Intelligence in Schizophrenia: Meta-analysis of the Research

430 Intelligence inSchizophrenia:Meta-Analysis of theResearch

by Elizabeth Aylward,Elaine Walker,and Barbara Bettes

Abstract

This article combines a review andmeta-analysis of research on IQ inschizophrenia, with emphasis onareas of convergence in the findings,as well as questions that remain to beanswered. Taken together, thefindings suggest that early-onset andadult-onset schizophrenia areassociated with intellectual deficitsacross the lifespan. Preschizophrenicchildren, adolescents, and youngadults perform below matchedcontrols on a variety of standardizedmeasures of intelligence. SignificantIQ deficits are also apparent after theonset of the disorder. Moreover, IQis positively related to several indicesof prognosis, and, among hospi-talized patients, there is negativewithin-subject covariance betweenintellectual performance andsymptom severity. Although there isfairly consistent evidence that VerbalIQ is higher than Performance IQamong schizophrenic patients, amore specific pattern of subtestperformance is not apparent. Acentral question raised by the resultsis whether IQ is an independentlydetermined factor that can serve tomitigate the vulnerability ofindividuals who are constitutionallypredisposed to schizophrenia, orwhether intellectual deficit is onemanifestation of the constitutionalpredisposition to the disorder. Thefindings also raise the question ofpossible sex differences in thedevelopmental determinants ofschizophrenia: Meta-analysesrevealed that premorbid IQ deficitsare more prevalent among malesthan females.

Since the early 1900s whenstandardized intelligence tests werefirst being developed, a great deal ofspeculation and research has beendevoted to determining whether

relationships exist between IQ andcertain forms of psychopathology.Much of the early research in thisarea sought to discover patterns ofsubtest performance that coulddistinguish among diagnosticcategories. For example, Wechsler(1958) characterized schizophrenicpatients as having high scores on theInformation and Vocabulary subtestsof, the Wechsler Adult IntelligenceScale, and low scores on the ObjectAssembly and Digit Symbol subtests.While more recent research hascontinued to look for ways in whichintelligence tests can be used diagnos-tically, a number of other importantquestions related to this topic havealso been addressed. Researchershave sought to determine whetherintelligence deteriorates as a result ofschizophrenia, whether individualswith lower IQs are more susceptibleto schizophrenia, or whether IQ maybe related to such variables asseverity or chronicity of the disorder.Unfortunately, despite the greatamount and variety of research onthe relation between intelligence andschizophrenia, there is currently nocomprehensive theory regarding thenature of this relationship. Onereason for this is the abundance ofpresumably contradictory results. Asa case in point, while one inves-tigation provided evidence thatchildhood intelligence is not relatedto length of hospitalization forschizophrenic patients (Watt andLubensky 1976), "the most strikingresult" of another investigation "isthat schizophrenics with lowchildhood IQs . . . remain institu-tionalized significantly longer thanschizophrenics with average IQs"(Offord and Cross 1971, p. 431).Similarly, while several investigators

Reprint requests should be sent to Dr.E. Walker, HDFS, Cornell University,Ithaca, NY 148S3.

by guest on January 16, 2016http://schizophreniabulletin.oxfordjournals.org/

Dow

nloaded from

http://schizophreniabulletin.oxfordjournals.org/

VOL.10, NO. 3,1984 431

have described patterns of subtestscores that distingush schizophrenicpatients from normals (Wechsler andJaros 1965), other studies fail tosupport the hypothesis of differentialimpairment in schizophrenic patients'test performance (Binder 1965).

The intent of this article is toreview the findings of research onintelligence and schizophrenia. Areview of the research is important atthis time for several reasons. First,the most recent comprehensivereviews of the literature in this area(Payne 1960; Winder 1960) are over20 years old. During the past 20years, a great deal of research hasbeen conducted that sheds new lighton the issue of intellectual factors inthe developmental course of schizo-phrenia. Another reason forreviewing the literature at this time isto provide a firmer ground for futureresearch. The notion that certaincognitive characteristics may serve aspremorbid indicators of risk forschizophrenia has received increasingattention (Zubin and Steinhauer1981), and a significant proportion ofcurrent research on the disorderfocuses on cognitive processes.Moreover, recent interest insymptomatological, and possibleetiological, heterogeneity in schizo-phrenia has led to the suggestion thata subgroup of patients may besuffering from structural centralnervous system impairment andconcomitant cognitive deficits(Andreasen and Olsen 1982). Abetter understanding of the lifespancharacteristics of intellectualperformance in schizophrenia mayenhance our ability to pursue thequestions of specific cognitive deficitsand subgroup distinctions. As Rieder,Broman, and Rosenthal (1977) pointout, a thorough understanding of theIQ-schizophrenia relationship isimportant in order to develop themost efficient and productive

strategies for future research andintervention. Certain developmentsin methodology, such as longitudinalhigh-risk studies, offer new oppor-tunities for investigating therelationship between schizophreniaand IQ. Thus, this review is intendednot only to resolve contradictions ofpast research, but also to highlightthe questions that remain to beanswered. Although our review willgenerally be restricted to studiesreported since the publication of thelast reviews (i.e., 1960), earlierstudies will be mentioned when theirprocedures or results are particularlysalient.

The discussion will be organizedaround the major questions that havebeen addressed by researchers. Thesequestions are:

• Do schizophrenic patients differfrom normals, or other psychiatricpatients, in IQ?

• Is schizophrenia associated withpremorbid IQ deficits?

• Is schizophrenia associated witha pre- or post-onset decline in IQ?

• Are clinical improvement, lengthof hospitalization, or prognosticindicators related to premorbid orpostmorbid • IQ?

• Are there patterns of premorbid, or postmorbid subtest performance(e.g., large Verbal IQ-PerformanceIQ discrepancies, large intratestscatter) that characterize and/orpredict schizophrenia?

The issue of premorbid intellectualfunctions in schizophrenia is ofparticular interest because of itsrelevance to theories of etiology.

1 Throughout this review, the term"postmorbid IQ" will be used to refer toIQs obtained from tests administered afterthe onset of psychiatric disorder.

Moreover, the literature addressingthe question of premorbid IQ inschizophrenia lends itself to theapplication of meta-analytictechniques, due to both the numberof studies and their relative methodo-logical consistency. In this review,meta-analyses are applied to thesestudies in order to determinecombined probabilities and effectsizes (Rosenthal 1978). Quantitativetechniques for summarizing theresults of independent investigationsoffer a more direct approach to theresolution of controversies in thisimportant area of the literature.

Before beginning our review of theliterature, we should note that it willbe limited to those studies that haveused standardized intelligence tests,such as the Wechsler Adult Intelli-gence Scale (WAIS), the WechslerIntelligence Scale for Children(WISC), and standardized group testsof intelligence. While numerousstudies have examined theperformance of schizophrenics orchildren at high risk for schizo-phrenia on other tasks thatpresumably tap intellectualfunctioning (e.g., object-sortingtasks, Piagetian tasks, subjectivemeasurements of creativity, andschool performance), these measuresare too disparate to provide resultsthat can meaningfully be comparedto the results from standardized IQtests.

Several methodological issuesshould also be noted. First, the inves-tigations discussed here have beenconducted over a broad pertod oftime and in a variety of settings; as aresult, they employ diversediagnostic criteria. Because of thevolume of literature to be reviewed,a comparative analysis of the effectsof diagnostic criteria on the outcomesof the studies is not possible. It cangenerally be assumed, however, thatthe diagnostic criteria used in most of


Dow

nloaded from


432 SCHIZOPHRENIA BULLETIN

these studies (usually hospitaldiagnoses) roughly correspond to thebroader schizophrenia categorydefined by DSM-II (AmericanPsychiatric Association 1968) asopposed to the narrower criteria ofDSM-III (American PsychiatricAssociation 1980). Second, althoughmany of the investigations focusexclusively on adult-onset patients,others use mixed samples of early(childhood and adolescence) andadult-onset patients. Very few focusexclusively on early-onset patients.This review will present informationregarding age at onset of illness whenit is provided by the authors. A thirdissue is that many reports fail tomention the medication status ofpatient subjects. However, based onthe results of research on the effectsof neuroleptics on cognitive functions(Gilgash 1961; Spohn, Lacoursiereand Thompson 1977; Braff andSaccuzzo 1982), medication tends toenhance at least some aspects ofcognitive performance in psychoticpatients and would therefore tend toreduce, rather than inflate, patientdeficits. Finally, most of the studiesdiscussed here do not examine theeffects of sex on their dependentmeasures. This is unfortunatebecause, as we shall see, there isevidence that gender is related topremorbid IQ in schizophrenia.

Do Schizophrenic PatientsDiffer From Normals orOther Psychiatric Patientsin IQ? „

Not surprisingly, many of the earlieststudies of schizophrenia wereconducted to determine whetherpersons suffering from the disordermanifested deficits in IQ. Only a fewstudies published within the past twodecades have addressed this question.Before 1939, when the Wechsler-

Bellevue scale was introduced, themost commonly used test in this kindof research was the Stanford-Binet.Unfortunately, as Hunt and Cofer(1944) point out, the Stanford-Binetdid not provide adult norms.Therefore early research using thistest compared the IQs of mentalpatients with the norms derived fromgroups of children. Roe and Shakow(1942) attempted to overcome thisproblem by including a control groupof 65 surgical and orthopedicpatients. The average mental age ofthis group was 160-164 months, ascompared with an average IQ of140.6 months among 300 schizo-phrenic patients. A study by Kendigand Richmond (1940) reports asimilar average of 138 months for agroup of 500 schizophrenics. Otherearly investigations (Wells and Kelley1920; Michaels and Schilling 1936)reported higher Stanford-Binet IQsfor schizophrenics than the Roe andShakow (1942) and Kendig andRichmond (1940) studies, and stillothers (Pressey 1917; Cornell andLowden 1923; Wentworth 1924;Jastak 1937) reported lower averageIQs. However, previous reviewers(Hunt and Cofer 1944; Winder 1960)consider the Kendig and Richmond(1940) and Roe and Shakow (1942)samples as being the most repre-sentative of the general population ofschizophrenic patients. Thus, as Huntand Cofer (1944) conclude, resultsfrom the best-controlled of the earlystudies using the Stanford-Binetindicate that schizophrenic patientsinstitutionalized in state hospitalsshow an average intellectual deficitof approximately 20 months.

Following its introduction in 1944,the Wechsler-Bellevue Scale was thepreferred measure in many studies ofschizophrenia. The majority of theseinvestigations were not, however,aimed at determining whether theaverage IQ of schizophrenic patients

was different from that of the generalpopulation. Instead, they wereinterested in examining patterns ofsubtest scores in order to determinewhether such patterns coulddistinguish between schizophrenicand other mental patients, andamong various subtypes of schizo-phrenia. Payne (1960) reviewed 28studies which reported Wechsler-Bellevue scores for groups ofpsychiatric patients with variousdisorders. For the 1,284 schizo-phrenic patients tested in thesestudies, Payne calculated an averageIQ of 96.08. Payne notes, however,"it is almost certainly not the case"that the groups of schizophrenicpatients tested in these studies wererepresentative of the general schizo-phrenic population. From the 14studies which reported demographicdata for schizophrenics, Paynedetermined that the patients sampledby these studies were considerablyyounger than the average schizo-phrenic patient. He also speculatedthat the more uncooperative andunresponsive psychotic patients werenot included in some of the studieshe reviewed, thus making thesamples even more unrepresentative.Payne therefore suggests that theaverage IQ of 96.08, derived fromstudies that tested schizophrenics onthe Wechsler-Bellevue, is probablyhigher than would be obtained froma randomly selected sample. Winder(1960) speculates that the results ofseveral studies using the Wechslertests are indicative of an averagedeficit of 10 IQ points.

Thus, early investigations using theStanford-Binet and the Wechsler testssuggest that schizophrenic patientsperform at least slightly belowaverage on tests of intelligence.Although these early studies failed tocompare schizophrenic patients tonormal controls matched on socio-economic status and geographic


Dow

nloaded from


VOL. 10, NO. 3,1984 433

origin, recent studies that usematched controls draw the sameconclusion (Pollack, Woerner, andKlein 1970; Lehman, Chelune, andHeaton 1979). In the study ofPollack, Woerner, and Klein (1970),scores obtained from schizophrenicpatients on the WISC or WAISduring the fourth week of hospi-talization were compared to scoresobtained from their siblings at thesame time. The average Full Scale IQscore for 27 schizophrenic patientswas 106.6; the average score for theirclosest-in-age normal sibling was114.7. Thus, while scores for boththe schizophrenic patients and theirsiblings were well within the normalrange, the schizophrenic patientsscored significantly lower. Theresults of the study of Pollack,Woerner, and Klein (1970), andearlier studies, lead us to concludethat the intellectual performance ofdiagnosed schizophrenic patients islower than would be predicted fromfamily and environmental variables.

There have been relatively fewwell-controlled studies of IQ inpatients with various psychiatricdiagnoses. The limited data that areavailable suggest that adult schizo-phrenic patients do not show deficitsin IQ when compared to patientswith other psychotic disorders(Payne 1960; Cohler et al. 1977), orpersonality disorders (Pollack,Woerner, and Klein 1970), but theydo manifest deficits when comparedto alcoholic and neurotic patients(Payne 1960; Holland, Levi, andWatson 1979). Schizophrenic childrenhave been found to score below bothneurotic and personality-disorderedchildren on the WISC (Schoonoverand Hertel 1970).

Given the pervasive cognitive,affective, and motoric symptomsfrequently associated with adiagnosis of schizophrenia, it is notsurprising or particularly illuminating

that these patients manifest meanperformance deficits on standardizedtests of intelligence. Recent investi-gations have been more concernedwith the etiological and prognosticimplications of intelligence forschizophrenia. Thus, researchershave focused upon premorbid intel-lectual development and postmorbidcorrelates of IQ.

Is Schizophrenia AssociatedWith Premorbid IQ Deficits?

A number of investigators havesought to determine whether lowerIQ occurs as a precursor to schizo-phrenia, or occurs only concomi-tantly with the overt expression ofthe illness. Most of these are retro-spective studies which comparepremorbid IQs of schizophrenicpatients with IQs of their peers,schoolmates, or siblings. In additionto these studies, results from some ofthe longitudinal prospective investi-gations of children at risk for schizo-phrenia (offspring of schizophrenicparents) are just beginning to becomeavailable.

Retrospective Studies. Becauseregularly scheduled IQ testing andsystematic record keeping in theschools are a fairly recent devel-opment, early investigations thatsought to determine the relationshipbetween premorbid IQ and schizo-phrenia looked at premorbid scoresfrom army intelligence tests or fromchild guidance clinic records. Both ofthese methodologies are limited bythe nonrepresentativeness of theirpreschizophrenic samples; yet, someof the studies yielded results that areconsistent with those that use morerepresentative samples. Based onArmy induction test scores, itappears that preschizophrenic malesdo show IQ deficits in young

adulthood. Mason (1956), forexample, compared premorbid scoreson the Army General ClassificationTest for 368 schizophrenic patients,188 nonschizophrenic psychiatricpatients, and 290 army inducteeswho had experienced no mentaldisorders. The comparison revealedthat the schizophrenic patients as agroup had scored significantly lowerthan the controls, while nonschizo-phrenic patients did not differ signif-icantly from the controls onpremorbid IQ. Similarly, Miner andAnderson (1958) found that men whowere discharged from military servicebecause of psychotic illness hadobtained lower Army GeneralClassification Test scores at the timeof induction than men who experi-enced no later disorder. No reliabledifferences were found, however,between premorbid scores ofpsychotics and psychoneurotics. Theone study that failed to findpremorbid IQ deficits in military menwho later developed schizophrenia isreported by Schwartzman, Douglas,and Muir (1962); however, theirsample included only 23 patients.

