Schizophrenia Research 65 (2003) 105 – 116

www.elsevier.com/locate/schres

Basic neuropsychological dimensions in schizophrenia

Judith Jaeger a,b,*, Pál Czobor c, Stefanie M. Berns a
a
Department of Psychiatric Rehabilitation, Center for Neuropsychiatric Outcome and Rehabilitation Research (CENORR),
Zucker Hillside Hospital, North Shore – Long Island Jewish Health System, Glen Oaks, NY, USA
b
Albert Einstein College of Medicine, USA
c
Nathan S. Kline Institute, Orangeburg, NY 10962, USA
Received 20 September 2002; received in revised form 26 December 2002; accepted 16 January 2003

Abstract

Neuropsychological (NP) studies in schizophrenia often require data reduction to avoid statistical type I error from multiple
comparisons. Typically, this involves grouping measures into domains defined by experts a priori based on delineations
validated in brain-injured but not schizophrenia samples (e.g. attention, executive functioning, memory, language visuospatial,
motor). Component measures are arbitrarily selected and validity may not generalize to different samples or within the same
sample over time.
One solution to these problems is illustrated using neurocognitive subdomains based on recent schizophrenia literature, and
validated with data from a longitudinal study (156 subjects) involving repeated NP testing (baseline—within 6 months of
hospital discharge—and 6 and 18 months later). A priori subdomains were grouped and submitted to principal component
analysis (PCA) at each time point. Longitudinal stability of the resulting factors was tested by computing congruency
coefficients. Six stable factors were extracted having good construct, divergent and predictive validity. Five neuropsychological
measures frequently studied in schizophrenia were not correlated with these factors, suggesting that they should be maintained
as independent neurocognitive subdomains. Distinct factors for executive functioning, verbal memory and motor functions
could not be validated; this raises concerns about conclusions of previous studies regarding the pattern, severity and correlates
of specific neurocognitive functions in schizophrenia.
D 2003 Elsevier Science B.V. All rights reserved.

Keywords: Neuropsychological domains; Neurocognitive subdomain; Schizophrenia

1. Introduction in schizophrenia has grown steadily in the last three

The use by researchers of neuropsychological
(NP) test batteries to profile neurocognitive deficits
* Corresponding author. Department of Psychiatric Rehabilita-
tion, Center for Neuropsychiatric Outcome and Rehabilitation
Research (CENORR), Zucker Hillside Hospital, North Shore—
Long Island Jewish Health System, 75 – 59 263rd Street, Glen Oaks,
NY 11004, USA. Tel.: +1-718-470-8342; fax: +1-718-962-2742.
E-mail address: jaeger@lij.edu (J. Jaeger).
decades. Flor-Henry and Yeudall were among the
first in the post-Kraepelinian era to employ NP
batteries, previously designed to examine localization
of function in focal brain-damaged populations, in an
effort to reveal pathophysiologic substrates of the
endogenous psychoses (Flor-Henry and Yeudall,
1979; Flor-Henry, 1969, 1976). Heaton et al. (1978)
further emphasized the importance of NP tests for
revealing neurocognitive deficits in the endogenous

