoms were also quite variable ranging from a low of -.111

for autistic thinking/delusions to 1.0 for tearfulness. Again, a number

of the correlations were quite low for some symptoms including suicidal

gestures (.087), chronicity (.199), psychomotor retardation (.239),
agitation (.305), inappropriate affect (.333), flat affect (.347),

mannerisms (.404), flight of ideas (.409) and mutism (.471), all of

which were nonsignificant (See Appendix 3). The very low agreement

between the two clinicians suggested that the ratings were too

unreliable to be used for analyzing the features of the cases. Looking

at the differences between the two clinicians reveals that one tended to

give higher ratings, especially to the schizophrenic symptoms (See

Appendix 4).

The weighted scores were then examined for each clinician using

just the symptoms the clinicians were certain were present. For the

schizophrenic symptoms, the correlation between the two clinicians

across the 17 schizoaffective cases was just .434 (T = 1.860, df = 151,

p < .05). The mean percentage of schizophrenic symptoms in the cases

was 37% and 47% for the two clinicians. When probable symptoms were
included, the correlation increased to .509 (T = 2.290, df = 151,

p < .05) and the percentage of schizophrenic symptoms increased to 47%

for one clinician and 61% for the other, who was more inclusive in

rating the presence of a symptom. When the weighted scale for affective

symptoms was examined for symptoms clinicians were certain were present,

the correlation between the two clinicians was nonsignificant (r = .404,

T = 1.711, p > .05). The mean percentage of affective symptoms rated by

the two raters was 22% and 24%. When probable affective symptoms were

included, the correlation increased to .692 (T = 3.709, p < .01) and the

mean percentage for affective symptoms for the cases was 33% and 37% for

the two clinicians. Therefore, using this scale, most of the

schizoaffective cases were seen as having a higher percentage of

schizophrenic than affective symptoms (See Appendix 5).

Cases were also assigned to one of the five regions described by

Stone (region 1: 80-100% schizophrenic symptoms to region 5: 80-100%

affective symptoms). Using just symptoms clinicians were certain of

resulted in very low agreement regarding region assignment (r = .292, T

= 1.182, df = 15, p > .05). Using both certain plus probably present

symptoms to assign the cases to the five regions resulted in higher

agreement (r = .570, T = 2.688, p < .05) but the percentage of the

variance accounted for was still very low (35%). The median region was

region 2 for both clinicians when only certain symptoms were used,

suggesting the cases were seen as having more schizophrenic than

affective symptomatology. When certain and probable symptoms were used,

the median was still region 2 for one clinician and 3 for the other

suggesting more affective symptoms were included.

Because of the inconsistency between the two clinicians in the

ratings of symptoms present in the cases, the Stone regions were

redetermined using symptoms that both clinicians endorsed. A criterion

of 3 for the summed certainty ratings of the two clinicians was selected

(i.e., one clinician rated the symptoms as definitely present while the

other rated it as at least probably present). This resulted in most of

the cases falling in region 2 with more schizophrenic than affective

symptoms and region 3 with an approximately equal percentage of

schizophrenic and affective symptoms. The remaining three cases were

divided between regions 1, 4 and 5 (See Table 3-2). Specific symptoms

endorsed by the two clinicians for each schizoaffective case are shown

in Table 3-3.

In general, the Stone ratings had low reliability, suggesting that

the data they yielded might not be very useful in examining differences

between prototypic and non-prototypic cases. Two reasons may account

for the low reliability. First, the symptoms are not operationally

defined or in any way described, but are simply a list of symptoms. In

addition, a number of the symptoms actually consist of a combination of

symptoms. Often, the combined symptoms did not co-exist and clinically

they did not seem to be logical to combine. For example, autistic

thinking and delusions of grandeur and persecution were combined as were

disorganized thinking, ambivalence, bad dreams and poor memory. The

narrow range of regions represented is also less useful when looking for

differences between cases. While the Stone continuum is a novel way of

classifying schizoaffective cases, the number of individual

schizophrenic and affective symptoms present in a case were too

unreliably rated to be useful. Therefore, it was decided to use another

set of criteria with operational definitions. The Feighner criteria,

one of the first sets of diagnostic criteria to utilize operational

definitions, was selected and the rating process was repeated for the

presence of schizophrenic and affective symptoms in the schizoaffective



Table 3-2

Percentage of Total Possible Schizophrenic and Affective Symptoms,
Proportion of Schizophrenic and Affective Symptomatology and Stone
Regions Based on 2 Clinicians Endorsing Symptom (minimum rating of 3)