Studies of child guidance clinicpopulations have yielded contra-dictory results. Frazee (1953)compared boys who had beenreferred to a child guidance clinicbetween the ages of 5 and 16, andwho later became schizophrenic, withboys from the same clinic whoexperienced no later psychiatricdisorder. She found a mean IQ of88.7 for the children who laterbecame schizophrenic, comparedwith a significantly higher mean of97.7 for those who experienced nolater disorder. In a similar study byBirren (1944), scores from childguidance clinics were obtained for 38psychiatric patients and 53 nondis-turbed adult controls (gender notindicated). Unlike Frazee (1953),Birren found that the schizophrenic


Dow

nloaded from



patients had significantly higherchildhood IQs than the controls andother psychiatric patients (whosedisorders included brain disease,paresis, severe mental deficiency withpsychosis, and psychoneurosis). It isinteresting to note, however, that theaverage premorbid IQ of the schizo-phrenic patients in Birren's study(87.5) was quite similar to that foundby Frazee (88.7). The largediscrepancy in control group means(80.8 for Birren vs. 97.7 for Frazee)may be explained by the types ofproblems exhibited by the twoguidance center populations. In theBirren study almost all children werereferred to the clinic for "schoolretardation or backwardness," whilemost of the children in the Frazeestudy were referred for antisocial,delinquent, or acting-out behavior.

As schools began routinely admin-istering group IQ tests, and keepingrecords of scores, it became possiblefor researchers to compare premorbidscores of schizophrenic patients withchildhood IQ scores of normal peersand siblings. Because scores wereavailable for children who were morerepresentative of the generalpopulation of school children thanthe samples obtained from childguidance clinics, the results of theselater studies are much easier tointerpret. The mean IQs yielded bythe studies are listed in table 1. In theOfford and Cross (1971), Offord(1974), and Watt and Lubensky(1976) studies, control subjects areclassroom peers matched with thepreschizophrenic children on age,sex, and socioeconomic status. Thereports by Lane, Albee, andcolleagues (Albee, Lane, and Reuter1964; Lane and Albee 1968) use themean IQ for all children in the samegrade and school district as the basisfor comparison. Bower, Shellhamer,and Daily (1960) used a randomlyselected control group of classroom

peers. As indicated in the table, fivestudies derived an average IQ foreach subject based upon multipletests administered over several gradelevels. In cases where preschizo-phrenic children had more than onesibling for whom IQ data wereavailable, Watt and Lubensky (1976)and the Lane and Albee groupderived an average of the siblings' IQscores and considered each set ofsiblings as a single observation.Offord (1974) and Offord and Cross(1971) treated each sibling as aseparate observation, and computedthe average IQ across siblings.Pollack, Woerner, and Klein (1970)used only the nearest-in-age sibling.Finally, it should be noted that someof the articles report data on samplesthat contain subjects from previouspublications. These overlappingsamples are noted in the table.

When male and female preschizo-phrenic children were combined, allstudies found significantly lower IQsfor preschizophrenic childrencompared to controls. However,when the sexes were consideredseparately, Watt and Lubensky(1976) failed to find differences formales with an IQ measure from theearly elementary years, and forfemales with an IQ measure fromhigh school. Yet the studies byOfford (1974) and Bower, Shell-hamer, and Daily (1960) includedlarger numbers of subjects, and bothreport significantly lower IQs formale and female preschizophrenicchildren when compared to same-sexcontrols. Taken together, the resultsof these investigations suggest thatpreschizophrenic children do manifestIQ deficits, whether contrasted withrandomly selected or matched-peercontrols.

A statistical summary of the resultsof the four nonoverlapping studiesthat compared preschizophrenic topeer control children is presented in

table 2. Two methods were used forcombining probabilities acrossstudies; the method of adding proba-bilities (p = (lp)N/N\) and themethod of adding unweighted "Z"s(Z = lZ/NVi)(see Rosenthal 1978,for a comparative analysis of variousmethods of combining probabilities).These methods were selected becauseof their appropriateness with smallnumbers of studies. Cohen's (1977)procedure was used to determine themean effect size. Cohen's d gives thesize of the effect defined as theproportion of the pooled standarddeviation estimate from the twosamples being compared.2 As table 2illustrates, the magnitude of theprobability across studies indicatesthat there is a real difference betweenthe IQs of preschizophrenics andtheir peers.

The results of comparisons ofpreschizophrenic children and theirsiblings are less consistent than thosecontrasting preschizophrenic childrenand peer controls. Examination ofthe means in table 1 reveals that,with the exception of females in theOfford (1974) study, all mean IQsfor preschizophrenic children arebelow those of their siblings.However, the differences betweenthese means fail to reach statisticalsignificance in 6 out of 16 compar-isons. Pollack, Woerner, and Klein(1970) and Watt and Lubensky (1976)found no differences between mixed-sex groups of preschizophrenicchildren and their siblings in IQmeasured from the third througheighth grades. Similarly, Schaffner,Lane, and Albee (1967) find nodifference with a measure of IQtaken from the fourth through sixthgrades. Offord (1974) finds nodifference for preschizophrenic

' Effect sizes were also determined usingthe procedure suggested by Glass (1980),and the results were comparable.


Dow

nloaded from


VOL. 10, NO. 3,1984 435

Table 1. Mean IQs oi

Study

MalesBower, Shellhamer, &

Daily (1960)

Offord (1974)

Watt & Lubensky(1976)

FemalesOfford (1974)


; preschizophrenic children, their siblings,

Grade level Preschizophrenicwhen tested

High school

Through 9th grade

Kindergarten-6th grade

7th-12th grade

Through 9th grade

Kindergarten-6th grade

7th-12th grade

Male and female subjects combinedAlbee, Lane, &

Reuter(1964)

Lane & Albee (1964)(subgroup of Albeeet al. 1964)

Lane & Albee (1965)(Subgroup of Albeeet al. 1964—preschizophrenicswith siblings)

Lane & Albee (1968)(Subgroup of Lane& Albee 1965-subjects who weretested in both

2nd grade

6th grade

8th grade

2nd grade

2nd grade

6th grade

8th grade

2nd grade

6th grade

IQ(SD)1

99.3(_)(n = 44)

88.2(17.5)(n = 51)

104.25(_)(n = 16)100.56(_)(n = 16)

97.1(13.2)(n = 65)

106.56(_)(n = 18)106.28(_)(n = 18)

93.9(11.5)(n = 122)89.4(15.0)(n = 154)90.6(16.2)(n = 103)

91.7(8.8)(n = 36)

91.6(_)(n = 55)

86.5(_)(n = 60)

89(_)(n = 42)

87.8(14.9)(n = 41)

87.9(11.6)(n = 41)

and peer controls

SiblingIQ(SD)

—

95.9(17.9)'(n = 50)

(male sibs only)

—

—

97.0(12.4)(n = 63)

(female sibs only)

—

—

—

—

—

100.8(9.1)2

(n = 36)

99.0(_)2

(n = 55). 94.2(_)2

(n = 60)96.2(_)J

(n = 42)

95.9(11.8)2

(n = 41)

92.3(10)(n = 41)

ControlIQ(SD)

106.3(_)J

(n = 44)

100.6(11.9)2

(n = 51)

106.38(_)(n = 16)

107.00(_)2

(n = 16)

101.8(14.7)2

(n = 65)

111.83(_)2

(n = 18)108.83(_)(n = 18)

100.8(10.1)2

(n = 2613)99.4(15.7)2

(n = 4166)99.6(17.2)2

(n = 4960)

—

100.9(9.8)2

(n = 4597)

98.9(18.0)2

(n = 4166)2nd and6th grade)


Dow

nloaded from



Table

Study

1. Mean IQs of preschizophrenic

Grade levelwhen tested

children, their siblings, and

PreschizophrenicIQ(SD)'

peer controls—Continued

SiblingIQ(SD)

ControlIQ(SD)

Schaffner, Lane, &Albee (1967)3

(high SES, suburbansample)

Pollack, Woerner, &Klein (1970)

Offord & Cross (1971)(Subgroup of Offord1974)

Watt & Lubensky(1976) (IQs of maleand femalepreschizophrenicscombined, and theirsibs and controls)

Kindergarten-3rd grade

4th-6th grade

7th-9th grade

Kindergarten-2nd grade

3rd-5th grade

6th-9th grade

Through 9th grade

3rd-8th grade

106.6(17.1)(n = 25)

107.6(20.5)(n = 26)

107.7(16.3)(n = 32)

104.4(16.4)(n = 15)

105.5(17.1)(n = 14)

113.4(15.2)(n = 11)

86.6(20)(n = 29)

102.86(_)(n = 50)

111(16)2

(n = 25)111.1(14.5)(n = 26)

114.3(16.8)2

(n = 32)

114.5(16.0)'(n = 15)

114.6(10.6)(n = 14)

114.9(13.1)(n = 11)

94.3(16.4)2

(n = 77)

—

(all preschizophrenics)102.91(_)(n = 33)

105.27(_)(n = 33)

(preschizophrenics with siblings)

—

—

—

—

—

—

98.5(13)2

(n = 87)

107.66(_)J

(n = 50)

—

1 ( ) = Standard deviation not listed by the authors.1 Significantly higher than mean IQ for preschizophrenics.1 The authors list mean 10 scores and standard deviations, but conduct their comparisons with transformed scores.

females and their siblings, but doesfind a difference for males.

The statistical summary of thenonoverlapping studies of preschizo-phrenic children and their siblings ispresented in table 3. Again, both themethod of adding probabilities andthe method of adding "Z"s yield asignificant effect across studies.Although the combined effect size of.55 is somewhat greater than theeffect size of .43 resulting fromcomparisons of preschizophrenicchildren and peer controls, thedifference in effect sizes is not statis-tically significant.

Offord (1974) has speculated onfactors that may be contributing tothe lack of consistent significant

findings across studies of preschizo-phrenic-sibling differences in IQ. Hismajor criticism of the early studiesby the Lane and Albee group is thatthese researchers excluded from theiranalyses the preschizophrenicchildren and siblings who were inspecial classes in school. Since thesiblings' IQ was calculated byaveraging the IQs of only thosesiblings who attended regular classes,it is quite possible that many siblingswith IQs lower than those of thepreschizophrenic children wereexcluded from the sample, thusraising the sibling IQ measure, andenhancing preschizophrenic-siblingIQ differences. The Offord (1974)study, on the other hand, included

all probands and siblings whoattended the school system, whetheror not in special classes.

Offord (1974) offers a similarcriticism of the study of Pollack,Woerner, and Klein (1970), notingthat these investigators compared thepreschizophrenic with his or hernearest-in-age normal sibling only. Anormal sibling was one who wasjudged free of current symptoma-tology, beyond a mild degree, andwho had no history of psychiatrictreatment or impairment of functionthat would have warrantedtreatment. Offord (1974), in contrast,compared each patient with all of hisor her siblings for whom data wereavailable, and did not exclude any of


Dow

nloaded from


VOL.10, NO. 3,1984 437

Table 2. Statistical summary: Preschlzophrenics vs. school controls

StudyGrade levelwhen tested di One-tailed p Effect size (d)

Bower, Schellhamer, &Daily (1960)

Albee, Lane, &Reuter(1964)


Of ford (1974)

High school(all males)2nd grade6th grade8th grade

K-6th gradeFemalesMales7-12th gradeFemalesMales

Through 9th gradeFemalesMales

2.10

7.337.745.26

1.86.89

.891.98

1.924.18

Sp*P

P

86

273343185061

3430

3430

128100

= .722= .072= .0000

.0182

.0000

.0000

.0000

.0341

.3078

.3083

.0270

.0264

.0002

Id =Xd =

4.3.43<3>

.45

.28

.24

.15

.64

.33

.31

.72

.34

.84

1 2 =XZ =

Z =(P<

2.09

6.406.504.90

1.82.50

.501.93

1.943.60

30.183.029.55<4>

.000)

' Values were recomputed and, If necessary, corrected when sufficient Information was available.' P = (X p) N/N!1 d = (X, - X.)/S«Z = ZZ/NVi

the siblings for reasons of poormental health.

In summary then, Offord arguesthat when preschizophrenic childrenare compared with all of their .siblings on IQ, no significant differ-ences are found. But when preschizo-phrenic children are compared withthose siblings who show no mentalretardation or behavioral disorder,the preschizophrenic children doshow IQ deficits. However, thisexplanation was not sufficient toaccount for the Offord and Cross(1971) finding of significantpreschizophrenic-sibling differences,or the Offord (1974) finding thatpreschizophrenic males from familieswith an average IQ of 95 or lessscore lower than their siblings, whilethose from families with higher IQs

do not. In neither of these studieswere subjects excluded because ofspecial class assignment or poormental health.

In order to explain these findings,Offord (1974) proposed a hypothesisbased on the assumption that bothIQ and the predisposition to schizo-phrenia are largely genetic, butindependently inherited, traits. Ahigh IQ, Offord suggests, can protecta predisposed person from becomingovertly schizophrenic, while a lowIQ offers no such protection. Conse-quently, if the predisposition for-schizophrenia is only slight, overtexpression of schizophrenia willoccur only in families of low IQ. Onthe other hand, in a high-IQ familywith a schizophrenic child, thepredisposition for schizophrenia must

be quite severe—if it were not, noneof the children would show overtsigns of the illness. Thus, Offordhypothesizes that IQ level acts as amediating factor in the overtexpression of schizophrenia, and iscrucial only in the low-IQ families.In addition, Offord proposes that thelack of preschizophrenic-siblingdifferences among females from low-IQ families may be due to anincreased vulnerability to schizo-phrenia among low-IQ males. Hesuggests that this increased vulner-ability may be mediated by greatersusceptibility to central nervoussystem damage in male offspring.The question of sex differences inpremorbid IQ will be addressed ingreater detail later in this section.