0920-9964/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0920-9964(03)00052-5
106 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116
psychoses, concluding that schizophrenia patients had
deficits comparable to those seen in brain-injured
patients. These two influential bodies of work, to-
gether with early brain imaging studies reporting
abnormal brain structure and function, led a number
of neuropsychologists to employ NP test batteries in
studies of schizophrenia. Early findings validated this
effort, demonstrating marked NP test deficits that
discriminated more severe state hospital patients from
ambulatory samples (Perlick et al., 1992). Imaging
studies subsequently demonstrated region-specific
functional activation with one widely used NP test,
the Wisconsin Card Sorting Test (Weinberger et al.,
1986). Studies suggesting that NP measures are
associated with functional disability (reviewed by
Green, 1996) recently fueled a trend in the pharma-
ceutical industry to target NP test performance as a
proxy for a drug’s potential to improve independent
functioning. These forces, combined with a marked
increase in the percentage of psychologists specializ-
ing in neuropsychology, have led to the current
almost ubiquitous use of NP tests in clinical studies
of schizophrenia.
The research strategy most widely used for discern-
ing patterns of neurocognitive deficit in schizophrenia
has been to administer a comprehensive NP battery
consisting of several measures tapping each of several
putative neurocognitive domains (e.g. Saykin et al.,
1991; Bilder et al., 2000; White et al., 1997; Hoff et al.,
2001; Heaton et al., 2001; Blanchard and Neale, 1994;
Censits et al., 1997). Such batteries typically yield an
impractically large number of variables. Arguments
have been made for reducing the length of the batteries
used; however, these generally center on the cost and
subject burden (Gold et al., 1999) rather than more
basic theoretical grounds. Redundancy in NP testing is
valued as an important method for assuring that a few
spurious test scores do not yield false conclusions
regarding the functioning of a particular cognitive
domain (Lezak, 1995). However, comprehensive bat-
teries yield many variables, raising the risk of statistical
type I error, and the need for data reduction. Ideally, test
scores would be aggregated into reliable indices each
reflecting the integrity of a single neurocognitive
domain.
The most frequently used method for NP data
reduction in the literature involves grouping test
measures into conventional domains (e.g. viz., at-
tention, memory, executive functioning, language,
visuospatial, motor) and averaging the individual
standardized measures within each domain. Howev-
er, this approach is problematic: The domains tradi-
tionally used derive from delineations valid in focal
brain-injured patients, but not necessarily patients
with schizophrenia. For example, schizophrenia
studies have revealed that memory, executive func-
tioning and attention are overlapping and not neces-
sarily valid independent constructs (Gold et al.,
1997). In practice, the tests within each domain
are often arbitrarily selected and open to debate.
Indeed our systematic analysis of key studies
employing this technique revealed surprising differ-
ences in both the domain grouping of NP measures
and the designation of domains themselves. For
example, Trails B, one of the most commonly used
NP measures, has been variously grouped under
‘‘Visuomotor’’ (Cuesta and Peralta, 1995), ‘‘Atten-
tion’’ (Censits et al., 1997), ‘‘Executive’’ (Bilder et
al., 2000), ‘‘Abstraction and Cognitive Flexibility’’
(Heaton et al., 2001), and ‘‘Concentration/Speed’’
(Hoff et al., 1999). These differences cannot be
accounted for by differences in the domains desig-
nated by each investigator: The domain of ‘‘Abstrac-
tion/Cognitive Flexibility’’ was distinguished in
reports by Saykin et al. (1991, 1994), Censits et
al. (1997), and Heaton et al. (2001), all of which
included Trails B in their battery, but only one of
which grouped this measure within this domain
(Heaton et al., 2001) while others instead grouped
it with a combined domain called ‘‘Visual Motor
Processing/Attention’’ (Saykin et al., 1991, 1994) or
‘‘Attention’’, as cited above.
Another concern is that even if distinguishable
domains are validated, this does not guarantee the
validity of the derived domain measures in a sample
selected differently (e.g. inpatients/outpatients, first-
episode/chronic) or the same sample at different time
points. Covariance matrices may differ between dif-
ferent samples or within the same sample over time.
For example, if the range on a particular measure is
truncated in one group, but is widely distributed in a
second group, the relationship between that variable
and other measures in the test battery may differ
between the two groups.
This report seeks to establish a valid data reduction
method for neuropsychological research in schizo-
 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116
phrenia. NP variable groupings, formed based on the
recent schizophrenia literature, are studied for their
validity and replicability.
2. Method
2.1. Subjects
Subjects are consenting patients in an ongoing 3-
year study of schizophrenia and schizoaffective dis-
order (diagnosed using SCID for DSM-IV) involving
repeated NP testing. Subjects are enrolled within 6
months of symptom exacerbation requiring hospitali-
zation, and received a comprehensive NP test battery
and Positive and Negative Symptom Scale (PANSS)
(Kay et al., 1987) ratings at baseline and again after 6,
18 and 36 months. Staff administering NP tests were
previously trained and observed in test battery admin-
istration to assure uniformity. PANSS raters had
demonstrated interrater reliability compared to an
expert (ICCs z 0.80). The Multidimensional Scale
for Independent Functioning (MSIF) developed and
validated in our laboratory (Jaeger et al., in press) was
administered monthly to capture recovery of life
functions. The MSIF distinguishes role responsibili-
ties from support received, and quality of perfor-
mance in work, school and residential roles, and
provides global ratings.
For the present analyses, the dataset from this
ongoing study (N = 250) was frozen. Subjects were
included if at that time, they had completed the
baseline NP assessment and at least one follow-up
NP assessment at 6 (N = 146) or 18 months (N = 124).
A total of 156 subjects was studied. Mean age at
baseline was 36.3 years (S.D. = 9.2), 41% were fe-
male, 65 (41.7%) were white, 62 (39.7%) African
American, 22 (14.1%) Hispanic, 2 (1.3%) Asian and 5
were ethnically classified as ‘‘other’’ (3.2%). Sixty-
four cases (41%) of the sample had schizoaffective
disorder. Age at first symptom presentation averaged
18.7 years (S.D. = 7.0) and 16.2 years (S.D. = 9.0) had
elapsed since first psychiatric treatment. Education