Case Percent of

Percent of


Percent of Total Symptoms
Schizophrenic Affective




Table 3-3
Consensus Stone Symptoms for 17 Schizoaffective Cases


Halluc(4) Inapp aff(4) Mann(4) Rep ep(4) Aut/dels(3)
Aut/dels(4) Disorg(4) Euph(4) Flight id(4) Halluc(4)
Agit(4) Aut/dels(4) Insom(4) Wt loss(4) Disorg(3) Euph(3) Flat
aff(3) Flight id(3)
Aut/dels(4) Disorg(4) Halluc(4) Agit(3) Flat aff(3)
Aut/dels(4) Disorg(4) Halluc(4) Insom(4) Tearful(4) Flat aff(3)
Inapp aff(3) Rep ep(3) Self-dep(3)
Dep(4) Halluc(4) Insom(4) Mann(4) Chronic(3) Retard(3) Wt loss(3)
Aut/dels(4) Disorg(4) Halluc(4) Inapp aff(4) Agit(3) Self-dep(4)
Flight id(3) Mann(3)
Agit(4) Aut/dels(4) Disorg(4) Insom(4) Mann(4)
Aut/dels(4) Dep(4) Disorg(4) Insom(4) Self-dep(4) Tearful(4)
Chronic(4) Dep(3) Retard(3)
Dep(4) Agit(3)
Aut/dels(4) Inapp aff(4) Insom(4) Self-dep(4)
Aut/dels(4) Disorg(4) Self-dep(4) Dep(3)
Aut/dels(4) Disorg(4) Euph(4) Halluc(3) Mann(3) Sui gesture(3)
Aut/dels(4) Dep(4) Disorg(4) Insom(4) Somat co(4) Tearful(4)
Agit(4) Aut/dels(4) Disorg(4) Mann(4) Euph(3) Flight id(3)
Aut/dels(4) Dep(4) Disorg(4) Halluc(4) Insom(4) Wt loss(3)

Agit = Agitation
Aut/dels = Autism, delusions
Dep = Depressed affect
Disorg = Disorganized thinking
Euph = Euphoria
Flat aff = Flat affect
Flight id = Flight of ideas
Halluc = Hallucinations
Inapp aff = Inappropriate affect

Insom = Insomnia
Mann = Mannerisms
Retard = Retardation
Re ep = Repeated episodes with
good remission
Self-dep = Self-deprecation
Somat co = Somatic complaints
Sui gesture = Suicide gestures
Wt loss = Weight loss

Note: Number in parentheses represents the degree of certainty of
the presence of the symptom for both clinicians combined


Schizophrenic and affective disorder symptoms abstracted from the

Feighner criteria resulted in a list of 22 symptoms, 16 affective and 6

schizophrenic symptoms (See Appendix 6). This time, three experienced

clinicians were asked to rate the cases. Again, there was variability

in the reliability with which symptoms were rated (See Appendices 7

and 8). Appendix 9 shows the same correlations between ratings for

symptoms when cases 1 and 2 were eliminated. This was done in order to

have exactly half of the cases be schizoaffective and because these two

cases were very short (7 and 8 lines) and seemed more like a list of

symptoms than real case histories.

In general, eliminating cases 1 and 2 resulted in the same or

higher reliability ratings between the 3 clinicians. Looking at

Appendix 8 shows that the ratings of the 3 clinicians correlated

significantly with each other for most cases. Exceptions were cases

#3,5,6 and 11. In cases #3 and 5, the correlation between clinicians 2

and 3 was nonsignificant, while in case #6, the correlation between

clinician 1 and 2 was not significant. In case #11 the ratings of both

clinicians 1 and 2 correlated nonsignificantly with clinician 3. In

general, the correlations between clinicians 1 and 2 were highest and

quite good, with an overall correlation across all cases and symptoms

of .663 (T = 16.057, p < .001). Correlations between clinicians 1 and 2

ranged between .41 and .93 (median = .68), between clinicians 1 and 3

from .32 to .73 (median = .58) and between clinicians 2 and 3, .22

to .79 (median = .58).