Offord's (1974) hypothesis


Dow

nloaded from



Table 3i. Statistical summary: Comparisons of preschizophrenics with their siblings1

StudyGrade levelwhen tested df One-tailed p Effect size

Lane & Albee (1965)


Of ford (1974)

Pollack, Woerner, &Klein (1970)

2nd grade6th grade8th grade3rd-8th

Through 9th gradeMalesFemales

K-2nd grade3rd-5th grade6th-8th grade

6.734.813.13

.89

2.19-.04

1.711.69.25

IPXp

P

CO

CO

C

MO

T-

CO

64

99126

CO

CO

O

CM

CM

C

M

= 1.34= .15= .000

.0000

.0000

.0014

.3092

.0145

.5165

.0474

.0497

.4009

Id =Xd =

4.96.55

1.30.89.69

.22

.44

.00

.65

.66

.11

IZ =xz =

z =

5.004.002.98

.50

2.18- . 0 4

1.671.64

.25

18.182.026.06

: .000)1 The Schaffner, Lane, & Albee (1967) results are not included in the statistical summary because the analyses were conducted ontransformed scores that are not published in their report; thus, it was not possible to derive an effect size.

regarding familial IQ level isgenerally compatible with the resultsof other studies that have looked forsibling-preschizophrenic IQ differ-ences. For example, the studies byLane and Albee (1964, 1965, 1968),which compared IQ scores of siblingsand pfeschizophrenics wereconducted in urban neighborhoods oflower socioeconomic status. In thesestudies (all of which found sibling-preschizophrenic differences) thepreschizophrenic children's mean IQwas below 95 across grade levels,and IQs were generally in the high80's. The siblings' mean IQ wasbelow 101 at all grade levels, andIQs generally fell in the mid-90's.Similarly, the study by Offord andCross (1971), which likewise foundsibling-preschizophrenic differences,reports a mean IQ of 86.6 for thepreschizophrenic children and 94.3for the siblings. In contrast, the Wattand Lubensky (1976) study, whichfound no preschizophrenic-sibling

differences, reports a mean of 102.91for preschizophrenic children, and105.27 for siblings. The average IQsfor the preschizophrenic children inthe study of Pollack, Woerner, andKlein (1970) were 104, 105, and 113at the K-2, 3-5, and 6-8 gradelevels, respectively, and were 114.5,114.6, and 114.9 for the siblings atthe three grade levels. (Significantsibling-preschizophrenic differenceswere found only for K-2 test scores.)Thus, in each of these studies,samples with lower than averagemean IQs showed sibling-preschizo-phrenic differences, while thosesamples with higher than averagemean IQs generally showed no suchdifferences. The only study whichprovides results contradictory to thisconclusion is the study of Schaffner,Lane, and Albee (1967), which foundpreschizophrenic-sibling differences insamples from middle and upper classsuburbs. The average IQs forpreschizophrenic children were

between 106.6 and 107.7 (at threegrade levels); the IQs for siblingswere between 111 and 114.3.

For a more direct evaluation of thehypothesis that preschizophrenicsfrom low IQ families are more likelyto manifest IQ deficits relative totheir siblings, we derived the coeffi-cient of correlation (Pearson r)between mean sibling IQs and theeffect sizes listed in table 3. Theresultant coefficient of -.22 does notreach statistical significance, but it isin the direction predicted by Offord'shypothesis. Also, it should be notedthat this negative coefficient is in thedirection opposite to what would bepredicted by regression toward themean.

There are alternative approaches tointerpreting the above results. Longi-tudinal studies of children whosuffered perinatal complications ormanifested early developmentaldelays indicate that familial socio-economic status (SES) is a significant


Dow

nloaded from


VOL. 10, NO. 3,1984 439

determinant of outcome (Drillien1964; Davie, Butler, and Goldstein1972; Rubin and Balow 1979). Whensuch children have been reared inlower SES homes by parents withlimited educational backgrounds,signs of cognitive and neuromotordeficits are more likely to persist. Incontrast, at-risk infants born tomiddle and upper SES families tendto improve and, as a group, areoften indistinguishable from thenorm by middle childhood. It ispossible that the genotype for schizo-phrenia has pleiotropic effects thatcan be manifested in intellectualdeficits, but only when the child isreared in a lower SES home wherethe parents have limited educationalbackgrounds. Thus, standardized IQtests may be tapping, at leastpartially, cognitive deficits associatedwith a genetic predisposition toschizophrenia. The apparent inter-action between preschizophrenic IQand familial IQ may therefore berelated to the ameliorative effects ofmiddle/upper SES child-rearingpractices, rather than the protectiveeffects of an independently trans-mitted genetic predisposition to highIQ. On the other hand, it may bethat acquired subclinical braindamage is the primary etiologicfactor for a subgroup of schizo-phrenic patients. Recent researchusing the positive/negative symptomdistinction suggests that schizo-phrenia characterized by negativesymptoms is associated with greatercognitive deficit and evidence ofstructural abnormalities in the centralnervous system (Andreasen andOlsen 1982). Thus, a subgroup ofpieschizophrenic children may besuffering from an acquired, ratherthan inherited, organic impairmentthat is manifested in low IQ whenthe individual is reared in a low SESenvironment. Additional researchwill be necessary to evaluate the

hypothesis that preschizophrenicchildren have lower IQs than theirsiblings only if they are from low-IQfamilies. Such research may provideimportant information regardinggenetic determinants of both IQ andschizophrenia, and how these deter-minants interact.

Regardless of the validity of theproposed explanations for thediscrepancy in results onpreschizophrenic-sibling IQ differ-ences, two important conclusions canbe drawn from this research. First, itis important to note that while somestudies showed the preschizophrenicsto have lower childhood IQs thantheir siblings, and other studiesshowed no preschizophrenic-siblingIQ differences, none of the studiesfound evidence for higher IQs amongthe preschizophrenic children. Thisobservation, along with strongevidence for preschizophrenic-controldifferences, suggests that it is notunreasonable to assume that lowchildhood IQ is associated with riskfor schizophrenia. Consistent withthis conclusion, Pollin and Stabenau(1968) found that for 20 monozygotictwin pairs discordant for schizo-phrenia, the schizophrenic twin had alower premorbid IQ four times asoften as the nonschizophrenicco-twin.

Furthermore, in those studieswhich included both sibling andcontrol comparisons (Lane and Albee1968; Offord and Cross 1971; Offord1974; Watt and Lubensky 1976), thematched controls always scoredhigher than the siblings of thepreschizophrenics, as well as thepreschizophrenic children themselves.However, the authors of these studiesdid not conduct statistical tests of thedifference between the IQs of siblingsand controls. Three of the reports intable 1 provide mean IQs andstandard deviations for sibling andcontrol groups. We conducted f-test

comparisons of these means andfound significant differences (one-tailed) in four comparisons and atrend toward significance in onecomparison (Offord and Cross1971: t = 1.83, p = .033; Offord1974: for males, t = 1.55, p = .06;for females, t = 1.99, p = .023;Lane and Albee 1968: 2nd grade,f = 3.25, p = .000; 6th grade,f = 2.34, p = .009). Thus, at leastmoderate IQ deficits may be presentin the siblings of preschizophrenics,regardless of their clinical outcome,when compared to children whoattended the same school. This resultis consistent with the notion that thegenotype for schizophrenia, althoughnot always manifested in the schizo-phrenic phenotype, may have impli-cations for cognitive development.On the other hand, it is also possiblethat exposure to environmentalstressors (i.e, exposure to thepreschizophrenic sibling or familialinstability) is producing IQ deficits inthe siblings of premorbid subjects.

Prospective Studies. Anotherresearch approach that holds promisefor providing information aboutpremorbid IQ is the longitudinalhigh-risk method. In this paradigm,children are considered to be at highrisk for schizophrenia if they have atleast one parent who has beendiagnosed as schizophrenic. For mostof the studies discussed here, thehigh-risk children have only oneschizophrenic parent, usually themother. While the types of variablesexamined in high-risk researchprojects vary greatly, almost all haveincluded some measure of IQ.Because the majority of these projectshave not yet followed their subjectsthrough the major risk period forschizophrenia, most of the reportsemanating from them rely oncomparisons between all high-riskchildren and various control groups.


Dow

nloaded from



However, it should be noted thatonly 10-15 percent of subjects withone biological schizophrenic parentare expected to manifest the disorder(Gottesman and Shields 1982).Consequently, group comparisonsare limited in the information thatthey can offer regarding premorbidindicators.

The mean IQs reported in the high-risk studies are listed in table 4. Tworeports not listed in the table statedthat they found no significant IQdeficits in high-risk children whencompared to controls; however, theauthors did not report mean scores(Hanson, Gottesman, and Heston1976; Grunebaum et al. 1978). Aninvestigation by Landau et al. (1972)is also not listed in table 4 becausethe authors did not report meanscores; however this study did revealsignificant IQ deficits in high-riskchildren. Another report showing IQdeficits in high-risk children (Winterset al. 1981) is not listed because itpresented only raw scores from asubset of Verbal and Performancesubtests.

A statistical summary of thenonoverlapping high-risk studies thatprovided adequate information forderiving effect sizes is presented intable 5. Because psychiatric controlgroups were included in only a smallnumber of studies, and the compo-sition of these groups is hetero-geneous across studies, table 5 isrestricted to comparisons of high-riskchildren and matched offspring ofnormal parents. Both methods ofderiving combined probabilitiesyielded statistically significantresults, and the estimated effect sizeacross studies is .39. Thus it appearsthat offspring of schizophrenicparents do score lower than childrenof normal parents; however, themean effect size is not so large asthose yielded by comparisons ofpreschizophrenic children with their

siblings and peer controls.When IQ scores of the offspring of

schizophrenic parents are comparedto those of children whose parentshave affective disorders, noconsistent differences are found.Winters et al. (1981) administeredtwo Verbal and two Performancesubtests of the WISC, and reportedthat offspring of unipolar depressivepatients, as well as offspring ofschizophrenic patients, had signif-icantly lower Verbal subtest scoresthan children of bipolar patients, andnormal controls. However,performance subtest scores werelower in children of unipolardepressive patients when comparedto offspring of bipolar patients,schizophrenic patients, and normalcontrols. Comparisons of the fullscale WISC IQs of the offspring ofschizophrenic patients with those ofthe offspring of manic-depressive(Worland and Hesselbrock 1980) andunipolar depressive (Oltmanns et al.1978) patients revealed no significantdifferences. Similarly, Worland et al.(1982) found no group differences inVerbal or Performance IQ whencomparing offspring of patients withschizophrenia, schizoaffectivedisorder, and affective disorder;however, Performance IQs of theoffspring of schizophrenic patientswere less stable across time thanthose of the other groups. Finally, aninvestigation by Cohler et al. (1977)showed lower full scale IQs forchildren of psychotic depressedmothers when compared withchildren of schizophrenic and normalmothers. In summary, then, itappears that IQ deficits are notunique to children at risk for schizo-phrenia, but instead may be charac-teristic of the offspring of parentswith a variety of psychotic disorders.

An examination of correlates of IQby Rieder, Broman, and Rosenthal(1977) indicated that perinatal

complications were significantlyassociated with IQ deficits in high-risk offspring, but not offspring ofnormal mothers. In contrast, SESwas positively correlated with IQ foroffspring of normals, but not foroffspring of patients with schizo-phrenia. The authors interpret theseresults as suggesting that "one aspectof schizophrenic inheritance is asusceptibility to other factors thatmay lower IQ" (p. 799). The ideathat genetic vulnerability mayinteract with external physicalstressors in the development ofschizophrenia is explored further in arecent article by Walker and Emory(1983). This idea offers yet anotherapproach to the general question ofthe nature of the relationshipbetween IQ and schizophrenia.

The high-risk studies thatexamined the relationship betweenparent and offspring IQs yieldedinconsistent results. Lane and Albee(1970) found no significantrelationship between the childhoodIQs of adult schizophrenics and theIQs of their offspring. However,childhood IQs of the nonschizo-phrenic parents were significantlycorrelated with their offspring's IQs.In contrast, Worland and Hessel-brock (1980) found that the IQs ofhigh-risk children showed a strongerrelationship with the IQs of theirschizophrenic parents than with thoseof their nondisturbed parents,although parent/child IQ correlationswere somewhat lower overall forhigh-risk children when compared tooffspring of normals and otherpsychiatric patients.

To date, only two longitudinalhigh-risk research projects have •reported psychiatric followups oftheir subjects. The New York high-risk project (Erlenmeyer-Kimling etal. 1980) has examined 80 children ofschizophrenic parents and 125children of parents with nonschizo-


Dow

nloaded from


VOL. 10, NO. 3,1984 441

Table 4. Mean IQs of offspring of parents with and without psychiatric disorders

StudyAge level

when testedOffspring of

schizophrenic parentsOffspring of

normals

Offspring ofparents with

other disorders

Cohleret al. (1977)

Rieder, Broman, &Rosenthal(1977)

Mean age 5V2 years

Males: 109.7(16.2)(n = 14)

Females: 105.6(10.3)(n = 10)

110.2(12.5)(n = 18)106.5(11)(n = 15)

7 years Offspring of chronic patients:Males: 101.5(11.7) 107.6(12.1)'

(n = 30) {n = 30)Females: 103.3(14.3) 102.6(12.3)

(n = 15) (n = 15)Offspring of acute patients:

Males: 101(4.9) 114.5(10.1)'{n = 6) (n = 6)

Females: 110.3(11.4) 102.1(16.8)(n = 9) (n = 9)

Depressed101.0(21.9)

(n = 5)92.3(22)(n = 5)

Oltmanns et al.(1978)

Worland &Hesselbrock(1980)

Griffith et al.(1980)

Watt, Grubb, &Erlenmeyer-Kimling(1982)

Worland et al.(1982)(Subgroup ofWorland &Hesselbrock1980)

6-15 years

6-20 years

13-21 years

7-12 years

Mean age 7.9 years

Mean age 15.6 years

94.32(24.18)(n = 156)

102.81(13.24)(n = 94)

100.7(_)(n = 207)

104.87(11.20)(n = 44)

112.45( )(n = 10)

107.30(_)(n = 10)

102.57(21.32)'(n = 139)

108.95(15.98){n = 119)

104.3(_)'(n = 104)

117.66(11.90)'(n = 70)

110.53( )(n = 75)

109.74(__)(n = 75)

Depressed99.12(22)

(n = 102)

Manic-depressive106.56(16.93)

(n = 53)Physically ill98.13(16.28)

{n = 73)—

—

Affective disorder103.93(_)(n = 20)

Schizoaffective disor101.86(_)(n = 17)

Physically ill104.52(_)(n = 31)

Affective disorder100.01 (_)(n = 20)


Dow

nloaded from



Table 4. Mean IQs of offspring of parents with and without psychiatric disorders— Continued

Study

Offspring ofAge level Offspring of Offspring of parents with

when tested schizophrenic parents normals other disorders

Schizoaffective disorder

(n = 17)Physically ill

99.81(_)(n = 31)

1 Significantly higher than the mean IQ for high-risk children.

phrenic psychiatric disorders or nomental disorder. Initial admin-istration of the WISC occurred whenthe children were between 7 and 12years. Ten years later, at the time ofthe most recent report, five of thesubjects had been hospitalized forpsychiatric treatment. The meanWISC IQ of the five high-risksubjects who were hospitalized was11 points below the mean for theentire high-risk group during initialtesting, a statistically significantdifference. In contrast, the Danishhigh-risk study (Griffith et al. 1980)found no premorbid IQ deficits. Thesubjects of the Danish study are 207offspring of schizophrenic mothers.Initial testing on the WISC and othermeasures occurred in 1962, whensubjects were adolescents. Followuppsychiatric evaluations occurred in1972. At the time of the 1972followup nine male and eight femalehigh-risk subjects had developedschizophrenia. When high-risksubjects who became schizophrenicwere compared with those who didnot, no significant premorbid IQdifferences were found (Griffith et al.1980).

Thus, initial findings from the twohigh-risk studies that have conductedfollowup psychiatric assessments areinconsistent. However, the subjects inthese studies have not yet passed

through the age period of greatestrisk for schizophrenia: As theseongoing research projects completeadditional followup evaluations oftheir high-risk samples, the resultswill shed greater light on thequestion of premorbid IQ. At thispoint, the findings of high-riskresearch provide evidence of IQdeficits in the offsprings of schizo-phrenic patients, when compared tothe offspring of normals. However,IQ deficits are also sometimes foundin children of parents with otherpsychotic disorders and, as statedearlier, Pollack, Woerner, and Klein(1970) report that patients whoreceive a diagnosis of personalitydisorder in adulthood score belowtheir siblings on childhood measuresof IQ. Thus, the available evidencesuggests that premorbid IQ deficitsare not unique to schizophrenia.