averaged 12 years (S.D. = 2.4) [49 did not finish high
school (HS), 72 had a HS diploma, 17 completed a
GED, 2 had an Associate’s degree, 15 completed a
bachelor’s degree, and one had completed graduate
school. Education was missing for one case].
107
With respect to treatment course, at the time of the
baseline assessment, an average of 81.5 days
(S.D. = 56) had passed since discharge from the index
hospitalization. Two subjects were not taking psycho-
active medications and six were on psychoactive
medications other than an antipsychotic. Of the 148
subjects taking antipsychotics, 59 were exclusively on
a new generation compound, 44 were on both old and
new generation compounds, and 45 were only on a
conventional drug. As a whole, psychopathology
ratings at baseline were in the mild range. Average
scores for each factor using the five-factor solution for
the PANSS of Bell et al. (1994) were 2.6 (S.D. = 8.2)
for positive symptoms, 2.4 (0.82) for negative symp-
toms, 2.8 for the cognitive factor, 2.9 (0.99) for
emotional discomfort and 2.1 (0.71) for hostility.
Ranges for these factors were from 1 to 5 and medians
ranged from 2.0 for hostility to 2.8 for emotional
discomfort. (PANSS items are rated on a seven-point
Likert scale with a 7 indicating ‘extreme’, 5 ‘moderate
to severe’, 3 ‘mild’ and 1 ‘absent’.)
2.2. Procedures
The first step was to a priori identify neuro-
cognitive subdomains based on recent schizophrenia
literature, refining the conventional NP domains
(see Table 1). A series of iterative procedures
ensued to identify the constituent NP measures,
Table 1
Conventional neuropsychological domains and neurocognitive
subdomains often distinguished in schizophrenia
Conventional Neurocognitive subdomains
neuropsychological for schizophrenia
domains
. Attention . Sustained Vigilance
. Short-Term Memory
Capacity/Span
. Executive Fx . Working Memory
. Set Shifting/
Cognitive Flexibility
. Ideational Fluency
. Memory . Verbal Learning
. Non-Verbal Learning
. Language . Verbal Knowledge
. Visuospatial . Non-Verbal Reasoning/
Problem Solving
. Motor . Motor
108 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116
and determine the validity and stability of the
resultant factors. All analytic procedures were con-
ducted first in the baseline dataset. The replicability
of valid subdomains was then evaluated using
follow-up datasets.
As mentioned above, neuropsychological test bat-
teries typically yield datasets with a large number of
variables, obtained on a relatively small number of
subjects. Due to the high number of variables, and to
the unfavorable case per variable ratio, such datasets
are not tractable by traditional multivariate statistical
methods, including factor and principal component
analyses (PCAs). In particular, in a single principal
component analysis, the number of cases should be at
least 5 to 10 times more than the number of variables
to derive a robust, generalizable solution. To deal with
this problem, in the present study, a two-stage proce-
dure was designed to implement the principal com-
ponent analysis in a stratified way.
In stage 1, the variables were divided into blocks
based on a priori knowledge about their observed
associations, derived from recent schizophrenia liter-
ature. To establish a clear delineation among adjacent
neuropsychological domains, these preselected blocks
of variables were intentionally chosen to be broad,
permitting the investigation of potential overlap
among hypothesized domains. In stage 2, the varia-
bles in each block were subjected to principal com-
ponent analysis. The outcome of this analysis
(variance explained, component loadings) helped us
to investigate the factors that explained most of the
variation within such a block of variables and to
identify the underlying neuropsychological domains.
A technique called ‘block principal component
analysis’ (BPCA) has been described recently in the
literature (Liu et al., 2002), which analogous to the
two-stage procedure employed in the current study,
relies on variable stratification. These authors pro-
posed the procedure ‘‘for extracting information from
a database with a large number of variables and
relatively small number of subjects’’. Using multivar-
iate statistical theory, they demonstrated that BPCA is
as efficient as ordinary principal component analysis
for dimensionality reduction. (For additional technical
details and numerical examples, please refer to Liu et
al. (2002).)
Table 2 provides details about the neuropsycholog-
ical test battery used (administration time: 4 –7 h),
Table 2
Neuropsychological tests used in the present study
Neuropsychological tests
Wechsler Adult Intelligence Scale-Revised (WAIS-R)
(Wechsler, 1981)
Wechsler Memory Scale-Revised (WMS-R) (Wechsler, 1987)
Letter Number Span (Gold et al., 1995)
Complex Ideational Material (Goodglass and Kaplan, 1983)a
Concentration Endurance Test (D2) (Brickenkamp, 1981)
Stroop Test (Uttl and Graf, 1997)
Wisconsin Card Sorting Test (WCST) (128 card manual version)
(Grant and Berg, 1995)
Trail Making Test (A and B) (Lezak, 1995)
Controlled Oral Word Association Test (COWAT)
(Benton and Hamsher, 1978)
Animal Naming Test (Goodglass and Kaplan, 1983)
Ruff Figural Fluency Test (Ruff et al., 1987)
Grooved Pegboard Test (Matthews and Love, 1964)
Finger Tapping Test (Reitan and Davidson, 1974)
Edinburgh Handedness Inventory (Oldfield, 1971)b
a
This measure was not used in subsequent analyses because it
revealed almost no variance among sample subjects.
b
This was used for determining preferred hand for motor tests.
which is comparable to batteries typically used in NP
studies of schizophrenia.
Table 3 identifies the NP measures from each of these
instruments used in the analyses, grouped according to
the hypothesized subdomains for schizophrenia. Most
NP tests generate many variables of potential interest.
Those measures most widely studied in the schizophre-
nia literature were selected for these analyses.
To validate the subdomain distinctions delineated
in Table 3, conceptually related (‘‘adjacent’’) subdo-
mains were combined and submitted to principal
component analysis with Promax rotation. By group-
ing potentially related subdomains, this procedure
offers a conservative method for determining the
validity of a priori subdomain delineations. In the
case of ‘‘Motor’’, the concern centered not on poten-
tial overlap of adjacent subdomains, but rather on the
validity of combining measures of motor dexterity and
fine motor speed into a single ‘‘Motor’’ domain, as
has been done in several studies (e.g. Bilder et al.,
2000; Heaton et al., 2001). Four PCAs were thus
computed consisting of the component variables that
comprised one or more of the a priori subdomains
listed in Table 3.
For Attention – Working Memory (ATTNWM),
variables making up the a priori subdomains of (1)
 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116 109