Appendix 9 reveals the variability in the reliability with which

the three clinicians rated the symptoms. Three symptoms had especially

poor reliability, agitation/retardation, chronicity and complaints of

diminished ability to think, all with nonsignificant ratings between the

clinicians. Two symptoms were not rated as present in any case by at

least one clinician so their reliability could not be calculated, family

history of schizophrenia and flight of ideas. Several symptoms were

rated with good reliability between clinician 1 and 2, but with poor

correlation with the ratings of clinician 3, decreased sleep,

distractibility, euphoria, hyperactivity and recurrent thoughts of

suicide. Symptoms with good reliability were delusions/hallucinations,

dysphoria, guilt, grandiosity, lack of understandable speech, loss of

interest in usual activities, onset before 40 and push of speech.

Symptom correlations ranged from 0 to 1.0 (median = .68) for clinicians

1 and 2, from .35 to .94 (median = .61) for clinicians 1 and 3, and .27

to .93 (median = .49) for clinicians 2 and 3.

Appendix 10 shows the correlation between the three clinicians in

the number of affective and number of schizophrenic symptoms they rated

as present in a case using all symptoms definitely plus probably present

and just symptoms clinicians were certain were present. In general,

there was much higher agreement for the schizophrenic symptoms,

especially when definitely and probably present symptoms were both used.

For number of affective symptoms, all correlations between the three

clinicians were nonsignificant. The means are also presented at the

bottom of the table. Looking at the schizophrenic symptoms, the

clinicians rated approximately the same number of symptoms. There is
considerable variability, however, in the number of affective symptoms

with clinician 2 rating more affective symptoms as probably present but

fewer as definitely present than the other two clinicians.

As with the Stone rating data a decision was made to use a

consensus between the three raters to arrive at the number of

schizophrenic and affective symptoms that will be considered as present

in a case. A cutoff of 4 for the combined certainty ratings for the

three clinicians was selected as a minimum criterion for deciding that a

symptom was present in a case. At least one clinician had to rate the

symptom as definitely present (2), while the other two clinicians rated

it as probably present (1), or two of the three clinicians had to rate

the symptom as definitely present (2). This resulted in the consensus

number of schizophrenic and affective symptoms for each case. Table 3-4

presents these results as well as the particular symptoms rated as

present in each case. Using consensus is a conservative method that may

tend to underestimate the number of symptoms in a case, especially if

all three clinicians rate the symptom as probably present (sum of 3).

An intraclass correlation (ICC) was also used to examine the

reliability between the three clinicians since Bartko and Carpenter

(1976) recommended it as the more appropriate statistic (See Appendix

11). The ICC could have ranged from -1/2 to 1.0 in this case and was

found to be .594, df=2/660. The percent of the variance due to

disagreement among the raters was 41%. The unbiased ICC was also

calculated using a correction formula and was .593. This suggests that

the reliability between the three raters was reasonably good.

A simple debriefing questionnaire was constructed consisting of

four questions: 1. Did the cases appear realistic to you? 2. Which case

was most vivid to you? 3. Which case seemed most believable to you?

4. Would you have liked any additional information about these cases

before making your diagnoses? If so, what?