Sex Differences in Premorbid IQ.Before attention is turned to adiscussion of pre- and post-onsetdecline in IQ, it should be noted thatstudies of preschizophrenic and high-risk children both suggest lower IQsin males. Offord (1974) found signi-ficantly lower IQs in male comparedto female preschizophrenic children.A difference in the same direction isindicated by Watt and Lubensky(1976), but they did not conduct a

statistical test of the significance ofthe difference. Similarly, Rieder,Broman, and Rosenthal (1977) foundthat only male offspring of schizo-phrenic patients showed IQ deficitswhen compared to controls, andLane, Albee, and Doll (1970)reported that male, but not female,offspring of schizophrenic patientsshowed a trend toward lower IQscores than their parents. Todetermine whether differences in theproportion of males included in thesamples of preschizophrenic andhigh-risk children contributed tovariability in effect sizes, coefficientsof correlation (Pearson r) werederived between the proportion ofmales in the samples and effect size.The correlation between theproportion of males in the samples ofpreschizophrenic children and theeffect sizes for comparisons with peercontrols (table 2) is .14—positive,but not statistically significant.However, for comparisons ofpreschizophrenic children and theirsiblings (table 3), the coefficient is.57 (p < .05). The correlationbetween the proportion of males inthe high-risk samples and the effectsize for comparisons with matchedcontrols (offspring of normals; table5) is .62 (p < .025). Thus, it seemsthat males are contributing morethan females to the size of the


Dow

nloaded from


VOL. 10, NO. 3,1984 443

Table 5. Statistical summary: Comparisons of

•Study

Rieder, Broman, &Rosenthal (1977)

Cohler et al.(1977)

Oltmanns et al(1978)

Worland & Hesselbrock(1980)

Griffith et al.(1980)

Watt, Grubb, &Erlenmeyer-Kimling(1982)

Offspring ofchronics:

MaleFemale

Offspring ofacutes:

MaleFemale

MaleFemale

high-

t

1.99-.144

2.951.21

.10

.213.11

2.99

2.00

5.71

I PXp

P

risk children with offspring of

df

5828

1016

3023

293

211

309

112

= 1.84= .184= .000

One-tailed

normals

p value Effect size (d)

.0242

.9211

.0071

.8788

.4589

.4148

.0012

.0017

.0218

.000

I d = 3.94Xd = .39

.52- .05

1.87- .61

.04

.09

.36

.41

.23

1.08

z

1.97-1.41

2.45-1.17

.10

.21

3.04

2.93

2.02

5.18

I Z = 13.91XZ =

Z =<P<

1.394.40

.000)

measured IQ deficit in the latter twocomparisons. It is possible, therefore,that female preschizophrenic childreneither do not manifest IQ deficitsrelative to their siblings or manifestlesser deficits than males. Similarly,female offspring of schizophrenicpatients may not differ significantlyin IQ from control children. Somepotential etiological implications ofthis sex difference are discussed later.

Is Schizophrenia AssociatedWith a Pre- or Post-OnsetDecline in IQ?

A number of studies have examinedchanges in IQ during the course ofschizophrenia. These studies have

looked for premorbid changes in IQbefore overt schizophrenic symptomsappear (Lane and Albee 1963;Pollack, Woerner, and Klein 1970;Jones and Offord 1975; Watt andLubensky 1976); differences betweenpremorbid scores and scores athospital admission (Lubin, Gieseking,and Williams 1962; Schwartzman,Douglas, and Muir 1962; Albee et al.1963; Pollack, Woerner, and Klein1970); and differences between scoresat hospital admission and scores afterseveral years of hospitalization or atremission (Haywood and Moelis1963; Smith 1964; Hamlin and Ward1965; Klonoff, Fibiger, and Hutton1970). Unfortunately, because of thediversity in methodology and time

periods covered, these studies do notlend themselves to statisticalsummary.

Premorbid IQ Stability. Addressingthe question of premorbid IQstability, the Lane and Albee studies(1964, 1968) compared IQ scoresobtained in second grade with thoseobtained in sixth grade. When scoresfrom preschizophrenic children werecompared with city-wide averages,Lane and Albee (1968) found a statis-tically significant decline in scoresbetween the second and sixth gradesfor the preschizophrenic subjects.However, when their scores werecompared with scores from siblingsand matched controls (Lane and


Dow

nloaded from



Albee 1968), it was found that a lossin IQ between the second and sixthgrades was also characteristic of thematched controls and siblings.Similarly, Watt and Lubensky (1976)examined IQ scores of preschizo-phrenic children from grades 3-6 andgrades 8-11 for downward andupward drift. They found thatpreschizophrenic children wereslightly more consistent than theircontrols, corroborating Lane andAlbee's (1968) conclusion thatpreschizophrenic development is notcharacterized by a significant drop inIQ between early and late childhood.

This conclusion is further corro-borated by results from the study ofPollack, Woerner, and Klein (1970).In this investigation childhood IQscores from grades K-2, 3-5, and 6-8were correlated with one another,separately for preschizophrenicchildren and for their nearest-in-agenormal siblings. For both thepreschizophrenic children and theirsiblings, all intertest correlationswere significant. Intertest correlationsfor the patients ranged from .56 to.97; correlations for the siblingsranged from .67 to .90. The authorsargue that if the schizophrenics hadexperienced a significant decline inIQ during childhood, one wouldhave expected their intertest corre-lations to be much lower than theintertest correlations for theirsiblings. Obviously, the absence of adifference in intertest correlationsdoes not rule out a premorbid declinein IQ, since an IQ loss that isconsistent across subjects would notaffect the magnitude of the coeffi-cient. Nevertheless, taken together,studies which looked for a prodromaldecline in IQ by comparing earlychildhood IQ scores with laterchildhood IQ scores, or by corre-lating scores from tests taken atdifferent ages, provide littleconvincing evidence for a decline in

IQ before the overt appearance ofschizophrenic symptoms.

In their 1965 study Lane and Albeecorrelated the childhood IQs ofschizophrenic patients with thechildhood IQs of their siblings. Theyfound nonsignificant correlations of.06 for the second grade test, .17 forthe sixth grade test, and .07 for theeighth grade test. When theyexamined the relationship betweenthe childhood IQs of two nonschizo-phrenic siblings of the patients, theresulting average correlation was .43for the three grade levels. Theauthors interpreted these data assupporting the hypothesis that theschizophrenic process lowers IQprodromally and, since it does soirregularly, weakens the correlationbetween the IQs of the schizophrenic-to-be and his siblings. In order to testthis hypothesis, Jones and Offord(1975) correlated scores of preschizo-phrenics and their siblings on highschool IQ tests. They found that thecorrelation between the IQs of twosiblings was essentially the same,whether or not one of them waspreschizophrenic. Correlations variedbetween .40 and .50 for both groups.Jones and Offord then concluded thefollowing:

If schizophrenia lowers IQ prodro-mally, the proband-sibling corre-lation should be less than the oneamong siblings, because theprobands' IQ would contain anadditional source of variationunrelated to the siblings' IQs,namely, variations having to dowith the clinical course andseverity of the proband's illness,[p. 188]

Furthermore, their finding thatpreschizophrenic-sibling correlationsdid not differ from sibling-siblingcorrelations led them to argue for theindependent transmission of IQ andschizophrenia. They explain thedifferences between their findings and

the Lane and Albee (1965) findingsby noting that Lane and Albee didnot include subjects who had been inspecial classes as children. WhenJones and Offord reanalyzed theirdata excluding special class subjects,they found that the preschizophrenic-sibling correlations were reducedfrom the mid-.40's to valuescentering around .20. However,another plausible explanation for thediscrepant findings that was notentertained by Jones and Offordconcerns the sex composition of thepreschizophrenic samples. Over 75percent of the preschizophrenicchildren in the Lane and Albee studywere male, whereas only 44 percentof Jones and Offord's preschizo-phrenic children were male. If it isthe case, as suggested above, thatpremorbid IQ deficits are primarilycharacteristic of male schizophrenicpatients, then Lane and Albee'sfindings of low correlations betweenthe IQs of preschizophrenic childrenand their siblings may reflect theinclusion of a high proportion ofpremorbid males with depressed IQs.Thus, although the results of theLane and Albee study do not neces-sarily demonstrate the existence of an"irregular" prodromal decline in IQ,they are compatible with theassumption that male vulnerability toschizophrenia is accompanied by IQdeficit.

Post-Onset IQ Stability. Todetermine whether IQ deteriorateswith the onset of schizophrenia,several studies compared premorbidand postmorbid IQs of schizophrenicpatients. In an early study byRappaport and Webb (1950), schizo-phrenic patients were administeredthe same intelligence tests they hadtaken during high school (beforehospital admission). The premorbidmean IQ was 97.6, compared with apostmorbid mean of 63.9 (significant


Dow

nloaded from


VOL.10, NO. 3,1984 445

at the .01 level). In more recentreports by Lubin, Gieseking, andWilliams (1962), and Schwartzmanand Douglas (1962), army intelli-gence tests were administered toschizophrenic patients, and theirscores were compared to premorbidscores obtained at the time of thepatients' induction into the army. Inthese studies, a deficit score wasobtained for each subject bysubtracting the current score fromthe premorbid score. Again, bothstudies found significantly greaterdeficits among the schizophrenicpatients than among the controls,although the amount of deficit wasnot so great as that reported byRappaport and Webb (1950).

The conclusion drawn from thesestudies of premorbid-postmorbid IQwas that the onset df schizophrenia ischaracterized by a drop in intel-lectual functions. However, a subse-quent follow-back study by Albee etal. (1963) does not support thisconclusion. In the Albee et al. study,the Wechsler-Bellevue test wasadministered to adult schizophrenicpatients. No significant differenceswere found between the adult IQscores and the childhood scores onthe Stanford Binet, Kuhlmann-Anderson, and Cleveland Classi-fication Test administered in thesecond and sixth grades. All meanscores were somewhat below normal(means ranged from 87.1 and 92.6).The authors conclude:

Intellectual deficit as a charac-teristic of schizophrenia does notappear to be a concomitant of thesevere or acute stages of thedisorder but rather is a charac-teristic of the total life picture, atleast from early childhood on, ofthose individuals who are even-tually admitted to public mentalhospitals, [p. 366]

Albee et al. suggest that thediscrepancy between their findings

and the findings of Lubin, Gieseking,and Williams (1962) is related to thenature of the tests administered afterthe onset of illness. The Wechsler-Bellevue tests used in the Albee et al.study are not so rigidly timed as thearmy tests used by Lubin, Gieseking,and Williams. Albee et al. argue thattests such as the army battery mightbe more affected than the Wechsler-Bellevue by the changes inmotivation, attention, and persis-tence that often accompany schizo-phrenia. However, whenSchwartzman and Douglas (1962)controlled for the speed factor in thearmy tests, the decrement manifestedby the schizophrenic subjects wasreduced, but still significantly greaterthan that of the controls.

Studies that have looked forpremorbid-postmorbid changes in IQare difficult to compare because ofmethodological inconsistencies. Thereare differences in the premorbid andpostmorbid tests administered, in theages at which premorbid tests wereadministered, in length of hospi-talization and chronicity before theadministration of postmorbid tests,and in demographic characteristics ofthe samples studied. Moreover, thestudies of Lubin, Gieseking, andWilliams and of Schwartzman andDouglas included only male subjects,while the Albee et al. study includedboth males and females. Thus, it ispossible that decline from premorbidIQ level is predominantly charac-teristic of schizophrenic males, andtherefore was only detected in thestudies of Lubin, Gieseking, andWilliams (1962) and Schwartzmanand Douglas (1962).

Results from studies that examinechanges in IQ level as schizophreniaprogresses are more consistent thanthe findings from research comparingpremorbid-postmorbid IQs. Inves-tigations of IQ stability following theonset of schizophrenia suggest that

symptomatological improvements arerelated to increases in IQ, while lackof symptom remission is associatedwith decrements. Payne (1960)obtained IQ scores for newlyadmitted schizophrenic patients andagain after 4 to 13 months withoutremission. The average IQ for thewhole group on admission was 82.8,and the average retest IQ was 75.2.The correlation between length ofillness and drop in IQ for the wholegroup was .39. Payne concludes:

Schizophrenic illness withoutremission tends to produce adeterioration of general level whichis progressive. The longer theillness, the more severe the deteri-oration tends to be. [p. 204]

In another early study, Rabin (1944)tested schizophrenic and nonschizo-phrenic psychiatric patients on theWechsler-Bellevue. The initial testingwas carried out approximately 1month after admission, and thefollowup testing occurred from 1 to35 months later (mean interval was13 months). The attending psychia-trists noted that there was a generalimprovement in clinical sympto-matology among the patients at thetime of the retesting. Of the 30schizophrenic patients, 16 showed arise of 5 or more points between thefirst test and the retest; eight showeda decline of 5 or more points. Thefigures were exactly the same for thenonschizophrenic patients. Theaverage initial test score was 81.6 forschizophrenic patients and 76.8 fornonschizophrenic patients; averageretest scores were 87.5 and 82.1,respectively. Rabin concludes that therise in mental level among themajority of patients was due to theirimproved clinical picture and,possibly, to practice effects.

Increases in IQ were also found ina subsequent study by Smith (1964),although he does not report on


Dow

nloaded from



symptom changes. This author testedyounger schizophrenic patients (meanage at first testing was 34 years) andolder schizophrenics (mean age was53) on the Wechsler-Bellevue test.Retesting occurred when the patientswere still hospitalized, 8.4 yearslater. Although the younger groupscored considerably lower on theinitial testing than the older group(means = 79.81 and 100.6, respec-tively), both groups showed slightincreases over the 8.4-year period(5.59 points for the younger schizo-phrenics, 1.96 points for the olderschizophrenics). Smith concludes thatthe study revealed no signs ofprogressive deterioration purely as afunction of hospitalization.

Hamlin (1969) followed up thesample described above in the Smith(1964) study. Hamlin was able toobtain Wechsler scores for 49patients on tests taken 14 years afterthe initial testing described by Smith(1964). Twenty-five of the subjects inthis study were younger nonparanoidschizophrenic patients (mean age atinitial testing was 33 years). Theremaining 24 subjects were olderparanoid schizophrenic patients(mean age at initial testing was 51).Hamfin first examined unweightedWechsler scores, which representactual changes in performance overthe years, with no allowance fornormal aging. He found significantgains during the 14-year period forthe younger nonparanoids, and nosignificant change for the olderparanoid patients. When the twogroups were compared on the finalfollowup IQ scores, with acovariance adjustment for initialscores, no significant difference wasfound between the scores of paranoidand nonparanoid patients. Hamlinconcludes that the stability of scoresover 14 years was essentially thesame for the paranoid and nonpara-noid subgroups—that is, paranoid

and nonparanoid patients do notrepresent two varieties of psychosiswith respect to IQ changes during thechronic stage of the illness.Furthermore, when subjects werecompared on Wechsler scores, whichdid take normal aging into account,significant gains were found for bothgroups. The effects of aging on IQ,he suggests, "may be counteracted bya general tendency toward alleviationof psychotic symptoms as the yearspass. Instead of following aninevitable downhill course, schizo-phrenia may become less severe"(p. 501).