Table 3 Set Shifting/Cognitive Flexibility and (5) Ideational

NP test measures comprising each hypothesized subdomain for Fluency were combined and submitted to a second
schizophrenia at the first stage of analysis (all scores are raw scores)
PCA for Executive Functions (EXEC). Similarly for
(1) Sustained Vigilance (6) Verbal Learning
General Knowledge and Memory (KNOWMEM),
D2—Fluctuation WMS-R Logical
(worst – best row) Memory I variables making up the subdomains of (8) Verbal
D2—Total letters WMS-R Logical Knowledge, (9) Non-Verbal Reasoning, (6) Verbal
minus errors Memory II and (7) Non-Verbal Learning were combined, and
Stroop—words only WMS-R Verbal measures of motor speed and dexterity (MOTSPDX)
Stroop—color only Paired Associates I
were subjected to a fourth PCA.
Trails A WMS-R Verbal
(2) Short-Term Memory Paired Associates II Subdomains were retained if their internal consis-
Capacity/Span (7) Non-Verbal Learning tency (Cronbach’s alpha) was acceptable and all
WAIS-R Digit Span WMS-R Visual item – total correlations exceeded 0.50. If not, alterna-
Forward raw Reproduction I tive groupings of measures were tested until either it
WMS-R Visual Memory WMS-R Visual
was concluded that a subdomain was not valid or a
Span Forward Reproduction II
WMS-R Figural Memory WMS-R Visual measure could not be reliably included in any of the
WMS-R Logical Paired Associates I factors. The stability of each subdomain (replicability
Memory Immediate WMS-R Visual of the factor structure) was then examined by repeat-
(3) Working Memory Paired Associates II ing each PCA at the 6- and 18-month time point and
Letter – Number Span test (8) Verbal Knowledge
calculating the congruency coefficient (CC; range 0 –
(LNS), # correct WAIS-R Vocabulary
Letter – Number Span test, WAIS-R Information 1) between each time point.
longest span WAIS-R Comprehension The final factors/subdomains were further examined
WAIS-R Arithmetic WAIS-R Similarities for construct and divergent validity by studying the
WAIS-R Digit (9) Non-Verbal Reasoning/ relationship of each with clinical symptoms using the
Span Backward Problem Solving
five-factor solution of the PANSS of Bell et al. (1994).
WAIS-R Digit symbol WAIS-R Block Design
(4) Set Shifting/ WAIS-R Object Assembly
Cognitive Flexibility WAIS-R Pict. Completion
WCST # Perseverative WAIS-R Pict. Arrangement 3. Results
Errors (10) Motor
WCST # Categories Grooved Pegboard
3.1. Attention –Working Memory (ATTNWM)
WCST Failure to
Maintain Set Grooved Pegboard
Stroop—color/word Measures selected a priori for Sustained Vigilance,
interference score Finger Tapping—NH Short-Term Memory Capacity/Span, and Working
(Trails A Trails B)/Trails A Memory subdomains yielded two valid factors: Atten-
Trails B number errors Finger Tapping—PH
tion and Working Memory (Table 4A –4D). WMS-R
Ruff Figural Fluency—Errors
(5) Ideational Fluency Figural Memory did not load on either factor. A
Ruff Figural Fluency— ‘‘Short-Term Memory’’ factor distinct from either
# Unique Designs Attention or Working Memory could not be discerned.
COWAT Phonemic Fluency Standardized alpha coefficients (a) for the constit-
Animal Naming—
uent measures in the Attention and Working Memory
Semantic Fluency
factors were 0.60 and 0.74, respectively, confirming
D2 = Concentration Endurance Test, LNS = Letter Number Span
Test, WAIS-R = Wechsler Adult Intelligence Scale-Revised, WMS-
adequate construct validity. As shown in Table 4A,
R = Wechsler Memory Scale-Revised, WCST = Wisconsin Card the same two factors were revealed with very similar
Sorting Test, COWAT = Controlled Oral Word Association Test. constituent variables at each of three time points.
PH = Preferred hand, NH = Nonpreferred hand. Congruency coefficients (CCs) ranged from 0.86 to
0.96 over the three time points, indicating that the two-
Sustained Vigilance, (2) Short-Term Memory Capac- factor solution is highly replicable and stable over an
ity/Span, and (3) Working Memory were combined 18-month period (Table 5 lists individual CCs for all
and subjected to PCA. Variables included within (4) analyses).
110 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116

3.2. Executive Functioning (EXEC) Ideational Fluency items and WCST Perseverative
Errors (see Table 4B).
A PCA was computed combining measures grouped Remaining variables loaded in different combina-
a priori into Set Shifting, Cognitive Flexibility and tions of Factors 2 and 3 at each time point (and a
Ideational Fluency subdomains. Factor 1 comprised all fourth factor at the third time point). In contrast,
Table 4
Factor groupings for each of the a priori subdomains at each time point
(A) Attention/Working Memory
A priori subdomains Individual variables Time point each variable loaded on resulting factorsa
entered into PCA
Factor 1 Factor 2
T1 T2 T3 T1 T2 T3
ATTNWM PCA generated two factors Attention Working Memory
(1) Sustained D2—fluctuation – * * –
Vigilance D2—letters minus errors * * *
Stroop—words only * * *
Stroop—color only * * *
Trails A * * *
(2) Short-Term WAIS-R Digit Span Forward * * *
Memory WMS-R Visual Memory * * *
Capacity/Span Span Forward
[WMS-R—Figural Memory]b – – – – – –
(3) Working LNS, # correct * * *
Memory LNS, longest * * *
WAIS-R Arithmetic * * *
WAIS-R Digit Span Backward * * *
WAIS-R Digit symbol * * *
WMS-R Logical – * – *
Memory I

(B) Executive
A priori subdomains Individual variables Time point each variable loaded on resulting factorsa
entered into PCA
Factor 1
T1 T2 T3
EXEC PCA generated one factor Ideational Fluency + WCST Perseverative Errors
(4) Set Shifting/ WCST # Perseverative Errors * * *
Cognitive [WCST # Categories] – – –
Flexibility [WCST Failure to – – –
Maintain Set]
[Stroop—color/word – – –
interference]
[(Trails A Trail B)/Trail A] – – –
[Trails B # errors] – – –
[Ruff Figural Fluency— – – –
Errors]
(5) Ideational Ruff Figural Fluency— * * *
Fluency Unique Designs
COWAT * * *
Animal Naming * * *
 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116 111

Table 4 (continued)
(C) Knowledge/Memory
A priori subdomains Individual variables Time point each variable loaded on resulting factorsa
entered into PCA
Factor 1 Factor 2 Factor 3
T1 T2 T3 T1 T2 T3 T1 T2 T3
KNOWMEM PCA generated three factors Learning Verbal Non-verbal
knowledge functions
(6) Verbal Learning [WMS-R Logical – – * – – – –
Memory I]
[WMS-R Logical – – * – – – –
Memory II]
WMS-R Verbal Paired I * – * – –
WMS-R Verbal Paired II * – * – –
(7) Non-Verbal WMS-R Visual Reproduction I * * *
Learning WMS-R Visual Reproduction II * * *
WMS-R Visual Paired I * – * – –
WMS-R Visual Paired II * – * – –
(8) Verbal WAIS-R Vocabulary * * *
Knowledge WAIS-R Information * * *
WAIS-R Comprehension * * *
WAIS-R Similarities * * *
(9) Non-Verbal WAIS-R Block Design * * *
Reasoning/ WAIS-R Object Assembly * * *
Problem WAIS-R Pict. Completion * * *
Solving WAIS-R Pict. Arrangement * * *

(D) Motor
A priori subdomain Individual variables Time point each variable loaded on resulting factorsa
entered into PCA
Factor 1 Factor 2
T1 T2 T3 T1 T2 T3
MOTSPDX PCA generated two factors Motor dexterity Motor speed
(10) Motor Grooved Pegboard Pref * * *
Grooved Pegboard NonP * * *
Finger Tapping Pref * * *
Finger Tapping NonP * * *
‘‘*’’ Indicates a variable loaded on the factor at that time point, ‘‘ – ’’ indicates that variable did not load on any factor at that time point.
D2 = Concentration Endurance Test, LNS = Letter Number Span Test, WAIS-R = Wechsler Adult Intelligence Scale-Revised, WMS-
R = Wechsler Memory Scale-Revised, WCST = Wisconsin Card Sorting Test, COWAT = Controlled Oral Word Association Test; for Motor
tests, Pref and NonP = Preferred and Nonpreferred hand, respectively.
a
T1 = baseline; T2 = 6-month follow-up; T3 = 18-month follow-up.
b
Variables in italics failed to load on any factor.