Clinicians were asked to read a series of 30 cases each presented

in random order on a single screen on the microcomputer. After reading

Table 3-4
Frequency of Feighner Criteria Symptoms in Schizoaffective Cases Based
on Endorsement by Three Clinicians (Minimum Rating of 4)
Case Schizophrenic Affective Case Schizophrenic Affective
3 2 6 11 1 4
4 2 3 15 2 4
5 3 6 16 2 5
6 4 6 17 1 4
7 2 6 18 2 3
8 4 4 19 1 4
9 4 4 21 2 6
10 2 2
Case Symptoms Endorsed in Order of Probability (Minimum rating of 4)
3 Onset<40(6) Hyperact(6) Dec sleep(5) Sleep diff(5) Dels/halluc(4)
Agit/retard(4) Euph(4) Push of Speech(4)
4 Dels/halluc(6) Onset<40(6) Grand(6) Hyperact(6) Push of Speech(4)
5 Dels/halluc(6) Onset<40(6) Grand(6) Distract(6) Dec sleep(5) Sleep
diff(5) Hyperact(4) Euph(4) Lacks Under Speech(4)
6 Dels/halluc(6) Onset<40(6) Dysph(6) Poor ap/wt loss(6) Chronic(5)
Guilt(5) Hyperact(5) Agit/retard(4) Poor premorb(4) Suic id(4)
7 Dels/halluc(6) Onset <40(6) Agit/retard(6) Grand(6) Euph(6)
Guilt(6) Push of speech(5) Hyperact(4)
8 Dels/halluc(6) Dec sleep(6) Sleep diff(6) Onset< 40(5) Dysph(5)
Lacks und speech(5) Grand(4) Poor premorb(4)
9 Dels/halluc(6) Dysph(6) Dec sleep(6) Sleep diff(6) Poor premorb(6)
Onset <40(5) Suic id(5) Lack und speech(5)
10 Dels/halluc(6) Onset<40(6) Dysph(6) Suic id(5)
11 Dysph(5) Distract(5) Loss of int(5) Onset<40(4) Dim thinking(4)
15 Dels/halluc(6) Dec sleep(6) Suic id(6) Agit/retard(5) Onset< 40(4)
Sleep diff(4)
16 Onset<40(6) Guilt(6) Dels/halluc(5) Dysph(5) Dec sleep(4) Agit/
retard(4) Poor ap/wt loss(4)
17 Dels/halluc(6) Agit/retard(6) Euph(6) Dysph(5) Grand(4)
18 Dels/halluc(6) Onset< 40(6) Dysph(6) Dec sleep(4) Sleep diff(4)
19 Dels/halluc(6) Agit/retard(5) Hyperact(5) Grand(5) Push of
21 Dels/halluc(6) Dysph(6) Dec sleep(5) Sleep diff(5) Loss of int(5)
Loss of energy(5) Dim thinking(5) Chronic(4)
Agit/retard = Agitated/retarded Lacks und speech = Lacks under-
Dec sleep = Decreased sleep standible, logical speech
Dels/halluc = Delusions/hallucinations Loss of int = Loss of interest
Dim thinking = Diminished thinking Poor ap/wt loss = Poor appetite
Distract = Distractibility or weight loss
Dysph = Dysphoric Poor premorb = Poor premorbid
Euph = Euphoric Sleep diff = Sleep difficulties
Grand = Grandiosity Suic id = Suicidal ideation
Hyperact = Hyperactivity

each case they were asked to assign a diagnosis by typing in a number

from 1 to 17 that corresponded with the diagnosis they selected from the

list and the next case automatically appeared. Appendix 12 presents

the list of their diagnostic choices which included bipolar disorder,

major depression, five subtypes of schizophrenia, schizoaffective

disorder, schizophreniform disorder, atypical and brief reactive

psychoses, and three types of organic disorders as well as non-psychotic

mental disorder and "no Axis I disorder." Subjects were asked to

familiarize themselves with the list before reading the first case.

Subjects were asked to assign a code name to preserve anonymity and to

help keep subjects' data separate. They were then asked to select a

three digit number in order to start the random number generator for

determining the order of presentation of the cases. Specific

instructions were:

This is a study concerning the classification of psychotic
disorders. You will be presented with 30 case histories and asked
to assign diagnoses to the persons described in these cases. The
list of possible diagnoses along with number codes for the
diagnoses are typed on a piece of paper to your right. Please use
the number code when assigning diagnoses. Notice that there is one
special number code: 999. This code should be used if you need to
take a break (e.g., to use the restroom or to answer your beeper).
If you type this code, the computer will stop and wait for your
return before proceeding with other case histories. If you have
any questions, do not hesitate to ask the investigator who is
Please type the code name you are using to identify yourself
in this study. To preserve your anonymity, you can use any code
name you wish. After typing your code name, please hit the ENTER
This program will ask your diagnoses of a set of case
histories. After being shown each case, type the number of the
diagnosis which you believe best fits the case and then hit ENTER.
When you are ready to proceed, type any three digit number and then
hit ENTER. Please do not use numbers such as 123 or 999 that are
likely to be generated by other subjects.