Hamlin also obtained information,at the time of followup testing, aboutthe clinical status of 44 of thepatients. The 14 patients who wererated as "improved" showed a meangain of 16.1 points over the 14 years.No statistical tests were used tocompare this gain with the gains ofthose patients who were not selectedas improved; however, the change inIQ of the improved patients appearsto be substantially higher than themean gains of 11.43 and 4.48, respec-tively, for the other nonparanoid andparanoid patients. Comparablefindings are reported by Klonoff,Fibiger, and Hutton (1970) whoadministered the WAIS to WorldWar II veterans who had beenclassified as schizophrenic forapproximately 20 years. Patientswere retested 8 years later and theresultant data indicated a significantimprovement from an average of93.9 during the initial testing to anaverage of 100.54 during thefollowup testing. Ratings on theMalamud and Sands psychiatricrating scale indicated that thepsychiatric status of schizophrenicshad also improved significantly overthe 8-year period.

In a study by Schwartzman,Douglas, and Muir (1962), theCanadian Army "M" test was admin-

istered to 23 schizophrenics after amean of approximately 4 years ofhospitalization (first postmorbidtesting), and again 8 years later(second postmorbid testing). Betweenthe first and second tests, 13 of thepatients were released while 10remained hospitalized. Schwartzman,Douglas, and Muir compared thechange between the first and secondpostmorbid test scores for thereleased and nonreleased patients.While the nonreleased patients had asignificant decline of 13.6 points, thereleased patients had a significantincrease of 13.4 points. Schwartz-man, Douglas, and Muir obtainedpremorbid scores from the time ofarmy induction for the 23 subjects.They found that the released patients'second postmorbid scores were notsignificantly different from theirpremorbid scores. Thus, it appearsthat released patients gained backmuch of the intellectual functioningthat they had lost when hospitalizedwith schizophrenia, while patientswho remained in the hospitalcontinued to decline. Schwartzmanand his collaborators found no corre-lation between length of hospitaliza-tion and amount of drop between thepremorbid and first postmorbid tests,between the premorbid and secondpostmorbid tests, or between the firstand second postmorbid tests. Theauthors conclude that there was noevidence to suggest that length ofhospitalization is correlated withamount of intellectual loss, but thatthe significant factor is whether thepatient is in the hospital at the timeof the testing.

In a similar study by Haywoodand Moelis (1963), 20 improved maleschizophrenic inpatients werematched on age and IQ, obtained attime of admission, with 20unimproved male schizophrenicinpatients. Improvement was definedas psychiatrists' judgments that


Dow

nloaded from


VOL. 10, NO. 3,1984 447

psychotic symptoms had subsided.All patients were retested after aminimum of 16 months of hospi-talization. The mean change for theimproved group was an increase of7.3 points; the mean change for theunimproved group was a decline of2.65 points. Of the 20 improvedpatients, 16 gained IQ points and 3lost. In the unimproved group, 5gained and 13 lost. The authorsconclude that "improved schizo-phrenics show improvement in intel-lectual functioning as compared withtheir performance on admission tothe hospital, while unimprovedschizophrenics may actually sufferfurther deficit" (p. 78). Haywood andMoelis's (1963) findings corroboratefindings of an earlier study byDavidson (1939) in which 18 schizo-phrenic patients were tested on theStanford-Binet and retested 6 monthslater. The three patients who showedclinical improvement had higherretest scores (compared with initialtest scores); the seven patients whoshowed clinical deterioration hadlower retest scores.

Overall, the results of these studiessuggest that continual deteriorationin intellectual functioning during thecourse of the illness is by no means anecessary characteristic of schizo-phrenia. Furthermore, it appears thatimprovement in IQ between initial(postmorbid) testing and retesting atsome later date is related to improve-ment in psychiatric symptoms. Thus,in answer to our question regardingthe covariation of IQ and clinicalstatus, the following conclusions canbe drawn from the literature:(1) There does not appear to be adecline in IQ before the overtappearance of schizophrenicsymptoms. (2) Results are equivocalregarding premorbid-postmorbiddeclines—results may depend on thepremorbid and postmorbid testsadministered, the ages at which

premorbid tests are administered, thelength of the patient's hospitalizationbefore the postmorbid test isadministered, and sex of the patient.(3) Differences between scoresobtained at admission and afterseveral years of hospitalizationappear to be related to changes in thepatient's symptomatology. AsHamlin (1969) notes:

Patients may lose abilities early inthe psychosis, but once they havebecome chronic, their mean intelli-gence scores remain stable or evenimprove. . . . [The] severity ofpsychotic symptoms has a demon-strable relationship to intelligencescores. When symptoms improve,test scores increase, [p. 502]

Are Clinical Improvement,Length of Hospitalization, orPrognostic Indicators ofSeverity and ChronicityRelated to Premorbid orPostmorbid IQs?

Several studies have looked at signsof clinical improvement (e.g.,discharge ratings, and incidence ofremission) and their relationship withIQ at hospital admission. In an earlystudy by Stotsky (1952), schizo-phrenic patients who had beenhospitalized for 2-4 years werecompared with patients who wereadmitted at the same time but whohad been released and had remainedin remission for at least 6 months. IQscores of remitted patients (obtainedwithin 90 days of hospital admission)were significantly higher than the IQsof nonremitted patients. Similarly,Carp (1950) found that thepretreatment IQs of patients whoresponded favorably to therapy weresignificantly higher than the pretreat-ment IQs of unimproved patients.

Pollack (1960) and Pollack,Levenstein, and Klein (1968) alsofound a relationship between clinical

improvement and IQ. Pollack (1960)obtained IQ scores and dischargeratings for adult and adolescentschizophrenic patients. (It is not clearwhen the IQ tests were administered,but they were apparently admin-istered sometime during the patient'shospitalization.) The dischargeratings consisted of the classificationof patients as unimproved,improved, much improved, andrecovered. The unimproved subgrouphad significantly lower IQs than thepatients with more favorable ratings.In fact, 50 percent of the unimprovedpatients had IQs below 90. No signif-icant differences in IQ were foundamong subgroups with ratings ofimproved, much improved, orrecovered.

In the subsequent study byPollack, Levenstein, and Klein(1968), adult and adolescent schizo-phrenic patients who were hospi-talized for an average of 9 monthswere followed up 3 years afterhospital discharge. At the time offollowup, each subject was ratedindependently by two judges asbelonging to one of six globaloutcome categories. Wechsler-Bellevue IQs had been obtainedduring hospitalization, at the end ofa 4-week, drug-free period. IQ scoreswere significantly higher for patientsin the excellent-good outcomecategories than for those in the verypoor category.

Prognosis has also been found tobe associated with childhood IQscores of schizophrenics. In a seriesof followup studies of boys treated ata child guidance clinic, Roff and hiscolleagues (Roff, Knight, andWertheim 1976a, 1976b; Roff andKnight 1980) found that preschizo-phrenic boys with low childhood IQshad disproportionately unfavorableclinical outcomes and never-marriedstatus. In summary, then, the resultssuggest that higher premorbid and


Dow

nloaded from



postmorbid IQs (obtained duringhospitalization) are related to betterclinical outcome for schizophrenicpatients.

Other investigators have looked atan alternative indicator of outcome,length of hospitalization, and itsrelationship to IQ. Heffner, Strauss,and Grisell (1975) obtained IQ scoresfrom tests administered to maleschizophrenic patients at the time ofhospital admission. Subjects weredivided into the categories of high(> 95) and low « 95) IQ. Followupinterviews were conducted todetermine whether the subject hadbeen readmitted during the thirdand/or fifth years following initialdischarge. Patients were assigned toone of three outcome categories: (1)short-term (not rehospitalized), (2)intermediate (rehospitalized eitherduring the third or fifth yearfollowing initial discharge), and (3)chronic (rehospitalized in both thethird and fifth years). Chi-squareanalysis revealed a significantrelationship between IQ andreadmission ratings. Over 80 percentof the chronic patients had been inthe lower IQ group at hospitaladmission. Although 70 percent ofthe high IQ group was never rehospi-talized, this was true of only half ofthe lower IQ group.

While Heffner, Strauss, and Grisellfound that IQ, as measured atadmission, is related to length ofhospitalization, the findings of threestudies of premorbid IQ and lengthof hospitalization are inconsistent.Offord and Cross (1971) examinedthe relationship between childhood(premorbid) IQ and length of hospi-talization for 29 schizophrenics atthree age levels: 20 to 29, 30 to 39,and 40 or older. These investigatorscalculated the number of months ofinstitutionalization for each of thesubjects. For all three age groups,patients with lower childhood IQs

had accumulated more institutionaltime. For patients who were at least30, for example, those withchildhood IQs under 80 had spent anaverage of 4 more years in institu-tional care than patients with IQs of100 or above.

However, Watt and Lubensky(1976) found no relation betweenpatient-sibling IQ discrepancies andlength of hospitalization. Asmentioned earlier, these authorsobtained a childhood IQ score for 50schizophrenic patients and theirsiblings by averaging scores from allIQ tests administered betweenkindergarten and high school.Subjects were matched with controlclassmates on the basis of sex, race,social class of'origin, and migratorystatus. Watt and Lubensky deter-mined the length of hospitalizationfor each patient during the first 4years after initial admission, andthen divided them into twogroups: those hospitalized less the260 days (short hospitalization) andthose hospitalized for 260 days ormore (long hospitalization).Although the preschizophrenicchildren had significantly lower testscores than their matched controls,long-hospitalized patients and short-hospitalized patients differed aboutequally from their siblings andmatched controls. However,premorbid IQs of the long- andshort-hospitalized groups were notdirectly compared; consequently theWatt and Lubensky study does notprovide a direct test of therelationship between length of hospi-talization and premorbid IQ. Theonly study that directly examinedthis relationship and found negativeresults was conducted bySchwartzman, Douglas, and Muir(1962). These authors found norelationship between premorbid testscores on the Canadian Army "M"test and length of hospitalization for

23 schizophrenic patients.In summary then, Schwartzman,

Douglas, and Muir (1962) found nopremorbid IQ-length of hospital-ization relationship, while the studyof Offord and Cross (1971) providedsome evidence for such arelationship. The difference betweenthese two studies in age of subjects atpremorbid assessment limits thecomparability of the results. It islikely, however, that the sample ofSchwartzman, Douglas, and Muirconstituted a more select andrestricted sample of preschizophrenicsubjects—namely, those who weredeemed eligible to serve in the armedforces. Offord and Cross's samplehad an average childhood IQ of 86.6,while the Schwartzman group'ssample had an average test score thatdid not differ significantly from theaverage score for all army inductees.It is possible, then, that premorbidIQ scores are related to length ofhospitalization only when a broaderrange of premorbid IQs isconsidered.

There is considerable evidence thatIQ is positively correlated with ageat onset of illness. Offord and Cross(1971) found that premorbid IQ wasassociated with age at first hospi-talization in their sample of schizo-phrenic patients: Patients withchildhood IQs below 80 were hospi-talized 4 years earlier than those withIQs between 80 and 99, and almost10 years earlier than those with IQsabov<; 100. Postmorbid IQs have alsobeen found to vary with age athospitalization. In the report byPollack (1960), discussed above,schizophrenic and nonschizophrenicpsychiatric patients (manic-depressives, psychoneurotics,behavior disorders) were divided intothree groups: preadolescents (6-9years), adolescents (13-18 years), andadults (19-44 years). At 1 yearbefore the current admission, most of


Dow

nloaded from


VOL. 10, NO. 3,1984 449

these patients had never been hospi-talized. Pollack found that for boththe schizophrenic and nonschizo-phrenic patients, the mean IQ scoresof the preadolescent patients weresignificantly lower than average (75.1and 93.9, respectively), and weresignificantly lower than the mean IQsof the other age groups; the samewas true for adolescents under 15years of age. The differences amongthe age groups were much smaller,however, for the nonschizophrenicpatients when compared to schizo-phrenic patients (a 40-point differ-ence between the preadolescent andadult schizophrenic patients vs. a 19-point difference between thepreadolescent and adult nonschizo-phrenic patients). The authorconcludes that age at onset ofpsychiatric disorder is positivelyrelated to IQ, and that this may beespecially true for schizophrenicpatients.

Several subsequent studies haveyielded results that are consistentwith this conclusion. Smith (1964)reports that younger schizophrenicpatients scored significantly lowerthan older patients both on initialtesting (mean = 80 vs. mean = 101)and on retesting 8 years later (mean= 86 vs. mean = 103). Pollack et al.(1968) found that schizophrenics whowere between 30 and 57 years old atfirst admission had significantlyhigher mean IQ scores than thoseadmitted before the age of 30. Theresults of this investigation alsorevealed that early onset (adoles-cence, early adulthood) schizophrenicpatients were predominantly male,and showed evidence of minimalbrain damage during childhood, aswell as low IQ. Similarly, a retro-spective study by Pollack,Levenstein, and Klein (1968)indicated that low IQ in adult schizo-phrenic patients is associated withacademic performance deficits in

childhood, and Belmont et al. (1964)report that schizophrenic patientswith low IQs were more likely tohave shown marked childhoodbehavior disorders. Consistent withPollack, Levenstein, and Klein(1968), Belmont et al. (1964) alsofound that childhood behaviordisturbance was more prevalentamong male schizophrenics andtended to be associated with signs ofcentral nervous system impairment.

In another study relating age ofonset and IQ, Pollack, Woerner, andKlein (1970) subdivided their samplesof schizophrenic and personality-disordered patients into three groups;those with age of onset of serioussymptoms before 13 years, between13 and 16 years, and after 16 years.For both diagnostic groups, patientswith symptom onset before age 13had the lowest mean IQ scores.When the patients' adult IQ scoreswere compared with the adult IQscores of their siblings, it was foundthat the schizophrenic patients withchildhood onset had significantlylower IQ scores than their siblings,whereas those with early adolescent(ages 13-16) or late adolescent-adult(16 years or older) onset did notdiffer from their siblings on IQ.Personality-disordered patients withchildhood or early adolescent onsetalso had significantly lower IQ scoresthan their siblings. Thus, the findingsof Pollack, Woerner, and Klein(1970) provide additional evidencefor a strong relationship between IQand age of onset of schizophrenia,and suggest that the relationship mayhold for other disorders as well.

Jones (1973) examined therelationship between the premorbidIQs of schizophrenic patients andindirect measures of anotherprognostic indicator, heterosexualadjustment. His measures of hetero-sexual adjustment were marital statusand fertility rates among 114 male

and 108 female schizophrenicpatients. For about one third of thissample, IQ scores were obtainedfrom elementary or junior highschool records. The rest of thepatients were tested during hospi-talization with either the WAIS orthe General Aptitude Test Battery(test "G" from the latter test wasrescored and used as the measure ofIQ). There was a significant corre-lation between IQ and number ofchildren for male patients, but notfor female patients. Similarly, maleswith higher IQs were significantlymore likely to have been married atsome time than males with lowerIQs. This relationship did not holdfor the female patients. These resultscould be interpreted as indicatingthat male, but not female, sex roleexpectations for heterosexualrelationships are linked to cognitivecompetency. Alternatively, it ispossible that male patients withlower IQs are more seriouslydisturbed and consequently lesscapable of, or less interested in,establishing a relationship.

In general, the studies reviewedhere suggest a fairly consistentrelationship between higher IQ andmore positive outcome for schizo-phrenic patients, whether outcome isindexed by psychiatric ratings,incidence of remission, or length ofhospitalization. Similarly, prognosticindicators such as age at onset andheterosexual adjustment arepositively related to IQ. Finally, itappears that a premorbid history ofbehavioral disturbance and low IQ ismore common among males.