Factor 1 was highly replicable, both in terms of the
constituent variables at each time point and as mea-
sured by the CCs (0.95, 0.96 and 0.96 for time point
pairs 1 – 2, 1 – 3 and 2 – 3, respectively) (Table 5).
However, standardized a equaled only 0.16 for Factor
1, indicating inadequate construct validity. Item –total
correlations (correlation of each variable to total factor
score computed with that item dropped) revealed a
value of 0.06 for WCST Perseverative Errors, while
ranging from 0.34 to 0.42 for the three ideational
112 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116

Table 5 3.4. Motor speed and dexterity (MOTSPDX)

Coefficient of congruence for each factor at each pair of time points
Neurocognitive Standardized Congruency coefficients at Motor speed and dexterity variables comprised the
factor alpha each pair of time points studied
final set. Two independent factors emerged from four
1–2 1–3 2–3 variables, and a virtually identical solution was ob-
Attention 0.60 0.93 0.96 0.96 served at each time point (Table 4D). CCs equaled
Working 0.74 0.90 0.92 0.86 0.99 for each pair of time points for each factor.
Memory
Standardized alphas could not be meaningfully
Learning 0.80 0.69 0.89 0.66
Verbal 0.88 0.95 0.95 0.96 computed because each factor contained only two
Knowledge variables. However, alpha was only 0.30 for the
Non-Verbal 0.82 0.93 0.96 0.88 original ‘‘Motor’’ factor, further supporting the con-
Functions clusion that a single ‘‘motor’’ factor could not be
Ideational 0.16a 0.95 0.96 0.96
reliably derived, and that motor dexterity and speed
Fluency +
WCST should be distinguished.
Perseverative
Errorsa
a
Standardized alpha was 0.69 after WCST Perseverative errors 4. Summary of new variable set
variable was removed. Thus, the retained factor excludes this
variable. Table 5 summarizes the six retained NP factors and
for each, the CC for each pair of time points and the
fluency variables. On this basis, WCST was eliminat- corresponding alpha. Note that the last factor listed
ed, resulting in a = 0.69 for the remaining variables. required removal of one variable to be retained as a
Thus, the subdomain Ideational Fluency was adopted reliable factor. Several NP measures would not be
with WCST Perseverative Errors excluded. realiaby combined with any of these factors or with
one another. Five of these have been widely studied in
3.3. General Knowledge and Memory (KNOWMEM) schizophrenia and should, in our view, be studied
separately. These include: WCST Perseverative
Verbal Knowledge, Non-Verbal Reasoning, Verbal Errors, Stroop Interference, Trails B Trails A/Trails
and Non-Verbal Learning variables were next submit- A, Grooved Pegboard Preferred + NonPreferred
ted to PCA, yielding three factors: Learning (a = 0.80), Hands, Finger Tapping Preferred + NonPreferred
Verbal Knowledge (a = 0.88) and Non-Verbal Reason- Hands.
ing/Problem Solving (a = 0.82). The latter two consist
of WAIS-R subtests and correspond closely with 4.1. Validation of NP factor set: relationship with
WAIS-R Verbal and Performance IQ, respectively. As PANSS ratings and functional disability
shown in Table 4C, the same three factors were
revealed with largely overlapping constituent variables The ongoing study from which the present dataset
at each time point. The exceptions were Visual Repro- is derived investigates relationships between NP def-
duction I and II which loaded on the Learning factor at icits, psychopathology and functional outcome for 3
time point 3, and Logical Memory I and II which only years following an acute exacerbation of illness.
loaded on one of the three retained factors at the third Preliminary analyses are reported to further examine
time point. Thus, the convention of distinguishing the construct, divergent and predictive validity of the
between ‘‘Verbal’’ and ‘‘Non-Verbal’’ secondary and NP factors.
immediate memory was not supported in this analysis. Consistent with the literature (e.g. (Addington et
All but two CCs exceeded 0.88 (Table 5). CCs for al., 1991; Berman et al., 1997; Harvey et al., 1996;
the Learning factor were somewhat lower for time O’Leary et al., 2000; Voruganti et al., 1997), relating
points 1– 2 (0.69) and for time points 2– 3 (0.66). negative symptoms to NP test performance, three of
However, overall, results indicate that these two the six NP factors were significantly correlated with
factors are replicable and stable. negative symptoms while none was correlated signif-
 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116 113

icantly with positive symptoms [employing the five-
factor PANSS solution of Bell et al. (1994) and a
significance threshold of P < 0.001 to correct for
multiplicity (30 tests)] (Table 6). All six NP factors
were correlated with the ‘‘cognitive’’ factor of the
PANSS, which is made up of items sensitive to many
of the deficits assessed by NP tests (e.g. goal-directed
sequencing of thought, alertness to external stimuli,
interpretation of proverbs and analogies, disorders of
thought involving fluidity, spontaneity, flexibility and
planning and motor abnormalities including unnatural
or awkward movements as well as tremors).
The predictive validity of the final NP variable set
was examined using multiple regression analyses with
baseline NP measures predicting the last available
MSIF global rating (the principle measure of func-
tional disability) covarying for negative and positive
symptoms. While positive symptoms were not related
to functional outcome, negative symptoms displayed a
significant association with the MSIF in all analyses
( P < 0.001) (i.e. higher negative symptom severity
was associated with poorer functional outcome). After
negative symptoms were accounted for, significant
associations remained for Attention ( F = 5.73, df =
1,142, P < 0.01), Working Memory ( F = 4.18, df =
1,142, P < 0.04), and Ideational Fluency ( F = 3.92,
df = 1,142, P < 0.05) factors. These analyses revealed
a pattern of relationships that is generally consistent
with the literature (e.g. Voruganti et al., 1997). The
fact that only some NP factors were predictive and
that the pattern of significant findings is consistent
with the literature thus far, suggests that the derived
factors are distinguishable from one another and can
be fruitfully employed to study relationships between
individual cognitive subdomains and a variety of
outcome variables (e.g. disability, medication treat-
ment response, etc.).
5. Discussion
An obvious alternative data reduction method
would be to conduct a single PCA on the entire
dataset. The problem with this approach (and the
reason it has not been widely adopted) is because
the first factor, corresponding roughly to ‘‘general
cognitive functioning’’ or ‘‘g’’, typically explains the
great majority of the variance. Such an approach does
not therefore permit the examination of group differ-
ences or treatment response with respect to distinct
cognitive operations. The approach of grouping var-
iables into indices corresponding to putatively dis-
cernable cognitive subdomains seems intuitively
optimal. The challenge to researchers is to discern
the most valid and replicable subdomains.
Several important findings emerged from the anal-
yses reported here. Our analysis of EXEC variables
did not reveal any reliable factor containing either
Trailmaking or Stroop, both widely studied in schizo-
phrenia. One possible explanation could be that these
measures were grouped with the wrong a priori
subdomain. While these measures seem to share the
requirement of response inhibition and set switching
(as does the WCST), they may require intact Working

Table 6
Correlations (Pearson r) between derived NP factor set and factor score ratings from the PANSSa
Neurocognitive factor N Positive Negative ‘‘Cognitive’’ Emotional Hostility
symptoms symptoms discomfort
(1) Attention 142 0.14 0.14 0.30*** 0.07 0.20*
(2) Working Memory 143 0.18 0.23** 0.38**** 0.12 0.22**
(3) Ideational Fluency 142 0.10 0.23* 0.39**** 0.02 0.21*
(4) Learning 141 0.08 0.21** 0.33**** 0.12 0.05
(5) Verbal Knowledge 143 0.11 0.26** 0.34**** 0.13 0.07
(6) Non-Verbal Functions 142 0.05 0.15 0.30**** 0.03 0.14
Two-tailed P values, indicating nominal level of significance.
a
Factors scores from five-factor solution of Bell et al. (1994).
* P < 0.05.
** P < 0.01.
*** P < 0.001.
**** P < 0.0001.
114 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116
Memory. However, an additional PCA conducted at
each time point, including the original EXEC and new
Working Memory factor variables, failed to yield any
new replicable factors. In fact, each time point yielded
an entirely different factor solution. For none of the
three time points did Trails or Stroop Interference load
on the same factor. We therefore conclude that, as
with WCST Perseverative Errors, these measures
must be treated as independent subdomains.
The widely held view that ‘executive deficits’
reflect impairment in a unitary cognitive operation
warrants critical review. Defined as impairment in
planning, organizing, sequencing and modulating
behaviors, and incorporating such conceptually dis-
sociable phenomena such as behavioral inertia, idea-
tional fluency, and working memory, it is surprising
that the concept of a unitary operation subsuming all
these functions is still favored in the absence of
supporting empirical data.
With respect to our findings regarding verbal
versus visuospatial memory, most reports in the liter-
ature that averaged standard scores within a priori
domains discriminated separate subdomains to reflect
this putative ‘modality specificity’ of memory perfor-
mance. (Bilder et al., 2000; Saykin et al., 1991, 1994;
Blanchard and Neale, 1994; Cuesta and Peralta, 1995;
Hoff et al., 1999; Censits et al., 1997). Our findings
failed to validate this distinction in schizophrenia, and
suggested instead that for this population, variation in
performance on visuospatial memory tasks is more
closely associated with other visuospatial problem-
solving tasks than other memory tasks. Another dis-
tinction applied in the NP literature (Heaton et al.,
2001; Bilder et al., 2000) not shown to be valid in our
analyses is that between immediate and delayed recall.
Our findings are consistent with Bilder et al. (2002)
who employed PCAs for NP data reduction and found
that WMS-R Visual Reproduction Immediate and
Delayed grouped together and in the same factor as
WAIS-R Block Design rather than into the memory
factor, which contained only Logical Memory and
Hopkins Verbal Learning Test-Immediate and -
Delayed.
Finally, our findings do not support the validity of
combining motor speed and dexterity measures into a
single motor factor (Bilder et al., 2000; Saykin et al.,
1991,1994; Blanchard and Neale, 1994; Censits et al.,
1997; Heaton et al., 2001).
Our investigation has several important limitations.
Principal component analysis unavoidably involves a
series of judgments on which experts will differ.
While our technique was designed to help us ‘follow
the data’, choices were made with respect to selection
of NP tests or individual measures, as well as with
respect to initial groupings of measures. While the
availability in our dataset of three time points offers
some further validation of judgments that were made,
different choices might nevertheless have yielded
different outcomes. Further, even employing the same
decisions, our findings may not generalize to a sample
studied either earlier or later in the course of illness, or
to a different diagnostic mix. The solutions offered
here should therefore be examined in other datasets.
Finally, in spite of the relatively large sample size,
given the number of potential NP variables, we were
still constrained by the case per variable ratio. Our
motivation in constructing the a priori factors was to
create a representative set of items within the limits
permitted by our case per variable ratio.
In conclusion, these findings may offer useful
guidance for future investigations of NP performance
in schizophrenia. Further, they raise a number of
important questions about findings to date. A sub-
stantial proportion of the large-scale NP studies in
schizophrenia employ the method of grouping varia-
bles into domains which, once constructed, are trea-
ted as valid and reliable measures of discrete
cognitive operations. Our findings did not validate
discrete ‘‘executive functioning’’ or ‘‘motor’’
domains. Also, several subdomain distinctions often
used (e.g. immediate memory/learning versus delayed
recall, visuospatial versus verbal memory) were not
validated.
Harvey and Keefe (2001) offer guidelines for
studying the cognitive enhancement properties of
new medications, and recommend selecting cognitive
domains with demonstrated correlations to functional
outcome as potential treatment targets. However,
validation of such target domains is essential—for
example, conclusions from the Green et al. (2000)
meta-analysis assume the existence of valid and
distinct subdomains for ‘‘Secondary Verbal Memory’’
and ‘‘Immediate Verbal Memory’’, not distinguished
in the present analysis. The guidelines also address
prior exposure effects resulting from repeated NP
testing. We would add to this caution the need to
 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116
validate the statistical stability of domain groupings if
calculated at more than one time point.
Future studies examining NP performance in
schizophrenia will need to consider the validity of
their data reduction methods in light of the findings
reported here.
Acknowledgements
The authors thank the patients who volunteered
their time for this study, as well as Cristina Gomes,
Anne-Marie Donovan-Lepore, Sherif Abdelmessih,
Stephen Panopolous, Susan Farella, and Drs. Rose-
marie Basile-Szulc and Rashmi Rastogi for their
contributions in assessing the patients. We are also
grateful to two anonymous reviewers for very helpful
comments on an earlier draft of this paper. The project
was supported by an NIMH Grant (R01 MH 55585)
entitled ‘‘Neuropsychology of Psychiatric Disability
and Service Needs’’ (J. Jaeger, Principal Investigator).
References
Addington, J., Addington, D., Maticka-Tyndale, E., 1991. Cogni-
tive functioning and positive and negative symptoms in schizo-
phrenia. Schizophrenia Research 5, 123 – 134.
Bell, M.D., Lysaker, P.H., Beam-Goulet, J.L., Milstein, R.M., Lin-
denmayer, J.P., 1994. Five-component model of schizophrenia:
assessing the factorial invariance of the positive and negative
syndrome scale. Psychiatry Research 52, 295 – 303.
Benton, A., Hamsher, K., 1978. Multiphasic Aphasia Examination
Manual. University Of Iowa, Iowa City.
Berman, I., Viegner, B., Merson, A., Allan, E., Pappas, D., Green,
A.I., 1997. Differential relationships between positive and neg-
ative symptoms and neuropsychological deficits in schizophre-
nia. Schizophrenia Research 25, 1 – 10.
Bilder, R.M., Goldman, R.S., Robinson, D., Reiter, G., Bell, L.,
Bates, J.A., Pappadopulos, E., Willson, D.F., Alvir, J.M.,
Woerner, M.G., Geisler, S., Kane, J.M., Lieberman, J.A.,
2000. Neuropsychology of first-episode schizophrenia: initial
characterization and clinical correlates. American Journal of
Psychiatry 157, 549 – 559.
Bilder, R.M., Goldman, R.S., Volavka, J., Czobor, P., Hoptman, M.,
Sheitman, B., Lindenmayer, J.P., Citrome, L., McEvoy, J.,
Kunz, M., Chakos, M., Cooper, T.B., Horowitz, T.L., Lieber-
man, J.A., 2002. Neurocognitive effects of clozapine, olanza-
pine, risperidone, and haloperidol in patients with chronic
schizophrenia or schizoaffective disorder. American Journal of
Psychiatry 159, 1018 – 1028.
Blanchard, J.J., Neale, J.M., 1994. The neuropsychological signa-
115
ture of schizophrenia: generalized or differential deficit? Amer-
ican Journal of Psychiatry 151, 40 – 48.
Brickenkamp, R., 1981. Concentration – Endurance Test Manual.
Verlag fur Psychologie, Gottingen.
Censits, D.M., Ragland, J.D., Gur, R.C., Gur, R.E., 1997. Neuro-
psychological evidence supporting a neurodevelopmental model
of schizophrenia: a longitudinal study. Schizophrenia Research
24, 289 – 298.
Cuesta, M.J., Peralta, V., 1995. Cognitive disorders in the positive,
negative, and disorganization syndromes of schizophrenia. Psy-
chiatry Research 58, 227 – 235.
Flor-Henry, P., 1969. Psychosis and temporal lobe epilepsy. A con-
trolled investigation. Epilepsia 10, 363 – 395.
Flor-Henry, P., 1976. Lateralized temporal – limbic dysfunction and
psychopathology. Annals of the New York Academy of Scien-
ces 280, 777 – 797.
Flor-Henry, P., Yeudall, L.T., 1979. Neuropsychological investiga-
tion of schizophrenia and manic-depressive psychoses. In: Gru-
zelier, J., Flor-Henry, P. (Eds.), Hemisphere Asymmetries of
Function in Psychopathology. Elsevier/North-Holland Biomed-
ical Press, Amsterdam, pp. 341 – 362.
Gold, J., Carpenter, C., Randolph, C., Goldberg, T., Weinberger, D.,
1995. Auditory Working Memory and the Wisconsin Card Sort-
ing Test. Schizophrenia Research 15, 117 – 118.
Gold, J.M., Carpenter, C., Randolph, C., Goldberg, T.E., Weinberg-
er, D.R., 1997. Auditory Working Memory and Wisconsin Card
Sorting Test performance in schizophrenia. Archives of General
Psychiatry 54, 159 – 165.
Gold, J.M., Queern, C., Iannone, V.N., Buchanan, R.W., 1999.
Repeatable battery for the assessment of neuropsychological
status as a screening test in schizophrenia: I. Sensitivity, reli-
ability, and validity. American Journal of Psychiatry 156,
1944 – 1950.
Goodglass, H., Kaplan, E., 1983. The Assessment of Aphasia and
Related Disorders. Lea & Febiger, Philadelphia.
Grant, D., Berg, E., 1995. A behavioral analysis of degree of re-
inforcement and ease of shifting to new responses in a Weigl-
type card-sorting problem. Journal of Experimental Psychology
38, 404 – 411.
Green, M.F., 1996. What are the functional consequences of neuro-
cognitive deficits in schizophrenia? American Journal of Psy-
chiatry 153, 321 – 330.
Green, M.F., Kern, R.S., Braff, D.L., Mintz, J., 2000. Neurocognitive
deficits and functional outcome in schizophrenia: are we measur-
ing the ‘‘right stuff’’? Schizophrenia Bulletin 26, 119 – 136.
Harvey, P.D., Keefe, R.S., 2001. Studies of cognitive change in
patients with schizophrenia following novel antipsychotic treat-
ment. American Journal of Psychiatry 158, 176 – 184.
Harvey, P.D., Lombardi, J., Leibman, M., White, L., Parrella, M.,
Powchik, P., Davidson, M., 1996. Cognitive impairment and
negative symptoms in geriatric chronic schizophrenic patients:
a follow-up study. Schizophrenia Research 22, 223 – 231.
Heaton, R.K., Baade, L.E., Johnson, K.L., 1978. Neuropsycholog-
ical test results associated with psychiatric disorders in adults.
Psychological Bulletin 85 (1), 141 – 162.
Heaton, R.K., Gladsjo, J.A., Palmer, B.W., Kuck, J., Marcotte, T.D.,
Jeste, D.V., 2001. Stability and course of neuropsychological
116 J. Jaeger et al. / Schizophrenia Research 65 (2003) 105–116
deficits in schizophrenia. Archives of General Psychiatry 58,
24 – 32.
Hoff, A.L., Sakuma, M., Wieneke, M., Horon, R., Kushner, M.,
DeLisi, L.E., 1999. Longitudinal neuropsychological follow-up
study of patients with first-episode schizophrenia. American
Journal of Psychiatry 156, 1336 – 1341.
Hoff, A.L., Kremen, W.S., Wieneke, M.H., Lauriello, J., Blankfeld,
H.M., Faustman, W.O., Csernansky, J.G., Nordahl, T.E., 2001.
Association of estrogen levels with neuropsychological per-
formance in women with schizophrenia. American Journal of
Psychiatry 158, 1134 – 1139.
Jaeger, J., Berns, S.M., Czobor, P., 2003. The multidimensional
scale for independent functioning: a new instrument for measur-
ing functional disability in psychiatric populations. Schizophre-
nia Bulletin (in press).
Kay, S.R., Fiszbein, S., Opler, L.A., 1987. The positive and neg-
ative syndrome scale (PANSS) for schizophrenia. Schizophrenia
Bulletin 13, 261 – 276.
Lezak, M., 1995. Neuropsychological Assessment. Oxford Univ.
Press, New York.
Liu, A., Zhang, Y., Gehan, E., Clarke, R., 2002. Block principal
component analysis with application to gene microarray data
classification. Statistics in Medicine 21, 3465 – 3474.
Matthews, C.G., Love, H., 1964. Instructions Manual for the Adult
Neuropsychology Test Battery. University of Wisconsin Medi-
cal School, Madison, WI.
Oldfield, R.C., 1971. The assessment and analysis of handedness:
the Edinburgh Inventory. Neuropsychologia 9, 97 – 113.
O’Leary, D.S., Flaum, M., Kesler, M.L., Flashman, L.A., Arndt, S.,
Andreasen, N.C., 2000. Cognitive correlates of the negative,
disorganized, and psychotic symptom dimensions of schizo-
phrenia. Journal of Neuropsychiatry and Clinical Neurosciences
12, 4 – 15.
Perlick, D., Mattis, S., Stastny, P., Teresi, J., 1992. Neuropsycho-
logical discriminators of long-term inpatient or outpatient status
in chronic schizophrenia. Journal of Neuropsychiatry and Clin-
ical Neuroscience 4, 428 – 434.
Reitan, R., Davidson, L., 1974. Clinical Neuropsychology: Current
Status and Applications. Hemisphere, New York.
Ruff, R.M., Light, R.H., Evans, R.W., 1987. The Ruff Figural
Fluency Test: a normative study with adults. Developmental
Neuropsychology 3, 37 – 51.
Saykin, A.J., Gur, R.C., Gur, R.E., Mozley, D., Mozley, L.,
Resnick, S.M., Kester, B., Stafiniak, P., 1991. Neuropsycho-
logical function in schizophrenia: selective impairment in
memory and learning. Archives of General Psychiatry 48,
618 – 624.
Saykin, A.J., Shtasel, D.L., Gur, R.E., Kester, D.B., Mozley, L.H.,
Stafiniak, P., Gur, R.C., 1994. Neuropsychological deficits in
neuroleptic naive patients with first-episode schizophrenia. Ar-
chives of General Psychiatry 51, 124 – 131.
Uttl, B., Graf, P., 1997. Color – Word Stroop Test performance
across the adult life span. Journal of Clinical and Experimental
Neuropsychology 19 (3), 405 – 420.
Voruganti, L.N., Heslegrave, R.J., Awad, A.G., 1997. Neurocogni-
tive correlates of positive and negative syndromes in schizo-
phrenia. Canadian Journal of Psychiatry 42, 1066 – 1071.
Wechsler, D., 1981. Wechsler Adult Intelligence Scale-Revised
Manual. Psychological Corporation, San Antonio.
Wechsler, D., 1987. Wechsler Memory Scale-Revised. Psycholog-
ical Corporation, New York.
Weinberger, D.R., Berman, K.F., Zec, R.F., 1986. Physiologic dys-
function of dorsolateral prefrontal cortex in schizophrenia: I.
Regional cerebral blood flow evidence [see comments]. Ar-
chives of General Psychiatry 43, 114 – 124.
White, L., Harvey, P.D., Opler, L., Lindenmayer, J.P., 1997. Empiri-
cal assessment of the factorial structure of clinical symptoms in
schizophrenia. A multisite, multimodel evaluation of the factorial
structure of the Positive and Negative Syndrome Scale. The
PANSS Study Group. Psychopathology 30, 263 – 274.