After they entered a three digit number, the first case appeared.

Reaction time to read each case and assign a diagnosis was automatically

recorded by the internal clock in the microcomputer. Afterwards,

subjects were asked to answer the 4 questions on the debriefing


Exemplars for a disorder were defined as cases in which at least

half of the clinicians assigned the same diagnosis in a relatively short

time period. RT to assign a diagnosis was automatically recorded using

the internal clock in the microcomputer and then changed to z scores for

each subject to control for across subject differences in reading time.

In order to be considered an exemplar, a case should be diagnosed

relatively quickly, which was defined as a mean diagnostic decision time

of less than z = 0.00 (i.e., less than the mean for reading the 30


Table 3-5 presents diagnoses selected by two or more clinicians for

each of the 30 cases in order of highest frequency along with the

intended diagnosis, while Table 3-6 presents the RT data. When subjects

selected from among 17 subordinate level categories in assigning

diagnoses to the 30 cases, no exemplars were found for schizoaffective

disorder. In fact, of the 120 diagnoses assigned to the 15 cases

intended to represent schizoaffective disorder, the diagnosis of

schizoaffective disorder was assigned only 7 times (6%). Cases with the

highest number of schizoaffective diagnoses (#7,21) (25%) did not meet

diagnostic agreement or RT criteria for being an exemplar. However,

three schizoaffective cases did meet criteria for exemplars for other

diagnoses with a negative mean RT and exactly 50% clinician agreement

on diagnosis, case #11 for nonpsychotic disorder, case #15 for brief

reactive psychosis and case #18 for paranoid schizophrenia.

Table 3-5
Most Frequent Diagnoses Assigned by 8 Clinicians to 30 Cases Using 17

Case Intended Diagnosi
3 Schizoaffective
4 Schizoaffective
5 Schizoaffective
6 Schizoaffective
7 Schizoaffective
8 Schizoaffective

9 Schizoaffective
10 Schizoaffective
11* Schizoaffective
15* Schizoaffective
16 Schizoaffective
17 Schizoaffective
18* Schizoaffective
19 Schizoaffective

21 Schizoaffective

1 Bipolar
2 Paranoid Schiz
12* Cyclothymic
13 Major Depression
14* Bipolar
20 Schizophreniform
22 Unspecified Psy
23* PCP Psy
24* Atypical Psy
25 Schizophreniform
26 Major Depression
27 Catatonic Schiz
28* Dementia
29 Paranoid dis
30 Schizotypal


Most Frequent Diagnoses

Paranoid dis(3) Paranoid schiz(2)
Atypical psy(2) Schizophreniform (2)
Disorganized schiz(3) Bipolar(2)
Bipolar (3)
Schizoaffective (2) Schizophreniform (2)
Undifferentiated schiz(3) Disorganized schiz(2)
Paranoid schiz(2)
Undifferentiated schiz(3) Catatonic schiz(2)
Major depression(2)
Major depression(3)
Non-psychotic dis(f) No Axis 1(3)
Brief reactive psy(4)
Paranoid dis(2) Schizophreniform(2)
Schizophreniform(3) Undifferentiated schiz(2)
Paranoid schiz(4) Paranoid dis(2)
Bipolar(3) Schizophreniform(2) Undifferentiated
Paranoid schiz(4) Schizoaffective(2)

Paranoid schiz(3) Bipolar(2)
Paranoid schiz(2) Paranoid dis(2) Disorganized
No Axis I dis(4) Non-psychotic dis(2)
Bipolar(5) Brief reactive psy(2)
Atypical psy(2) Brief reactive psy(2)
Disorganized schiz(2) No Axis I dis(2)
Substance-induced psy(5)
Brief reactive psy(6)
Atypical psy(2) Paranoid dis(2) Disorganized
Major depression(3) Catatonic schiz(2)
Catatonic schiz(4)
Undifferentiated schiz(2) Disorganized schiz(2)
No Axis I dis(3) Undifferentiated schiz(2)

dis = disorder
psy = psychosis
schiz = schizophrenia
*met criteria as an exemplar

Table 3-6

Mean Decision RT for 30 Cases by 8 Clinicians

Case RT (mins.)