Are There Patterns ofPremorbid or PostmorbidSubtest Performance ThatCharacterize or PredictSchizophrenia?

As previously mentioned, many of


Dow

nloaded from



the early (i.e., pre-1960) studiesinvestigating the relationship betweenschizophrenia and IQ were concernedwith determining whether specificpatterns of test performance charac-terize various mental disorders. Thepurpose of these studies was toprovide information that wouldallow the IQ tests to be used as atool in diagnosis. In a review ofstudies that looked for differentialpatterns of performance on theStanford-Binet, Brody (1942)concluded that "the mental testpattern in the developed psychoses isremarkably constant. . . . Thetypical pattern is: vocabularyhighest, verbal test ability second,non-verbal test ability lowest" (p.255). Similarly, Wechsler (1958)characterized the performance ofschizophrenics on the Wechsler-Bellevue and WAIS tests as generallyshowing a higher Verbal IQ thanPerformance IQ score.

Based on the observation thatverbal scores, particularly scores onthe Vocabulary subtest, are generallyhigher than nonverbal scores, severalearly investigators attempted todesign methods for estimatinggeneral intellectual deterioration inschizophrenic patients. Payne (1960)reviews several of these methods,including the Babcock-Levy test forthe measurement of efficiency ofmental functioning (Babcock andLevy 1940), the Shipley-Hartford testof deterioration (Shipley 1940), andthe Hunt-Minnesota test for organicbrain damage (Hunt 1943). All ofthese tests, Payne notes, are based onthe assumptions that vocabularymeasures have a high correlationwith other measures of general intel-ligence and, when scores of normaland psychotic groups are compared,the test yielding the smallest meandifference is generally a vocabularytest. Thus, early investigators in thisarea assumed that vocabulary is the

ability least affected by mentalillness, and that a patient's currentvocabulary score is a good indicationof his pre-illness level of overall intel-ligence. Payne notes, however, thatnone of the early methods forestimating pre-illness intelligencewere ever directly validated or stand-ardized on psychiatric patients.

Much of the early research thatexamined the issue of differentialdeterioration of intellectual abilitiesused the Babcock-Levy test, or othersimilar methods, fo estimatepremorbid intelligence. By comparingthe schizophrenic patients'postmorbid scores on varioussubtests with their estimatedpremorbid level of intelligence,investigators drew conclusions aboutthe intellectual abilities most affectedby the disorder. Because of the lackof validation and standardization ofthe methods used to estimatepremorbid intelligence, one mustquestion the findings of studies thatused this design. It is possible thatthe postmorbid pattern ofperformance detected in these studiesactually predates the onset ofsymptoms and reflects the predis-position to schizophrenia. Also, thesestudies did not attempt to matchsubtasks on discriminative power(Chapman and Chapman 1973), sothe results may represent spuriousconsequences of the differentialpsychometric properties of the tasks.One cannot, however, discount thepossibility that differentialpostmorbid deterioration does occurin schizophrenia. In reviewing thestudies that used the Wechsler,Stanford-Binet, or Thurstone tests,Winder (1960) concludes:

It seems very plausible that differ-ential deficit will be found only incertain schizophrenics, not as acharacteristic of all or even themajority of schizophrenic patients,[p. 204]

Subsequent investigations haveused somewhat more sophisticatedparadigms to study differentialdeterioration in schizophrenia. Theseinvestigations can be divided intothree groups: studies comparingsubtest patterns of schizophrenicpatients with those of controls;studies comparing patterns of schizo-phrenic subgroups; and thosecomparing subtest patterns on teststaken before the onset of thedisorder, or at time of hospitaladmission, with patterns on teststaken after several years of hospi-talization. The literature relevant tothese three topic areas is discussedseparately.

Differentiating Schizophrenic PatientsFrom Normal Controls and OtherPatient Groups. One of the fewreplicated findings from research onsubtest performance is that schizo-phrenics, and some other patients,tend to show higher Verbal thanPerformance IQs. Parker andDavidson (1963) compared WAISVerbal (VIQ) and Performance (PIQ)scores of 50 psychotic patients(mixed diagnosis) with scores of 50student nurses who were matchedwith the patients on Full Scale IQ.For both groups, Verbal IQ scoreswere significantly higher thanPerformance IQ scores; however, theVIQ-PIQ difference was significantlygreater for the psychiatric patientsthan for the controls. The mean VIQ-PIQ difference for 30 patientsdiagnosed as psychotic (includingpatients with schizophrenia, manic-depression, psychopathic person-alities with psychosis, and involu-tional psychosis) was 9.96; the meandifference for the controls was 3.9.(No significance tests were conductedon the difference scores.)

Comparisons of schizophrenicpatients with brain-damaged patientsindicate that both groups are charac-


Dow

nloaded from


VOL. 10, NO. 3,1984 451

terized by higher VIQ than PIQ;however, there is limited evidence tosuggest that these two groups differon more specific subtest patterns.Watson (1965) compared WAISscores of schizophrenic patients withthose of patients with organic braindamage. Patients in the two groupswere matched for mean age,education, Full Scale IQ, and totallength of hospitalization. For bothschizophrenic and brain-damagedpatients, the mean Verbal IQ scorewas significantly higher than themean Performance IQ score. Thesubtest X diagnosis interaction wassignificant only for the long-termhospitalized group (hospitalization >37.5 months). Post hoc analysesindicated that the Digit Span scorewas significantly higher for long-termschizophrenics than for long-termbrain-damaged patients.

Another similar pattern of subtestperformance was reported byDeWolfe et al. (1971) who comparedWAIS subtest scores of chronicschizophrenic patients with those ofpatients with nonlateralized braindamage. Patients in the twodiagnostic groups were dividedaccording to age, with patients in theolder group being 60 years and over,and patients in the younger groupbeing ages 26-59. For each subtest, amean deficit score was calculated bysubtracting each subject's subtestscore from the mean of all his or hersubtests. DeWolfe et al. found signif-icant differences between schizo-phrenic and brain-damaged patientson deficit scores for the Digit Spanand Comprehension subtests, withschizophrenic patients showinggreater deficit on Comprehension andless deficit on Digit Span than thebrain-damaged patients. For the olderpatients, significant differencesbetween the two diagnostic groupswere also found for the PictureCompletion and Block Design deficit

scores. Schizophrenic patientsshowed more deficit on PictureCompletion than the brain-damagedpatients, but less deficit on the BlockDesign subtest. However, subsequentattempts to replicate these findingshave proved unsuccessful (Davis,DeWolfe, and Gustafson 1972;Chelune et al. 1979).

In a recent study, Lehman,Chelune, and Heaton (1979)compared WAIS scores of normalcontrol subjects with scores fromschizophrenic patients, patients withacute brain damage, and chronicbrain-damaged patients (the lattertwo groups being distinguishedaccording to the suddenness of onsetand rapidity of progression). Thenormal controls obtained signif-icantly higher Full Scale IQ scoresthan the other three groups, and theschizophrenics scored significantlyhigher than the acute brain-damagedpatients. When groups werecompared on intertest variability, thenormals differed significantly onlyfrom the acute brain-damaged group.Comparisons of the normal groupwith the schizophrenic and thechronic brain-damaged groupsreached marginal significance, withthe normals showing the smallestamount of intertest variability. Inanother report from this researchteam, subtest performance patternsof the patient groups were compared(Chelune et al. 1979). A discriminantfunction analysis indicated no groupdifferences in patterns of subtestscores.

Thus, comparisons of schizo-phrenic and brain-damaged patientssuggest that both groups show intel-lectual deficits, and no specificpattern of subtest performance distin-guishes them. Similarly, the results ofresearch investigating subtest patternsamong various psychiatric diagnosticgroups indicate that there are nodistinguishable patterns. For

example, Pollack, Woerner, andKlein (1970) compared adult IQscores of schizophrenic patients,personality-disordered patients, andtheir normal siblings. The subtestpatterns were similar for all groups,with Verbal IQ scores being higherthan Performance scores. For bothdiagnostic groups, patients' scores onVocabulary, Digit Span, DigitSymbol, Comprehension, PictureCompletion, and Object Assemblywere significantly lower than those oftheir siblings. The mean scores ofschizophrenic patients did not differsignificantly from those of thepersonality-disordered group on anyof the subtests. Similar results arereported by Schoonover and Hertel(1970), who examined WISC scoresfor children in the following ninediagnostic categories: mentaldeficiency, chronic brain syndrome,schizophrenic reaction, psycho-neurotic disorder, personality patterndisturbance, personality traitdisturbance, sociopathic personalitydisturbance, special symptomreaction, and transient situationalpersonality disorder. The children inthe schizophrenic reaction, mentaldeficiency, and chronic brainsyndrome groups scored significantlylower than the other six groups onFull Scale IQ. No systematicrelationship was found betweendiagnostic categories and VIQ-PIQdifferences or subtest patterns. Theauthors conclude that their data donot support the hypothesis thatdiagnostic categories are differ-entiated readily by WISC scoresalone.

Attempts to use Wechsler's (1958)subtest criteria for differentiatingamong diagnostic groups have alsobeen unsuccessful. Wechsler andJaros (1965) conducted a study todetermine whether IQ patterns coulddiscriminate between normal andschizophrenic children. For ages 8, 9,


Dow

nloaded from



10, and 11 years, WISC scores of 25normal boys were compared withscores of 25 young schizophrenicmales. It was concluded that fivepattern "signs" could discriminatesignificantly between the two groupsat each age level. These were: (1)three subtests, each deviating by 3 ormore scaled score points from themean; (2) Picture Arrangement scoregreater than Picture Completionscore, and Object Assembly scoregreater than Coding score, each by 3scaled score points; (3) Compre-hension score greater than Arithmeticscore, and Similarities greater thanArithmetic by 3 scaled score points;(4) Verbal IQ minus Performance =±16 points; and (5) Sign #1 plus anyone of the other signs. In a laterstudy by Kissel (1966), four of thefive signs found in the Wechsler andJaros (1965) study were applied toWISC scores of delinquent andacting-out children. The signs werefound to be almost as sensitive indetecting nonschizophrenic,emotionally disturbed children,suggesting that these IQ pattern"signs" are not unique to schizo-phrenia.

A recent investigation of subtestpatterns among psychiatric patientswas reported by Holland and Watson(1980). These authors administeredthe WAIS and the Minnesota Multi-phasic Personality Inventory (MMPI)to 84 patients with organic brainsyndrome, 162 schizophrenics, 86neurotics, and 91 alcoholics shortlyafter hospital admission. The schizo-phrenic patients were subdivided intoprocess and reactive types on thebasis of whether they had ever beenmarried. One-way analyses ofvariance on each of the WAISsubtests indicated that patients withprocess schizophrenia scored signifi-cantly lower than the mean for allpatients on the Comprehension andArithmetic subtests. Patients with

reactive schizophrenia did not differsignificantly from the mean on anyof the WAIS subtests. When amultiple discriminant analysis wasperformed, three discriminantfunctions for the combination ofWAIS and MMPI scores were statis-tically significant. One of these,which was most highly correlatedwith the Digit Symbol, Block Design,and Object Assembly subtests of theWAIS, significantly discriminated theprocess and reactive schizophrenicpatients from the brain-damaged andalcoholic patients, but not from theneurotic patients. No directcomparisons were made to determinewhether the groups differed signifi-cantly on any of these scales.

In summary then, studies thatcompare subtest patterns for variousdiagnostic groups suggest that VIQ-PIQ differences may be greater forall psychiatric patients than fornormal controls, but that subtestscore patterns do not consistentlydifferentiate among schizophrenicpatients, other psychiatric patients,brain-damaged patients, andnormals.

Differentiating Among SchizophrenicPatients. Several studies haveattempted to determine whethercertain subtest patterns candistinguish between schizophrenicpatients who differ on one or moreof the following characteristics: age,length of hospitalization, neuro-logical damage, presence of paranoidsymptoms, premorbid status, andchronicity. DeWolfe et al. (1971)compared scores of older andyounger psychiatric patients (schizo-phrenic and brain-damaged patients)and found a significant age X subtestinteraction, indicating that subtestpatterns differ for the young and oldpatients. As mentioned above,DeWolfe et al. found significantsubtest differences between older

schizophrenic and brain-damagedpatients; schizophrenic patientsshowed more deficits on PictureCompletion than brain-damagedpatients; but less on Block Design.No direct comparisons of deficitscores for older vs. younger schizo-phrenics were performed, however.In a study that did directly compareolder and younger patients. Smith(1964) found significant differencesbetween young (mean age = 34years) and older schizophrenicpatients (mean age = 53) on theComprehension, Arithmetic,Vocabulary, Picture Completion, andBlock Design subtests of theWechsler-Bellevue test. On all ofthese subtests, the younger patientsperformed better. Since thesemeasures include verbal andnonverbal subtests, Smith concludedthat the nature of impairment in theolder schizophrenic patients isnonspecific. It should be noted,though, that the Smith study did notcontrol for normal changes thataccompany aging. Hamlin and Ward(1965) also found significantdifferences in patterns of subtestscores between older and youngerschizophrenics: this patterndifference involved loss onperceptual-motor tasks, with stabilityor slight gains on symbolic tasks.However, they found the samedifferences for aging nonpatients,suggesting that differences foundbetween young and old schizophrenicpatients were not necessarily relatedto changes in the course of theirillness^

We know of only two inves-tigations (Davis, Dizzonne, andDeWolfe 1971; Goldstein andHalperin 1977) that have attemptedto distinguish among diagnosticsubgroups of schizophrenic patientson the basis of subtest patterns.Davis, Dizzonne and DeWolfe (1971)compared WAIS subtest scores for 40


Dow

nloaded from


VOL. 10, NO. 3,1984 453

schizophrenic patients equallydivided among four groups on thebasis of premorbid history (reactivevs. process) and the length of timethey had been hospitalized (meanlength of hospitalization for theshort-term group was 1.08 years;mean hospitalization for the long-term group was 5.04 years). In allfour groups, the patients performedmore poorly on the performancemeasures than on verbal measures.Relative subtest scores were obtainedfor each patient by subtracting eachsubtest score from the subject'soverall subtest score mean. Short-term patients tended to do better onthe Similarities and Comprehensionsubtests and more poorly on theArithmetic subtest than did the long-term patients. Patients with reactiveschizophrenia scored significantlyhigher on the Digit Symbol test thandid the patients with process schizo-phrenia. It should be noted thatHolland and Watson's (1980)discriminant function analysis(described earlier) also found that theDigit Symbol test differentiatedbetween patients with process andreactive schizophrenia.

Goldstein and Halperin (1977)administered the WAIS, Halstead-Reitan battery, and several othermeasures to 140 schizophrenicpatients. Patients were divided intosubgroups according to threecriteria: paranoid vs. nonparanoid,long-term vs. short-term (< 1 year ofhospitalization vs. > 1 year); andneurologically normal vs. abnormal.These authors used WAIS subtestscores, scores from the Halstead-Reitan battery, and associatedmeasures as independent variables instepwise discriminant analysesdesigned to distinguish betweenpatients on the three criterionvariables. Among the 10 measureswhich, when taken together,significantly discriminated between

paranoid and nonparanoid subjectswere the WAIS Arithmetic,Information, Comprehension, andVocabulary subtest scores. The threeWAIS subtests that contributed mostto the discrimination betweenneurologically normal and abnormalpatients were Object Assembly,Block Design, and Arithmetic Scores.Long-term vs. short-term hospitali-zation was best discriminated byWAIS Digit Span, Digit Symbol,Object Assembly, PictureArrangement, and PictureCompletion. (No direct comparisonswere made to determine the directionor significance of differences betweenthe groups.) In contrast, it should benoted that the study of Davis,Dizzonne, and DeWolfe (1971) foundthat Comprehension, Similarities,and Arithmetic were related to lengthof hospitalization. Thus, the resultsof these studies show littleconvergence and suggest that lengthof hospitalization is not related tospecific subtest scores. However,additional research is needed toestablish that premorbid status,neurological impairment, or theparanoid-nonparanoid distinction arerelated to subtest patterns.

Changes in Subtest PerformancePatterns Over Time. Longitudinalstudies have also examined subtestpattern changes in schizophrenicpatients. Two studies (Lubin,Gieseking, and Williams 1962;Schwartzman and Douglas 1962)have compared premorbid subtestpatterns from the time of armyenlistment with postmorbid subtestpatterns. Others have looked atsubtest pattern changes that occurafter 8 years of hospitalization(Smith 1964; Klonoff, Fibiger, andHutton 1970). In the study of Lubin,Gieseking, and Williams (1962),premorbid scores on the ArmyClassification Battery (ACB) were

compared to scores taken 1-3 monthsafter hospital admission for paranoidschizophrenic, nonparanoid schizo-phrenic, brain-injured, andnonpsychiatric patients. Nosignificant differences in amount orpattern of deficit were found betweenthe two schizophrenic groups. Forboth paranoid and nonparanoidschizophrenic patients, all the ACBtests, except Pattern Analysis,showed a significant drop frompremorbid levels. In general, thenonspatial tasks (Reading andVocabulary, Arithmetic, Reasoning,and Army Clerical Speed) showedmore premorbid-postmorbid declinethan the spatial tests (PatternAnalysis and Mechanical Aptitudesubtests). Comparing schizophrenicpatients and controls, Lubin,Gieseking, and Williams found thatschizophrenic patients showed signifi-cantly more decline than controls,and that this differentiation wasparticularly dependent on the non-spatial subtests of Reading andVocabulary and Army ClericalSpeed. Finally, in comparisons ofschizophrenic and brain-damagedpatients, it was found that schizo-phrenic patients have less overallimpairment than brain-damagedpatients, but that the general patternof decline is similar. The results ofthe study of Lubin, Gieseking, andWilliams are strikingly inconsistentwith the assumption that verbalabilities, particularly vocabulary, areless affected by mental illness thanother abilities. However, the findingsof a similar study by Schwartzmanand Douglas (1962) are consistentwith the assumption that verbalabilities are less affected.

Schwartzman and Douglascompared subtest scores from theCanadian Army "M" tests forhospitalized schizophrenic patients,discharged schizophrenic patients,and normal veterans. For each


Dow

nloaded from



patient, scores on the "M" tests wereobtained from the time of armyinduction and at postmorbidretesting. The "M" test is dividedinto three main parts: nonverbal,mechanical knowledge, and verbalabilities. The nonverbal sectionconsists of three tests: picturecompletion, picture absurdities, andpaper formboard. The twomechanical tests are tool recognitionand mechanical information. Theverbal section consists-of arithmetic,vocabulary, and verbal analogies.The control subjects manifested nosignificant losses at retesting. Incontrast, the patients showedsignificant decreases betweenpremorbid and postmorbid testing onall subtests except paper formboard,vocabulary, and verbal analogies; theex-patients showed significant lossonly on tool recognition andmechanical information.

Thus, when examining premorbid-postmorbid subtest patterns, theSchwartzman and Douglas studylends support to the assumption thatschizophrenic patients' verbal abilitiesremain intact while other abilitiesdecline. But the results of the studyby Lubin, Gieseking, and Williams(1962) contradict this assumption.There are at least two possibleexplanations for these contradictoryfindings. First, the inconsistent resultsmay be due to differences in themeasures used in these studies. Sincethe tests used by both Schwartzmanand Douglas, and Lubin, Giesekingand Williams were strictly timed andcontained similar content, it isunlikely that measurement differencesare the crucial factor. Anotherpossible source of the inconsistency isthe difference in premorbid-postmorbid test intervals. Lubin,Gieseking, and Williams administeredthe postmorbid test 1 to 3 monthsafter hospitalization, whereasSchwartzman and Douglas tested

their patients more than 6 to 12months after hospitalization. Thus itmay be that all abilities show adecline with the onset of schizo-phrenia, but that after 6 or moremonths, premorbid verbal abilitiesare regained.

The findings from studies thathave looked at subtest patternchanges over the course of the illnessare also contradictory. Smith (1964)examined subtest changes of youngnonparanoid schizophrenic and olderparanoid schizophrenic patients onthe Wechsler-Bellevue. Patients hadbeen hospitalized for a mean of 8-9years before the initial tests and wereretested 8.4 years later. The youngerschizophrenic patients showed slightlosses on the Object Assemblysubtest. The older group showedlosses on Information, Compre-hension, Similarities, Vocabulary,Picture Arrangement, PictureCompletion, and Block Design. Usingthese same subjects, Hamlin (1969)found that the patterns of changeover 14 years for Verbal andPerformance scores were generallythe same, and this held for both theyounger nonparanoid and the olderparanoid schizophrenic patients.Hamlin concludes that "deteriorationdoes not occur in some intellectualfunctions, with other functionsremaining intact" (p. 502).

In contrast, Klonoff, Fibiger, andHutton (1970) found that VIQs ofschizophrenics were more stable thanPIQs. These authors compared WAISscores upon hospital admission withscores from the same test admin-istered 8 years later. VIQs showedsignificantly less change over theyears than PIQs. As with the studieson premorbid-postmorbid changes,the inconsistencies in the findings ofthe studies of Smith (1964) andKlonoff, Fibiger, and Hutton (1970)may be due to differences in thetimes of testing. In this case, the

studies differ in the times of initial aswell as followup testing.

To date, the results of studies thathave attempted to determine whethersubtest patterns are related tochronicity or schizophrenic subtypemust be considered inconclusive.There have been no reportedattempts to replicate most of thefindings presented here. However,the issue of subtest patterns hasgained increasing relevance in light ofrecent theories regarding unilateralhemispheric dysfunction in schizo-phrenia (Walker and McGuire 1982).

Summary and Conclusions

This review has focused on researchspanning the past two decades inorder to address five major questionsabout the relationship betweenintelligence and schizophrenia.Although there is sufficient evidenceto answer some of the questions withconfidence, others must await theresults of future research. Weattempt, here, to summarize what wedo and do not know.

There is strong evidence thatdiagnosed schizophrenic patientsscore lower on standardized measuresof intelligence than would bepredicted from family and envi-ronmental variables. Moreover,premorbid scores of schizophrenicpatients, obtained during childhood,adolescence, and early adulthood,are lower than the scores of theirsiblings and peers with similar socialclass origins. Thus IQ deficits appearto be associated with schizophreniaacross the lifespan. However, theresults of our meta-analysis indicatethat premorbid IQ deficit may be anexclusive, or more pronounced,characteristic of schizophrenic males.The data also provide some supportfor the hypothesis that IQ deficits ofpreschizophrenic children, relative to


Dow

nloaded from


VOL. 10, NO. 3,1984 455

their siblings, are less pronouncedwhen the average IQ score for thesiblings is higher. Given the likelyhigher SES of the families of thesechildren, the results are consistentwith the notion that environmentalfactors can ameliorate cognitivedeficits (Rubin and Balow 1979).

There does not appear to be aprodromal decline in IQ for schizo-phrenic patients. That is, althoughpremorbid IQ may be lower thanwould be expected from family andenvironmental variables, it does notappear to drop before the overtappearance of schizophrenicsymptoms. Results from studies thathave examined changes betweenpremorbid and postmorbid IQs areequivocal. It is not known, then,whether deterioration in IQ is aconcomitant of the overt manifes-tation of schizophrenia. However, itcan be concluded that continualdeterioration in intellectualfunctioning during the course of theillness is by no means a necessarycharacteristic of schizophrenia.Furthermore, when increases in IQdo occur during the course of theillness, they are usually accompaniedby improvement in psychiatricsymptoms.

Higher IQs obtained at hospitaladmission or during the course ofhospitalization are associated withgreater clinical improvement, betteroutcome ratings, shorterhospitalization, later onset of illness,and better heterosexual adjustment.It is not clear whether premorbid IQscores are equally related to theseoutcome indicators. It is possible thatthe relationship between premorbidIQ and outcome was not detected bysome investigators because of therestricted range of IQ scoresrepresented in their samples (e.g.,Schwartzman, Douglas, and Muir1962).

Verbal IQ-Performance IQ discrep-

ancies and intertest variability appearto be greater for schizophrenicpatients than for normals. However,these indicators are not unique toschizophrenia, as other psychiatricand brain-damaged patients showsimilar patterns. It should also benoted that the consistent tendency forschizophrenic patients to manifesthigher VIQs than PIQs would seemto weaken the analogy drawnbetween schizophrenic patients andpatients with left hemisphere lesions(Flor-Henry 1976).

Along these same lines, it has notbeen established that any particularpattern of subtest scores canconsistently distinguish amongschizophrenics who differ on suchcharacteristics as length ofhospitalization, age, neurologicaldamage, existence of paranoidsymptoms, and reactive vs. process,chronic vs. acute, and discharged vs.nondischarged status. Nor do thereappear to be consistent changes insubtest patterns between premorbidand postmorbid testing or betweentests administered at hospitaladmission and tests administeredafter several years of hospitalization.These results may be interpreted asindicating that schizophrenia isassociated with generalized deficits incognitive functions. Alternatively, itmay be that the measures used in theresearch, to date, were not capable ofdetecting the specific pattern ofintellectual impairment associatedwith schizophrenia.

In summary, then, it appears thatat least some cases of schizophreniaare associated with moderateintellectual deficits across thelifespan, that prognosis is related toIQ, and that variations in sympto-matology are associated with changesin IQ. This pattern of results isinterpretable from at least threeperspectives. First, intellectual abilitycan be viewed as an independent

mediating factor: High IQ mayreduce susceptibility to schizophrenicbreakdown, and enhance thelikelihood of remission in those whodevelop the disorder. Thus,intellectual ability may determine theindividual's capacity to cope with theenvironmental stressors thatprecipitate breakdown or worsenprognosis. From this perspective, thefluctuations in IQ that accompanysymptomatological changes might beattributed to transient alterations inmotivational state. However, thehypothesis that IQ is an independent,mediating factor is not sufficient toaccount for the apparent linkbetween premorbid IQ and gender,or the IQ deficits noted in siblings ofschizophrenic patients. As a secondapproach, intellectual deficit can beviewed as one manifestation of agenetically determined constitutionalvulnerability to schizophrenia. Thisinterpretation is compatible with thefinding of IQ deficits in siblings ofschizophrenics. Individual differencesin premorbid and postmorbid IQwould therefore reflect individualdifferences in degree of liabilitypredisposition, as well as episodicfluctuations in symptomatology.Third, if etiologic heterogeneityexists, IQ deficit may be associatedwith a specific subtype of schizo-phrenia. Acquired central nervoussystem dysfunction may be theprimary determinant of this subtype,thus resulting in a higher incidence ofpremorbid IQ deficit in males.Finally, it is possible that all threeperspectives have merit: Intellectualcapacity may be partially determinedby acquired or inherited consti-tutional vulnerability to schizo-phrenia, but at the same time act tomediate the effects of stressors on thevulnerable individual. Obviously,differentiating these effects is not aneasy task. .

Several issues deserve attention in


Dow

nloaded from



future research. One question oftremendous importance is whetherenvironmental enhancement of intel-lectual abilities can serve to mitigatevulnerability in individuals presumedto be constitutionally predisposed. Asecond area warranting furtherinquiry is schizophrenic subgroupdifferences in patterns of intellectualfunctioning. A few studies haveaddressed this issue; however, theyhave tended not to focus on the samesubgroups. Consequently, there is noreplicative support for the findingsthat have been reported to date.Studies of IQ using newer subgroupdistinctions, such as positive versusnegative symptom patients, may beilluminating. Finally, an extremelyimportant issue that has yet to bedirectly addressed is the possible roleof sex differences. Several studieshave anecdotally reported sexdifferences in premorbid IQ; yet,there has been no systematicinvestigation of sex differences in theintelligence of schizophrenic patientsor preschizophrenic children. Theresults of our correlational analysesindicate that the proportion of malesin the samples of preschizophrenicand high-risk subjects is a significantpredictor of the effect size obtained.Combined with the demonstratedassociation between sex and age atonset of schizophrenia (Lewine,1981), these results point to thepossibility of divergent etiologicalsubtypes of schizophrenia in malesand females. Specifically, it ispossible that there is a subtype ofschizophrenia that is characterized bypremorbid intellectual deficits and towhich males are more vulnerablethan females.

References

Albee, G.; Lane, E.; Corcoran, C;and Werneke, A. Childhood and

intercurrent intellectual performanceof adult schizophrenics. Journal ofConsulting Psychology, 27:364-366,1963.

Albee, G.; Lane, E.; and Reuter, J.Childhood intelligence of futureschizophrenics and neighborhoodpeers. Journal of Psychology,58:141-144, 1964.

American Psychiatric Association.DSM-Ii. Diagnostic and StatisticalManual of Mental Disorders. 2nd ed.Washington, DC: The Association,1968.

American Psychiatric Association.DSM-III: Diagnostic and StatisticalManual of Mental Disorders. 3rd ed.Washington, DC: The Association,1980.

Andreasen, N., and Olsen, S.Negative versus positive symptomschizophrenia: Definition andvalidation. Archives of GeneralPsychiatry, 39:789-794, 1982.

Babcock, H., and Levy, L. Manual ofDirections for the RevisedExamination of the Measurement ofEfficiency of Mental Functioning.Chicago: Stoelting, 1940.Belmont, I.; Birch, H.; Klein, D.L.;and Pollack, M. Perceptual evidenceof CNS dysfunction in schizophrenia.Archives of General Psychiatry,10:395-408, 1964.

Binder, A. Schizophrenic intellectualimpairment: Uniform or differential?Journal of Abnormal and SocialPsychology, 52:11-18, 1956.

Birren, J.E. Psychological examina-tions of children who later becamepsychotic. Journal of Abnormal andSocial Psychology, 38:84-96, 1944.

Bower, E.; Shellhamer, T.; andDaily, J. School characteristics ofmale adolescents who later becameschizophrenic. American Journal ofOrthopsychiatry, 30:712-729, 1960.

Braff, D., and Saccuzzo, D. Effects

of antipsychotic medication on speedof information processing in schizo-phrenic patients. American Journal ofPsychiatry, 139:1127-1130, 1982.

Brody, M.B. A survey of the resultsof intelligence tests in psychoses.British Journal of MedicalPsychology, 19:215-261, 1942.

Carp, A. Performance on theWechsler-Bellevue Scale and insulinshock therapy. Journal of Abnormaland Social Psychology, 45:127-136,1950.

Chapman, L., and Chapman, J.Problems in the measurement ofcognitive deficit. PsychologicalBulletin, 79:380-385, 1973.

Chelune, G.; Heaton, R.; Lehman,R.; and Robinson, A. Level versuspattern of neuropsychologicalperformance among schizophrenicand diffusely brain-damaged patients.Journal of Consulting and ClinicalPsychology, 47:155-163, 1979.

Cohen, J. Statistical Power Analysisfor the Behavioral Sciences. NewYork: Academic Press, 1977.

Cohler, B.; Grunebaum, H.; Weiss,J.; Gamer, E.; and Gallant, D.Disturbance of attention amongschizophrenic, depressed and wellmothers and their young children.Journal of Child Psychology andPsychiatry, 18:115-135, 1977.

Cornell, E., and Lowden, G. Acomparison of the Stanford andPorteus tests in several types ofsocial inadequacy. Journal ofAbnormal and Social Psychology,18:33-42, 1923.

Davidson, M. Studies in theapplication of mental tests topsychotic patients. British Journal ofMedical Psychology, 18:44-52, 1939.

Davie, R.; Butler, N.; and Goldstein,H. From Birth to Seven. London:Longman Group Limited, 1972.

Davis, W.; DeWolfe, A.; and


Dow

nloaded from


VOL.10, NO. 3,1984 457

Gustafson, R. Intellectual deficit inprocess and reactive schizophrenicsand brain injury. Journal ofConsulting and Clinical Psychology,38:146, 1972.Davis, W., Dizzonne, M.; andDeWolfe, A. Relationships amongWAIS subtest scores, patient'spremorbid history, and institu-tionalization. Journal of Consultingand Clinical Psychology, 36:400-403,1971.

DeWolfe, A.; Barrell, R.; Becker, B.;and Spaner, F. Intellectual deficit inchronic schizophrenia and braindamage. Journal of Consulting andClinical Psychology, 36:197-204,1971.

Drillien, C. The Growth andDevelopment of the PrematurelyBorn Infant. Baltimore: Williams &Wilkins Company, 1964.

Erlenmeyer-Kimling, L.; Kestenbaum,C; Bird, H.; and Hilldoff, U."Assessment of the New York HighRisk Project Subjects in Sample AWho Are Now Clinically Deviant."Presented at the Risk ResearchConsortium Plenary Conference, SanJuan, Puerto Rico, 1980.

Flor-Henry, P. Lateralized temporal-limbic dysfunction and psycho-pathology. Annals of the New YorkAcademy of Sciences, 280:777-795,1976.

Frazee, H. Children who laterbecome schizophrenic. Smith CollegeStudies in Social Work, 23:125-149,1953.

Gilgash, C.A. Thorazine therapywith catatonic schizophrenics inrelation to Wechsler Verbal andPerformance subtests comparison.Journal of Clinical Psychology,17:95, 1961.

Glass-, G.V. Summarizing effect sizes.In: Rosenthal, R., ed. NewDirections for Methodology of Socialand Behavioral Science. San

Francisco: Jossey Bass, 1980.Goldstein, G., and Halperin, K.Neuropsychological differencesamong subtypes of schizophrenia.Journal of Abnormal and SocialPsychology, 86:34-40, 1977.

Gottesman, 1.1., and Shields, J.Schizophrenia: The EpigeneticPuzzle. New York: CambridgeUniversity Press, 1982.

Griffith, J.; Mednick, S.; Schulsinger,F.; and Diderichsen, B. Verbalassociative disturbances in children athigh risk for schizophrenia. Journalof Abnormal Psychology,89:125-131, 1980.

Grunebaum, H.; Cohler, B.J.;Kaufman, C ; and Gallant, D.Children of depressed and schizo-phrenic mothers. Child Psychiatryand Human Development,8:219-228, 1978.

Hamlin, R. The stability ofintellectual function in chronicschizophrenia. Journal of Nervousand Mental Disease, 149:495-503,1969.Hamlin, R.M., and Ward, W.D.Aging, hospitalization, and schizo-phrenic intelligence. AmericanPsychologist, 20:491-492, 1965.

Hanson, D.R.; Gottesman, 1.1.; andHeston, L.L. Some possiblechildhood indicators of adult schizo-phrenia inferred from children ofschizophrenics. British Journal ofPsychiatry, 129:142-154, 1976.

Haywood, H., and Moelis, I. Effectof symptom change on intellectualfunction in schizophrenia. Journal ofAbnormal Psychology, 67:76-78,1963.Heffner, P.; Strauss, M.; and Grisell,J. Rehospitalization of schizophrenicsas a function of intelligence. Journalof Abnormal Psychology,84:735-736, 1975.

Holland, T.; Levi, M.; and Watson,C. Multivariate structure of

associations between verbal andnonverbal intelligence among brain-damaged, schizophrenic, neuroticand alcoholic patients. Journal ofAbnormal Psychology, 88:354-360,1979.Holland, T., and Watson, C. Multi-variate analysis of WAIS-MMPIrelationships among brain-damaged,schizophrenic, neurotic, andalcoholic patients. Journal of ClinicalPsychology, 36:352-359, 1980.

Hunt, H. A practical clinical test fororganic brain damage. Journal ofApplied Psychology, 27:375-386,1943.Hunt, J., and Cofer, C.N. Psycho-logical deficit. In: Hunt, J., ed.Personality and the BehaviorDisorders. New York: The RonaldPress Co., 1944.

Jastak, J. Psychometric patterns ofstate hospital patients. DelawareState Medical Journal, 9:7-10, 1937.Jones, M. IQ and fertility in schizo-phrenia. British Journal ofPsychiatry, 122:689-696, 1973.Jones, M., and Offord, D.Independent transmission of IQ andschizophrenia. British Journal ofPsychiatry, 126:185-190, 1975.

Kendig, I., and Richmond, W.Psychological Studies in DementiaPraecox. Ann Arbor, MI: Edward,1940.

Kissel, S. Schizophrenic patterns onthe WISC: A missing control.Journal of Clinical Psychology,22:201, 1966.

Klonoff, H.; Fibiger, C ; and Hutton,G.H. Neuropsychological patterns inchronic schizophrenia. Journal ofNervous and Mental Disease,150:291-300, 1970.

Landau, R.; Harth, P.; Othnay, N.;and Sharfhertz, C. The influence ofpsychotic parents in their childrens'development. American Journal ofPsychiatry, 129:70-75, 1972.


Dow

nloaded from



Lane, E., and Albee, G. Childhoodintellectual development of adultschizophrenics. Journal of Abnormaland Social Psychology, 67:186-189,1963.

Lane, E., and Albee, G. Earlychildhood intellectual differencesbetween schizophrenic adults andtheir siblings. Journal of Abnormaland Social Psychology, 68:193-195,1964.

Lane, E., and Albee, G. Childhoodintellectual differences betweenschizophrenic adults and theirsiblings. American Journal of Ortho-psychiatry, 35:747-753, 1965.

Lane, E., and Albee, G. Onchildhood intellectual decline of adultschizophrenics: A reassessment of anearlier study. Journal of Abnormaland Social Psychology, 73:174-177,1968.

Lane, E., and Albee, G. Correlationsbetween the IQs of schizophrenicparents and their children. ThePsychological Record, 20:81-82,1970.

Lane, E.; Albee, G.; and Doll, L. Theintelligence of children of schizo-phrenics. Developmental Psychology,2:315-317, 1970.Lehman, R.; Chelune, G.; andHeaton, R. Level and variability ofperformance on neuropsychologicaltests. Journal of Clinical Psychology,35:358-363, 1979.

Lewine, R. Sex differences in schizo-phrenia: Timing or subtypes?Psychological Bulletin, 90:432-444,1981.

Lubin, A.; Gieseking, C ; andWilliams, H. Direct measurement ofcognitive deficit in schizophrenia.Journal of Consulting Psychology,26:139-143, 1962.

Mason, C. Pre-illness intelligence ofmental hospital patients. Journal ofConsulting Psychology, 20:297-300,1956.

Michaels, J., and Schilling, M.E. Thecorrelations of the intelligencequotients of the Porteus Maze andBinet-Simon tests in 200 neuropsychi-atric patients. American Journal ofOrthopsychiatry, 6:71-74, 1936.

Miner, J., and Anderson, J.Intelligence and emotionaldisturbance: Evidence from Armyand Veterans Administration records.Journal of Abnormal and SocialPsychology, 56:75-81, 1958.

Of ford, D. School performance ofadult schizophrenics, their siblings,and age mates. British Journal ofPsychiatry, 125:12-19, 1974.

Offord, D., and Cross, L. Adultschizophrenia with scholastic failureor low IQ in childhood. Archives ofGeneral Psychiatry, 24:431-435,1971.

Oltmanns, T.F.; Weintraub, S.;Stone, A.A.; and Neale, J.M.Cognitive slippage in childrenvulnerable to schizophrenia. Journalof Abnormal Child Psychology,6:237-245, 1978.

Parker, R., and Davidson, N. Acomparison of students of nursingand hospitalized patients on scoresderived from an intelligence test(WAlSf. Psychiatric QuarterlySupplement, 37:298-306, 1963.

Payne, R. Cognitive abnormalities.In: Eysenck, H., ed. Handbook ofAbnormal Psychology. London:Pitman Medical Publishing Co., Ltd.,1960.

Pollack, M. Comparison ofchildhood, adolescent, and adultschizophrenias. Archives of GeneralPsychiatry, 2:652-659, 1960.Pollack, M.; Levenstein, S.; andKlein, D.F. A three-year posthospitalfollow up of adolescent and adultschizophrenics. American Journal ofOrthopsychiatry, 38:94-109, 1968.

Pollack, M.; Woerner, M.; andKlein, D.F. A comparison of

childhood characteristics of schizo-phrenics, personality disorders, andtheir siblings. In: Roff, M., andRicks, D., eds. Life History Researchin Psychopathology. Minneapolis:University of Minnesota Press, 1970.

Pollin, W., and Stabenau, J.Biological, psychological andhistorical differences in series ofmonozygotic twins discordant forschizophrenia. In: Rosenthal, D.,and Kety, S.S., eds. TheTransmission of Schizophrenia.London: Pergamon Press, 1968.

Pressey, S.L. Distinctive features inpsychological test measurementsmade upon dementia praecox andchronic alcoholic patients. Journal ofAbnormal and Social Psychology,12:130-139, 1917.

Rabin, A. Fluctuations in the mentallevel of schizophrenic patients.Psychiatric Quarterly, 18:78-91,1944.

Rappaport, S., and Webb, W. Anattempt to study intellectual deteri-oration by premorbid and psychotictesting. Journal of ConsultingPsychology, 14:95-98, 1950.Rieder, R.; Broman, S.; andRosenthal, D. The offspring ofschizophrenics. Archives of GeneralPsychiatry, 34:789-799, 1977.

Roe, A., and Shakow, D. Intelligencein mental disorder. Annals of theNew York Academy of Sciences,42:361-490, 1942.

Roff, J., and Knight, R. Preschizo-phrenics: Low IQ and aggressivesymptoms as predictors of adultoutcome and marital status. Journalof Nervous and Mental Disease,168:129-132, 1980.

Roff, J.; Knight, R.; and Wertheim,E. Disturbed preschizophrenics.Journal of Nervous and MentalDisease, 162:274-281, 1976a.

Roff, J.; Knight, R.; and Wertheim,E. A factor-analytic study of


Dow

nloaded from


VOL. 10, NO. 3,1984 459

childhood symptoms antecedent toschizophrenia. Journal of AbnormalPsychology, 85:543-549, 1976b.

Rosenthal, R. Combining results ofindependent studies. PsychologicalBulletin, 85:185-193, 1978.Rubin, R., and Balow, B. Measuresof infant development and socio-economic status as predictors of laterintelligence and school achievement.Developmental Psychology,15:225-227, 1979.

Schaffner, A.; Lane, E.; and Albee,G. Intellectual differences betweensuburban preschizophrenic childrenand their siblings. Journal ofConsulting Psychology, 31:326-327,1967.

Schoonover, S., and Hertel, R.Diagnostic implications of WISCscores. Psychological Reports,26:967-973, 1970.Schwartzman, A., and Douglas, V.Intellectual loss in schizophrenia:Part I. Canadian Journal ofPsychology, 16:1-10, 1962.Schwartzman, A.; Douglas, V.; andMuir, R. Intellectual loss in schizo-phrenia: Part II. Canadian Journal ofPsychology, 16:161-168, 1962.Shipley, W. A self-administeringscale for measuring intellectualimpairment and deterioration.Journal of Psychology, 9:371-377,1940.

Smith, A. Mental deterioration inchronic schizophrenia. Journal ofNervous and Mental Disease,139:479-487, 1964.

Spohn, H.; Lacoursiere, R.; andThompson, K. Phenothiazine effectson psychological and psychophysio-logical dysfunction in chronic schizo-phrenia. Archives of GeneralPsychiatry, 34:633-644, 1977.

Stotsky, B. A comparison ofremitting and nonremitting schizo-phrenics on psychological tests.Journal of Abnormal and SocialPsychology, 47:489-496, 1952.Walker, E., and Emory, E. Infants atrisk for psychopathology: Offspringof schizophrenic parents. ChildDevelopment, 54:113-125, 1983.

Walker, E., and McGuire, M. Intraand interhemispheric informationprocessing in schizophrenia. Psycho-logical Bulletin, 92:701-725, 1982.

Watson, C. WAIS profile patterns ofhospitalized brain-damaged andschizophrenic patients. Journal ofClinical Psychology, 21:294-295,1965. •

Watt, N.; Grubb, T.; andErlenmeyer-Kimling, L. Social,emotional, and intellectual behaviorat school among children at high riskfor schizophrenia. Journal ofConsulting and Clinical Psychology,50:171-181, 1982.

Watt, N., and Lubensky, A.Childhood roots of schizophrenia.Journal of Consulting and ClinicalPsychology, 44:363-375, 1976.

Wechsler, D. The Measurement ofAdult Intelligence. 4th ed.Baltimore: Williams & WilkinsCompany, 1958.Wechsler, D., and Jaros, E. Schizo-phrenic patterns on the WISC.Journal of Clinical Psychology,21:288-291, 1965.

Wells, F., and Kelley, C. Intelligenceand psychosis. American Journal ofInsanity, 77:17-45, 1920.

Wentworth, M. Two hundred casesof dementia praecox tested by theStanford Revision. Journal ofAbnormal and Social Psychology,18:378-384, 1924.

Winder, C. Some psychologicalstudies of schizophrenics.In: Jackson, D., ed. The Etiology ofSchizophrenia. New York: BasicBooks, 1960.Winters, K.; Stone, A.; Weintraub,S.; and Neale, J. Cognitive and atten-tional deficits in children vulnerableto psychopathology. Journal ofAbnormal Child Psychology,9:435-453, 1981.Worland, J., and Hesselbrock, V.The intelligence of children and theirparents with schizophrenia andaffective illness. Journal of ChildPsychology and Psychiatry,21:191-201, 1980.

Worland, J.; Weeks, D.; Weiner, S.;and Schechtman, J. Longitudinal,prospective evaluations of children atrisk. Schizophrenia Bulletin,8:135-141, 1982.

Zubin, J., and Steinhauer, S. How tobreak the logjam in schizophreniaresearch. Journal of Nervous andMental Disease, 169:477-492, 1981.

Acknowledgments

The writing of this article wassupported in part by grants from theScottish Rite Foundation and theW.T. Grant Foundation. Special

thanks to Vicki Griffin for her expertassistance.

The Authors

Elizabeth Aylward, Ph.D., ElaineWalker, Ph.D., and Barbara Bettes,B.A., are in the Department ofHuman Development and FamilyStudies, Cornell University, Ithaca,NY.


Dow

nloaded from


Intelligence in Schizophrenia: Meta-analysis of the Research

Documents