Case RT (mins.)




Five of the nonschizoaffective cases met both diagnostic agreement

and RT criteria as exemplars. Cases #12 and 14 were exemplars for

bipolar disorder with 75% and 63% diagnostic agreement and fast RTs.

Cases #23 met criteria for a substance-induced disorder with 63%
clinician agreement while #28 was diagnosed as dementia by 75% of

clinicians, each also being diagnosed quickly. Case #24 intended to

represent an atypical psychosis was found to be an exemplar for brief

reactive psychosis with 75% interclinician agreement and a relatively

fast mean diagnostic decision time.

Looking at the diagnoses assigned to the schizoaffective cases more

closely reveals that clinicians tended to diagnose these cases as a

schizophrenic subtype (42 of 120 = 35%) more often than an affective
disorder (bipolar or major depression) (20 of 120 = 17%). These cases

were also perceived as schizophreniform more than twice as often as

schizoaffective disorder (16 vs. 7). Examining individual cases shows

that a schizophrenic subtype diagnosis was assigned by 88% for case #8,

75% for case #9 and by 63% for case #5. Only case #6 was seen as an

affective disorder by half of the clinicians. All of the clinicians saw

these cases as functional disorders; not one organic diagnosis

(dementia, delirium, substance induced psychoses) was assigned.

The symptoms present in the schizoaffective cases were then

examined. While the Feighner criteria ratings suggested that the cases

contained mostly affective (mean = 4.5) instead of schizophrenic (mean =

2.3) symptoms, the weighted Stone ratings suggested that the cases

contained mostly schizophrenic symptomatology. The major difference

between the ratings was that the Stone ratings were weighted to offset

the fewer possible schizophrenic (6 and 8) than affective (12 and 16)

symptoms on both scales. In addition, the schizophrenic symptoms

contained in most of the cases, hallucinations and delusions, had among

the highest weights.

Schizoaffective cases that received the highest number of

schizoaffective diagnoses, #7 and 21, fell in Stone regions 2 and 3,

respectively. While case #7 had 56% of possible schizophrenic and only

17% of possible affective symptoms, case #21 had only 36% of possible

schizophrenic symptoms and an approximately equal percentage of possible

affective symptoms (29%). Both cases contained the 3 schizophrenic

symptoms autism/delusions, hallucinations and disorganized thinking, but

case #7 had primarily manic affective symptomatology while #21 had 3

depressive symptoms. Feighner criteria ratings showed both cases had 2

schizophrenic and 6 affective symptoms. Each shared only one

schizophrenic symptom, delusions/hallucinations, and again, #7 had 6

manic symptoms (e.g., euphoria, push of speech) while #21 had 6

depressive symptoms (e.g., dysphoria, decreased energy).

Schizoaffective cases that met criteria as exemplars for other

disorders were #11 (nonpsychotic mental disorder), #15 (brief reactive

psychosis) and #18 (paranoid schizophrenia). Stone ratings gave some

insight into why they received these diagnoses. Case #11 fell in region

5 with 100% affective symptomatology and no schizophrenic symptoms.

Case #15 was in region 1 with 86% schizophrenic symptoms. Perhaps one

reason it was not diagnosed as schizophrenia was because it did not have

hallucinations and had only 25% of total possible schizophrenic

symptoms. Case #18 was in region 3 with an equal percentage of

affective and schizophrenic symptomatology, but contained 3 important

schizophrenic symptoms, autism/delusions, hallucinations and

disorganized thinking. Feighner criteria ratings for case #11 were for

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd