Schizophrenia Research 72 (2004) 29 – 39

www.elsevier.com/locate/schres

Identification of separable cognitive factors in schizophrenia

Keith H. Nuechterleina,b,*, Deanna M. Barchc, James M. Goldd, Terry E. Goldberge,
Michael F. Greena,f, Robert K. Heatong
a
Department of Psychiatry and Biobehavioral Sciences, Geffen School of Medicine at UCLA, United States
b
Department of Psychology, UCLA, United States
c
Department of Psychology, Washington University, St. Louis, United States
d
Department of Psychiatry, University of Maryland Baltimore School of Medicine, Maryland Psychiatric Research Center, United States
e
Clinical Brain Disorders Branch, NIMH Intramural Program, United States
f
VA Greater Los Angeles Healthcare System, United States
g
Department of Psychiatry, University of California, San Diego, United States
Received 3 September 2004; accepted 3 September 2004

Abstract

One of the primary goals in the NIMH initiative to encourage development of new interventions for cognitive deficits in
schizophrenia, Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS), has been to
develop a reliable and valid consensus cognitive battery for use in clinical trials. Absence of such a battery has hampered
standardized evaluation of new treatments and, in the case of pharmacological agents, has been an obstacle to FDA approval
of medications targeting cognitive deficits in schizophrenia. A fundamental step in developing such a battery was to identify
the major separable cognitive impairments in schizophrenia. As part of this effort, we evaluated the empirical evidence for
cognitive performance dimensions in schizophrenia, emphasizing factor analytic studies. We concluded that seven separable
cognitive factors were replicable across studies and represent fundamental dimensions of cognitive deficit in schizophrenia:
Speed of Processing, Attention/Vigilance, Working Memory, Verbal Learning and Memory, Visual Learning and Memory,
Reasoning and Problem Solving, and Verbal Comprehension. An eighth domain, Social Cognition, was added due to recent
increased interest in this area and other evidence of its relevance for clinical trials aiming to evaluate the impact of potential
cognitive enhancers on cognitive performance and functional outcome. Verbal Comprehension was not considered
appropriate for a cognitive battery intended to be sensitive to cognitive change, due to its resistance to change. The remaining
seven domains were recommended for inclusion in the MATRICS-NIMH consensus cognitive battery and will serve as the
basic structure for that

* Corresponding author. UCLA Neuropsychiatric Institute, 300 Medical Plaza, Rm 2251, Los Angeles, CA 90095-6968, United States.
Tel.: +1 310 825 0036; fax: +1 310 206 3651.
E-mail address: keithn@ucla.edu (K.H. Nuechterlein).

0920-9964/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.schres.2004.09.007
30 K.H. Nuechterlein et al. / Schizophrenia Research 72 (2004) 29–39

battery. These separable cognitive dimensions also have broader relevance to future research aimed at understanding the
nature and structure of core cognitive deficits in schizophrenia.
D 2004 Elsevier B.V. All rights reserved.

Keywords: Cognition; Schizophrenia; Factor analysis; Neuropsychology; Psychopharmacology

1. Introduction However, there has until now been no consensus

agreement on a standard way to divide the most
One of the primary initial goals of the NIMH prominent cognitive deficits into key dimensions or
contract on Measurement and Treatment Research to domains.
Improve Cognition in Schizophrenia (MATRICS) Extensive research using neuropsychological test
has been to develop a consensus cognitive battery for batteries has demonstrated that schizophrenia
use in clinical trials, as the absence of such a battery patients, as a whole, do show certain characteristic
has been a major impediment to standardized profiles of cognitive deficits (Heaton et al., 2001;
evaluation of new treatments designed to alleviate Heinrichs and Zakzanis, 1998; Saykin et al., 1991;
cognitive deficits in this disorder. In particular, Saykin et al., 1994). Some investigators have
development of a standard cognitive battery through suggested that deficits in a number of cognitive
a consensus of experts would establish an accepted domains may share a common cause (Andreasen et
measurement instrument for evaluation of al., 1998; Cohen et al., 1999; Goldman-Rakic,
pharmacological agents that target cognitive deficits 1994; Nuechterlein and Dawson, 1984). However,
in schizophrenia, thereby allowing a clear pathway individual patients may show meaningful variations
for FDA approval of such new medications. in the modal profile (Palmer et al., 1997, Weickert et
In considering the development of a MATRICS- al., 2000) and other researchers have suggested that
NIMH consensus cognitive battery for use in clinical different types of cognitive deficits may have
trials, a key step was to determine which cognitive different neurobiological substrates (e.g., working
domains should be represented. As reported by Kern memory versus episodic memory) and may respond
et al. (2004) elsewhere in this issue, the broad sample to different treatment approaches. Thus, it is
of experts in relevant fields that were surveyed by the important to represent the major separable cognitive
MATRICS Neurocognition Committee clearly stated domains in a battery for clinical trials. Unfortunately,
that they believed that reliable coverage of the major the ways that the cognitive domains have been
separable cognitive deficits in schizophrenia was an divided has varied from study to study, depending on
bessentialQ feature for the consensus cognitive battery. the preferences of the individual investigative teams
Ideally, separable in this context would refer to and the range of measures employed. To reach a
cognitive deficits with distinct causes or neural consensus on the major separable cognitive factors in
substrates that may therefore respond to different schizophrenia, the authors of this article served as a
types of pharmacological interventions. However, at subcommittee of the MATRICS Neurocognition
a minimum, separable should refer to cognitive Committee to review the available empirical
deficits in schizophrenia that are distinguishable at evidence and to make recommen- dations regarding
the statistical or analytic level, such as in the results the separable cognitive domains to be included in the
of factor analytic studies. There is an extensive MATRICS consensus cognitive battery for clinical
literature seeking to delineate the nature of cognitive trials.
deficits in schizophrenia (Braff, 1993; Goldberg and
Green, 2002; Nuechterlein and Dawson, 1984) and it
is clear that these deficits are enduring, core features 2. Method
of this illness (Censits et al., 1997; Gold, 2004;
Green and Nuechterlein, 1999; Nuechterlein et al., To delineate the major dimensions of cognitive
1992). deficit in schizophrenia, the authors established a set
 K.H. Nuechterlein et al. / Schizophrenia Research 72 (2004) 29–39 31

of principles for identification of the cognitive et al., 2002; Hobart et al., 1999; Keefe et al., 2004;
domains. The following principles were emphasized: Kremen et al., 1992; Mirsky, 1987; Nuechterlein et
al., 2003; Spaulding et al., 1989; Strauss and
(1) Dimensions that were independent or only Summerfelt, 2003). The analytic methods, character-
weakly intercorrelated were sought, such that istics of samples, and cognitive tests used in these 13
they could be viewed as separable contributors studies are listed in Table 1. As is evident there, the
to functional outcome and as potentially most popular method for extraction of dimensions (9
separate targets for new treatments. of 13 studies) was principal components analysis,
(2) Only cognitive dimensions that had been repli- usually accompanied by a Varimax rotation method
cated across several studies of schizophrenia that seeks independent factors (Jolliffe, 1986; Nun-
patients were included. nally, 1978). When principal components analysis was
(3) Research on separable neurocognitive factors in used, all studies except one used eigenvalue N1.0 as
large normal samples [e.g., research with the an initial guide to the number of factors to extract.
Wechsler Adult Intelligence Scale-III (WAIS-III) Confirmatory factor analysis (Bollen and Long,
and Wechsler Memory Scale-III (WMS-III) 1998; Hoyle, 1995) was employed in three studies,
standardization samples] was considered rele- typically using prior factor analytic solutions with
vant as an initial source of candidate normal subjects as a source of hypothesized factors
dimensions (Tulsky and Price, 2003). in schizophrenia. Sample sizes varied substantially,
(4) Many sources of information are relevant to from 34 to 209, and samples occasionally included
deciding whether or not specific cognitive patients with other psychoses or severe mental
domains should be considered separable, disorders rather than solely patients with
includ- ing the animal and human cognitive schizophrenia. As can be seen in Table 1, the subject-
neuro- science literatures on the neurobiological to-variable ratios also show a range across studies, so
substrates of different cognitive domains and the stability of loadings on factors may vary. While
differential responsivity to pharmacological most of the samples involved outpatients, a few
intervention. However, results of factor analytic involved inpa- tients or a combination of outpatients
studies of cognitive performance in schizophre- and inpatients. Because consistent factors could be
nia patients were viewed as a direct form of detected across these differences in factor analytic
evidence regarding cognitive dimensions in methods, sample composition, specific tests, and
schizophrenia, and one of the few sources of subject-to-variable ratio, the authors feel that greater
information available for all of the domains of confidence can be assigned to the conclusions.
interest for schizophrenia. Thus, all known Our MATRICS subcommittee examined the
factor analytic studies were sought, including results of factor analytic studies to seek replicated
both published and unpublished results to factors at the level of the cognitive test measures
broaden the available database as much as loading on factors. That is, the factor names assigned
possible. by the authors of the individual studies were
(5) When relevant, the likely sensitivity of a considered less important than the combination of
cognitive dimension to intervention attempts cognitive measures that each factor represented.
was considered, based on the neuropsycholog- Tables of factors and the measures loading on them
ical and cognitive neuroscience literatures. were constructed to summarize results across the 13
studies to examine consistency. In a few instances, it
Computer searches of the published scientific was not clear whether certain groups of cognitive
literature and individual inquiries of investigators measures should be combined in one factor or split
who might have relevant unpublished results yielded into two factors. When possible, additional targeted
13 factor analytic studies of cognitive performance in factor analyses were conducted to resolve these
schizophrenia (Allen et al., 1998; Dickinson et al., issues, using relevant original data sets available to
2004; Gladsjo et al., in press; Gold, unpublished; subcommittee members. Another method of
Goldberg, Egan, and Weinberger, unpublished; resolving such ambiguities was to examine other
Green studies of the interrelationships among
Table 1 32
Characteristics of factor analytic studies of cognitive performance in schizophrenia and related disorders
Study Analytic Sample Tests No of Percent of
method variables in variance
factor accounted
analyses for by factors
Allen et al. (1998) CFA 169 male schizophrenia WAIS-R subtests 11 NA
inpatients
Dickinson et al. Single 97 stable schizophrenia WAIS-III and WMS-III subtests 18 NA
common
(2004) FA outpatients
Gladsjo et al. CFA 209 outpatients with WAIS-R subtests, Trail Making Test, Boston Naming Test, 21 NA
K.
(in press) psychotic disorders Letter Fluency, Story Memory, CVLT, Figure Memory, H.
Grooved Pegboard, WCST, Booklet Category Test N
Gold et al. PCA, varimax 56 schizophrenia patients AFV fluency, FAS fluency, Symbol Search, Digit Symbol, 8 in each of 83% and 76% ue
two ch
(unpublished) Logical Memory I, Word List I, Woodcock–Johnson PCA analyses te
Analysis/Synthesis, WCST, Arithmetic, Digit Span, CPT-IP rl
Goldberg, Egan, PCA, varimax 86 schizophrenia patients Arithmetic, Similarities, Picture Completion, Digit Symbol, 16 64% ei
and n
Weinberger WRAT Reading, Trail Making Test, verbal fluency for et
(unpublished) categories, Gordon CPT, N-Back, Logical Memory I, Visual al.
Reproduction I, Verbal Paired Associates, WCST, CVLT /
Sc
Green et al. (2002) PCA, varimax 62 schizophrenia DS-CPT, Span of Apprehension, spatial work and reference 13 63%
hi
outpatients
memory tests, FAS fluency, CVLT, Digit Span Distractibility,
zo
ph
Pin Test, WCST, Block Design, Trail Making Test
re
Hobart et al. PCA, varimax 150 outpatients with WAIS-III Vocabulary, Picture Arrangement, and 22 74% ni
(1999) a
schizophrenia or Comprehension, WMS-III Logical Memory I and II, Visual R
major mood disorder Reproduction I and II, Woodcock–Johnson Listening es
Comprehension, Stroop, Trial Making Test, Gordon CPT, ea
WCST, Finger Tapping, Purdue Pegboard rc
Keefe et al. (2004) PCA, oblique 150 patients with BACS Digit Sequencing, Symbol Coding, Tower of London, 6 74% h
schizophrenia Token Motor Task, Verbal Fluency, Verbal Memory 72
Kremen et al. PCA 34 patients with major Trail Making Test, Digit Symbol, Digit Span, WRAT-R 11 77%
(1992)
psychotic disorders Arithmetic, WMS Mental Control, WCST, auditory CPT,
dichotic listening
Mirsky (1987) PCA 86 psychiatric patients and Digit Span, Arithmetic, Digit Symbol, Talland Letter 10 71%
normal subjects Cancellation Test, Stroop, Trail Making Test, WCST, X and
AX CPT
Nuechterlein et al. PCA, varimax 47 outpatients with a recent 3–7 CPT, DS-CPT, Span of Apprehension, backward masking, 10 65%
(2003) onset of schizophrenia CVLT, Digit Span Distractibility, Trail Making Test
Spaulding et al. CFA 125 patients with chronic COGLAB subtests for backward masking, simple reaction 10 29%
(1989) schizophrenia time, reaction time redundancy effect, rebound effect, and
anticipatory errors, a combined CPT and Span of Apprehension
Strauss and PCA 83 patients with task 6 35%
schizophrenia Trail Making Test, WCST, letter fluency, Calev word list
 Summerfelt or bipolar disorder learning, Moss visuospatial memory span, Mooney Closure Test
(2003)
Note: CFA=confirmatory factor analysis. PCA=principal component analysis. WAIS=Wechsler Adult Intelligence Test. WMS=Wechsler Memory Scale. CVLT=California Verbal Learning Test. AFV
fluency=fluency for names of animals, fruits, and vegetables. FAS fluency=fluency for words starting with F, A, and S. WCST=Wisconsin Card Sorting Test. CPT-IP=Continuous Performance Test,
Identical Pairs. WRAT=Wide Range Achievement Test. DS-CPT=Degraded Stimulus Continuous Performance Test. X and AX CPT=continuous performance test with X or AX targets. 3-7
CPT=continuous performance test with 3-7 target. COGLAB=computerized cognitive laboratory.
 K.H. Nuechterlein et al. / Schizophrenia Research 72 (2004) 29–39 33

the measures in schizophrenia, even if these studies Table 3

had not employed factor analytic methods. Attention/vigilance
Study Measures
Gold (unpublished) CPT-IP 2-, 3-, and 4-digit conditions
3. Results Goldberg et al. Gordon CPT, 1-back, WCST persev.
(unpublished) errors

The result of this process was the identification of Green et al. (2002) DS-CPT, Span, Spatial Memory
six separable cognitive dimensions in schizophrenia. Hobart et al. (1999) Gordon CPT vigilance and distractibility
Kremen et al. (1992) Auditory CPT, Dichotic Listening
We present these cognitive domains ordered from Mirsky (1987) X CPT, AX CPT
relatively basic to high-level cognitive processes, Nuechterlein et al. DS-CPT, 3–7 CPT, Backward Masking
(2003)
rather than in the order in which factors emerged
from individual studies, as the latter varied from
study
to study. To help the reader see the support for each highly on this factor emphasize the speed with which
dimension across studies and the nature of the digit/symbol pairings can be completed, target sym-
cognitive domain, a table showing the studies bols can be located, number or number/letter sequen-
identifying a similar factor and the primary measures ces on a page can be identified and connected, and
loading on the factor is provided for each dimension. colors can be named. Verbal fluency, often measured
It is important to note that the absence of a factor in by the number of words starting with a given letter
any particular study may reflect either the absence of that can be generated in a brief time period, also
measures in that cognitive domain or the lack of a loads on this factor, although it is not traditionally
factor that would be expected given the presence of thought of as a measure of processing speed. The
relevant measures. Thus, statements that a similar authors considered whether fluency should be a
factor was represented in a certain number of the 13 separate factor, but found that factor analytic results
studies should not be taken to imply that all 13 indicated that fluency most commonly loaded on this
studies would be expected to yield each factor. In factor. The cognitive processes tapped by the tasks
judging whether individual studies supported a given loading on this factor are relatively simple, often
separa- ble cognitive dimension, the Neurocognition involve perceptual and motor components, and
Com- mittee took into account the nature of the always empha- size speed of performance. This
measures included in those studies. dimension has also been identified in studies of
The first dimension, Speed of Processing, was normal cognitive per- formance (Tulsky and Price,
represented within 8 of the 13 factor analytic studies, 2003).
as shown in Table 2. Cognitive measures loading A second dimension, Attention/Vigilance, was
found in 7 of the 13 studies (Table 3). The most
prominent measures loading highly on this perform-
Table 2
ance dimension were indices from versions of the
Speed of processing
Continuous Performance Test (CPT), a sustained
Study Measures
attention task originally developed by Rosvold et al.
Dickinson et al. (2004) Digit Symbol, Symbol Search (1956) and refined for research on schizophrenia
Gladsjo et al. (in press) Digit Symbol, Trails A and B,
(Cornblatt et al., 1988; Nuechterlein, 1983). Other
Grooved Pegs
Gold (unpublished) Digit Symbol, Symbol Search measures loading on this dimension included shad-
owing one voice in a dichotic listening task and
target
Goldberg et al. Trails A and B, Digit Symbol, identification in early perceptual processing in the
(unpublished) Fluency forced-choice Span of Apprehension task (Asarnow et
Hobart et al. (1999) Stroop color and color-word, al., 1991) and under backward masking conditions
Trails A and B
Keefe et al. (2004) Fluency, Symbol Coding,
(Breitmeyer, 1984; Green et al., 1994). In studies of
Token Motor Test WAIS and WMS versions in normal subjects, a
Kremen et al. (1992) Trails A and B, Digit Symbol factor combining attention and working memory is
Mirsky (1987) Digit Symbol, Talland Letter often identified, including subtests, such as
Cancellation, Trails, Stroop Arithmetic, Digit Span, Letter–Number Sequencing,
and Spatial
Span (Tulsky and Price, 2003). However, when CPT Table 5
performance indices and these subtests are examined Verbal learning and memory
in the same schizophrenia samples, the CPT indices Study Measures
tend to load together on a separate factor (Gold, Dickinson et al. (2004) Logical Memory, Verbal
unpublished; Goldberg et al., unpublished; Kremen et Pairs Gladsjo et al. (in press) CVLT recall, Story Learning
Goldberg et al. CVLT recall, Logical Memory,
al., 1992; Mirsky et al., 1991; Nuechterlein et al., (unpublished) Verbal Pairs
2003), although some overlap with CPT versions with Green et al. (2002) CVLT, Fluency, Digit
immediate memory components is evident (Nuechter- Span Distractibility
lein et al., 2003). Thus, we concluded that an Keefe et al. (2004) Verbal Memory, Digit
attentional factor that emphasizes vigilance is separa- Sequencing Nuechterlein et al. (2003) CVLT recall, CVLT
recognition
ble from a working memory factor in schizophrenia.
A Working Memory dimension was represented in
8 of the 13 studies, as shown in Table 4. The most
common measures loading on this factor were the schizophrenia (Barch et al., 2002; Gold et al., 1997;
subtests Arithmetic, Digit Span, Letter–Number Kim et al., 2004; Park et al., 1995), but there is
Sequencing, and Mental Control from versions of debate as to whether deficits are more severe in one
the WAIS and WMS (Tulsky and Price, 2003; domain than another (Coleman et al., 2002; Tek et
Wechsler, 1981, 1987, 1995). However, other meas- al., 2002). However, available factor analytic results
ures from an experimental psychopathology tradition reveal that verbal and nonverbal tests tend to form a
in schizophrenia, such as the more complex levels of single factor when both are represented (Dickinson et
n-back tasks (Callicott et al., 2000; Cohen et al., al., 2004), and other available research has generally
1999) and spatial working memory tasks (Goldman- found moderate to high correlations of verbal and
Rakic, 1994; Keefe, 2000; Park et al., 1995), would nonverbal working memory deficits in schizophrenia
be expected to load on this dimension. All involve (Strauss and Summerfelt, 2003). Thus, the
temporary on-line storage of information and, in Committee con- cluded that one dimension of
most cases, mental manipulation of information. working memory in schizophrenia fits available
The MATRICS Neurocognition Committee con- evidence. Although verbal measures of working
sidered whether separate verbal and nonverbal work- memory have been used more extensively in
ing memory dimensions should be included among schizophrenia research, nonverbal measures of
the primary cognitive deficit dimensions in schizo- working memory have the advantage of readily
phrenia. Impairments in both verbal and nonverbal available animal models for drug develop- ment. As a
working memory domains have been documented in result, the Committee recommended that both verbal
and nonverbal subtests be included in the MATRICS-
NIMH consensus cognitive battery to measure the
Table 4 working memory dimension.
Working memory
A fourth dimension, Verbal Learning and
Study Measures Memory, was evident in 6 of the 13 factor analytic
Allen et al. (1998) Arithmetic, Digit Span studies of schizophrenia (Table 5). Immediate and
Dickinson et al. (2004) Letter Number Sequencing, delayed recall of word lists that exceed working
Spatial Span
Gladsjo et al. (in press) Arithmetic, Digit Span
memory capacity (e.g., California Verbal Learning
Gold (unpublished) Letter Number Sequencing, Test, Hopkins Verbal Learning Test) and immediate
Arithmetic, Digit Span and delayed recall of paragraph-length story
information (e.g., WMS-III Logical Memory I and II)
were the most typical measures loading on this
factor. Indices of paired
Goldberg et al. Digits Forward, Backward associate learning, recognition measures of secondary
(unpublished) or long-term memory, and recall of digit sequences
Kremen et al. (1992) Arithmetic, Digit Span,
Mental Control also loaded on this factor in some studies.
Mirsky (1987) Arithmetic, Digit Span The MATRICS Neurocognition Committee exam-
Nuechterlein et al. Digit Span, Trails B, 3–7 CPT ined the evidence concerning episodic memory
(2003)
deficits in schizophrenia to determine whether verbal
and nonverbal learning/memory should be Table 7
represented as a single dimension or as two Reasoning and problem solving
dimensions. Paral- leling the situation with Study Measures
working memory, impair- ments in verbal learning Dickinson et al. (2004) Matrix Reasoning, Block Design,
and memory (Gold et al., 1999; Heinrichs and Pict. Comp., Pict. Arrangement
Zakzanis, 1998; Saykin et al., 1991; Seidman et al., Gladsjo et al. (in press) Block Design, Category, WCST
1994) and in visual learning and memory (Heinrichs Green et al. (2002) WCST categories and persev.
errors Hobart et al. (1999) WCST categories and persev.
and Zakzanis, 1998; Nestor et al., 1993; Saykin et al.,
errors Keefe et al. (2004) Tower of London
1991, 1994) have been docu- mented in Kremen et al. (1992) WCST categories and errors
schizophrenia. However, factor analytic studies that Mirsky (1987) WCST errors
include verbal and visual tests of episodic memory in
schizophrenia tend to isolate separable verbal and
nonverbal factors (Dickinson et al., 2004; Gladsjo et as shown in Table 7. Seven of 13 studies identified
al., in press), paralleling their separation in large such a dimension, sometimes labeled Executive
normal samples (Tulsky and Price, 2003). Functioning rather than Reasoning and Problem
Furthermore, a supplementary analysis of a sample Solving. The label Reasoning and Problem Solving
of schizophrenia patients with episodic memory has the advantage of distinguishing this domain from
deficits (b 1 SD below normal on verbal and/or visual the executive processes of working memory (Badde-
memory; Nayak et al., 2004) showed that 47% of the ley, 1986), which would fall on the Working
patients had either verbal or visual episodic memory Memory factor in the current delineation of cognitive
deficits rather than both. Thus, the Committee domains. Cognitive measures loading highly on this
concluded that verbal learning and memory deficits factor involve sorting cards by an abstract principle
and visual learning and memory deficits should be that changes over time (Wisconsin Card Sorting
represented as separate dimensions. Test), nonverbal reasoning to complete a sequence of
The factor analytic evidence for a Visual Learning visual patterns (Matrix Reasoning) or to construct a
and Memory dimension in schizophrenia is summar- visual pattern (Block Design), moving round disks
ized in Table 6. Most factor analytic studies of between pegs in the smallest number of steps to
cognitive performance in schizophrenia have not achieve a specific order (Tower of London), and
included visual memory tests, so the fact that only 3 similar verbal and nonverbal problem-solving tasks.
of 13 studies identified such a factor is more a These high- level cognitive processes often demand
reflection of test selection rather than evidence that relatively intact lower-level processes, but also
such visual memory tests do not form a distinct involve addi- tional complex strategic planning and
factor. The tests falling on this dimension require decision-mak- ing skills.
recognition of faces either immediately or after a One additional cognitive dimension, Verbal Com-
delay, recall of family scenes either immediately or prehension, has also been isolated in factor analytic
after a delay, memory for nonfamiliar figures, and studies of schizophrenia, typically receiving high
reproduction of line drawings. loadings from measures such as the WAIS-III
The sixth major cognitive dimension in schizo- subtests for Vocabulary, Similarities, and
phrenia that has been isolated by replicated factor Information (Allen et al., 1998; Gladsjo et al., in
analytic evidence is Reasoning and Problem Solving, press; Hobart et al., 1999). However, because this
general verbal ability dimension is considered to be
Table 6 extremely resistant to change in schizophrenia and in
Visual learning and memory neuro- logically impaired samples (Lezak, 1995), it
Study Measures was omitted from the cognitive domains
Dickinson et al. (2004) Facial Recognition I and II, recommended for the MATRICS-NIMH consensus
Facial Pictures I and II cognitive bat- tery as the Committee did not feel that
Gladsjo et al. (in press) Figure Learning, Figure Delay it would be sufficiently sensitive to existing or future
Hobart et al. (1999) WMS-III Visual Reproduction
1 and 2
treatment effects in clinical trials.
4. Discussion press). Thus, we concluded that the evidence, across
all available studies, favored sufficient separation of
As detailed above, the review of available factor the selected cognitive dimensions to include them as
analytic evidence for separable cognitive dimensions different cognitive domains in a battery intended to
in schizophrenia yielded six domains that could be examine treatment effects in clinical trials. The
recommended for inclusion in the MATRICS-NIMH results of such factor analyses provide evidence
consensus cognitive battery for clinical trials: Speed about the number of discrete dimensions that are
of Processing, Attention/Vigilance, Working assessed by typical cognitive batteries. These factor
Memory, Verbal Learning and Memory, Visual analytic findings do not directly address the between-
Learning and Memory, and Reasoning and Problem group differences that may be observed across these
Solving. A seventh dimension, Verbal discrete dimensions. Indeed, another study using a
Comprehension, was iden- tified but not specialized statistical modeling procedure, single
recommended for inclusion due to the extreme common factor analysis, found that a common factor
resistance to change that characterizes such accounted for the majority of the variance
overlearned verbal skills. Results of two additional distinguishing schizo- phrenia patients from healthy
factor analytic studies of cognitive performance in subjects in six cognitive domains from the WAIS-III
schizophrenia that we were not aware of at the time and the WMS-III (Dick- inson et al., 2004). The
of our discussions are consistent with these presence of a substantial common factor in between-
conclusions (Friis et al., 2002; Kurtz et al., 2001). As group variance, however, should not be interpreted as
described in the introduction, many other sources of implying that this factor is due to a unitary, common
information are relevant to understanding whether underlying neural cause or a common etiological
specific cogni- tive domains are separable at a neural source. The common factor documented by
level and potentially responsive to different types of Dickinson et al. does, however, imply that a
pharmaco- logical interventions, and we do not claim substantial portion of between-group variability
to have bcarved cognition at its jointsQ based on the observed across relatively independent cognitive
results of these factor analytic studies. However, domains is shared. Further research will be needed
these studies provide an important starting point for to determine the extent to which variance in these
generating a consensus cognitive battery that begins statistically separable cognitive domains is due to
to capture the variations in cognitive function that common contributors and can be impacted by single
exist among individuals with schizophrenia. versus separate interventions.
A particular issue that is worthy of note is debate The results of this review of the empirical
over the extent to which a generalized cognitive evidence, as well as the results of the survey of a
deficit cutting across many cognitive functions group of 68 experts in relevant fields (Kern et al.,
accounts for impairment in schizophrenia. Two of 2004), were presented to the experts from academia,
the studies using exploratory factor analytic methods government, and industry who met at the initial
found that a single factor accounted well for the MATRICS conference in Potomac, MD, in April of
common variance in cognitive measures among 2003. As explained in detail in the Kern et al. article
chronic schizophrenia patients (Spaulding et al., in this issue, the survey revealed that the cognitive
1989) or a mixed sample of schizophrenia and domains that the experts chose most often for
bipolar patients (Strauss and Summerfelt, 2003). In inclusion in a consensus cognitive battery for schiz-
contrast, the clear majority of the exploratory factor ophrenia were executive, concept formation, and
analytic studies found that cognitive performance cognitive control processes, attention/vigilance,
variations in schizophrenia were best accounted for short-term memory, long-term memory, problem
by multiple independent or only weakly correlated solving and decision making, speed of processing,
factors, based on considerations of amount of and social cognition. Thus, while the grouping of the
variance accounted for, eigenvalues, and scree plots. cognitive domains was somewhat different in the
Furthermore, con- firmatory factor analysis in a study survey of experts than in the factors emerging from
with the largest sample of schizophrenia patients the review of the empirical evidence, all domains
indicated that a six- factor model produced the best identified in this survey, with the exception of social
fit (Gladsjo et al., in
cognition, were included in the dimensions that were tee decided that the next step in the construction of
derived from the review of factor analytic research the battery was to consider candidate tests within
on cognitive performance in schizophrenia. each of these seven domains, using the criteria for
A discussion followed at the April 2003 test selection that we have described elsewhere
MATRICS meeting to debate the pros and cons of (Green et al., 2004; Kern et al., 2004). This approach
including social cognition as a domain in the will allow the examination of a profile of cognitive
consensus cognitive battery. Because studies of deficits, using brief measures with strong
social cognition in schiz- ophrenia are relatively new, psychometric proper- ties for each domain. While our
the absence of a social cognition factor in factor categorization of the major separable cognitive
analytic studies of cognitive deficits in this disorder deficits in schizophrenia was completed for the
reflects the absence of such social cognition purpose of selecting domains for the MATRICS-
measures in these studies rather than evidence against NIMH consensus cognitive battery, we believe that
such a dimension. The recent rise in interest in social this identification of cognitive factors in
cognition impairment in schizophre- nia (Corcoran, schizophrenia has broad implications for further
2001; Kee et al., 1998, 2003; Lancaster et al., 2003; research on the nature of cognitive impairments in
Mueser et al., 1996; Penn et al., 1997) was viewed as this disorder.
an indication that social cognition should be
considered as an additional cognitive domain for the
consensus cognitive battery. It is, of course, not Acknowledgements
possible at this point to determine the extent to which
social cognition is a relatively unitary dimension that is An earlier version of this paper was presented at:
separable from the neurocognitive dimensions Measurement and Treatment Research to Improve
already summarized, as we know of no factor Cognition in Schizophrenia (MATRICS) Conference
analytic studies that have addressed this issue. #1: Identifying Cognitive Targets and Establishing
However, there are some indications that social Criteria for Test Selection, Bethesda, MD. Support
cognition deficits may serve as a mediator between for this conference came from NIMH Contract
neurocognitive deficits and functional outcome in MH22006 (S.R. Marder, P.I., M.F. Green, Co-P.I.).
schizophrenia (Brekke et al., 2003), suggesting that Additional support for the preparation of this article
improvements in social cognition might play an came from Center grant MH66286 (K.H.
important role in increasing everyday functioning in Nuechterlein, P.I.).
schizophrenia. Because the absence of social
cognition measures in the battery would mean that
the next generation of clinical trials for cognition-
References
enhancing treatments for schizophrenia would have
no standardized assessment for this potentially Allen, D.N., Huegel, S.G., Seaton, B.E., Goldstein, G., Gurklis Jr.,
important domain, the MATRICS Neuro- cognition J.E., van Kammen, D.P., 1998. Confirmatory factor analysis of
Committee concluded that the MATRICS- NIMH the WAIS-R in patients with schizophrenia. Schizophr. Res.
consensus battery should include social cogni- tion as 34, 87 – 94.
a seventh cognitive domain. Andreasen, N.C., Paradiso, S., O’Leary, D.S., 1998. Cognitive
dysmetria as an integrative theory of schizophrenia: a dysfunc-
In summary, the review of the empirical evidence tion in cortical–subcortical–cerebellar circuitry. Schizophr.
from factor analytic studies, the survey of a broad Bull. 24, 203 – 218.
sampling of experts, and discussion at the MATRICS Asarnow, R.F., Granholm, E.L., Sherman, T., 1991. Span of
conference led to the conclusion that seven major apprehension in schizophrenia. In: Steinhauer, S.R., Gruzelier,
cognitive deficits in schizophrenia should be repre- J.H., Zubin, J. (Eds.), Neuropsychology, Psychophysiology
and Information Processing, vol. 5. Elsevier Science
sented as separate domains in the MATRICS-NIMH Publishers, Amsterdam, pp. 335 – 370.
consensus cognitive battery for schizophrenia: Speed Baddeley, A., 1986. Working Memory. Clarendon Press/Oxford
of Processing, Attention/Vigilance, Working University Press, Oxford, England.
Memory, Verbal Learning and Memory, Visual Barch, D.M., Csernansky, J.G., Conturo, T., Snyder, A.Z., 2002.
Learning and Memory, Reasoning and Problem Working and long-term memory deficits in schizophrenia: is
there a common prefrontal mechanism? J. Abnorm.
Solving, and Social Cognition. The MATRICS
Psychology 111, 478 – 494.
Neurocognition Commit-
Bollen, K.A., Long, J.S., 1998. Testing Structural Equation Charney, D., Coyle, J.T., Nemeroff, C. (Eds.), Neuropsycho-
Models.
pharmacology: The Fifth Generation of Progress. Williams &
Sage Publications, Thousand Oaks, CA.
Wilkins, Philadelphia, Lippincott.
Braff, D., 1993. Information processing and attention dysfunctions
Goldman-Rakic, P.S., 1994. Working memory dysfunction in
in schizophrenia. Schizophr. Bull. 19, 233 – 259.
schizophrenia. J. Neuropsychiatry Clin. Neurosci. 6, 348 – 357.
Breitmeyer, B.G., 1984. Visual Masking: An Integrative Approach.
Green, M.F., Nuechterlein, K.H., 1999. Should schizophrenia
Oxford University Press, New York, NY.
be treated as a neurocognitive disorder? Schizophr. Bull. 25,
Brekke, J.S., Kay, D., Kee, K., Green, M.F., 2003. Biosocial
309 – 318.
pathways to functional outcome in schizophrenia: a path
Green, M.F., Nuechterlein, K.H., Mintz, J., 1994. Backward
analytic model. International Congress on Schizophrenia
masking in schizophrenia and mania: I. Specifying a mecha-
Research, Colorado Springs, CO.
nism. Arch. Gen. Psychiatry 51, 939 – 944.
Callicott, J.H., Bertolino, A., Mattay, V.S., Langheim, F.J., Duyn, J.,
Green, M.F., Marder, S.R., Glynn, S.M., McGurk, S.R., Wirshing,
Coppola, R., Goldberg, T.E., Weinberger, D.R., 2000. Physio-
W.C., Wirshing, D.A., Liberman, R.P., Mintz, J., 2002. The
logical dysfunction of the dorsolateral prefrontal cortex in
neurocognitive effects of low-dose haloperidol: a two-year
schizophrenia revisited. Cereb. Cortex 10, 1078 – 1092.
comparison with risperidone. Biol. Psychiatry 51, 972 – 978.
Censits, D.M., Ragland, J.D., Gur, R.C., Gur, R.E., 1997. Neuro-
Green, M.F., Nuechterlein, K.H., Gold, J.M., Barch, D.M., Cohen,
psychological evidence supporting a neurodevelopmental
J., Essock, S., Fenton, W.S., Frese, F., Goldberg, T.E., Heaton,
model of schizophrenia: a longitudinal study. Schizophr. Res.
R.K., Keefe, R.S., Kern, R.S., Kraemer, H.C., Stover, E.,
24, 289 – 298.
Weinberger, D.R., Zalcman, S., Marder, S.R., 2004. Approach-
Cohen, J.D., Barch, D.M., Carter, C., Servan-Schreiber, D., 1999.
ing a consensus cognitive battery for clinical trials in schizo-
Context-processing deficits in schizophrenia: converging evi-
phrenia: The NIMH-MATRICS conference to select cognitive
dence from three theoretically motivated cognitive tasks. J.
domains and test criteria. Biol. Psychiatry 56, 301 – 307.
Abnorm. Psychology 108, 120 – 133.
Heaton, R.K., Gladsjo, J.A., Palmer, B.W., Kuck, J., Marcotte,
Coleman, M.J., Cook, S., Matthysse, S., Barnard, J., Lo, Y., Levy,
T.D., Jeste, D.V., 2001. Stability and course of
D.L., Rubin, D.B., Holzman, P.S., 2002. Spatial and object
neuropsychological deficits in schizophrenia. Arch. Gen.
working memory impairments in schizophrenia patients: a
Psychiatry 58, 24 – 32.
Bayesian item-response theory analysis. J. Abnorm.
Heinrichs, R.W., Zakzanis, K.K., 1998. Neurocognitive deficits in
Psychology 111, 425 – 435.
schizophrenia: a quantitative review of the evidence. Neuro-
Corcoran, R., 2001. Theory of mind and schizophrenia. In:
psychology 12, 426 – 445.
Corrigan, P.W., Penn, D.L. (Eds.), Social Cognition and
Hobart, M.P., Goldberg, R., Bartko, J.J., Gold, J.M., 1999.
Schizophrenia, (1st ed.), American Psychological Association,
Repeatable battery for the assessment of neuropsychological
Washington, DC, pp. 149 – 174.
status as a screening test in schizophrenia: II. Convergent/
Cornblatt, B.A., Risch, N.J., Faris, G., Friedman, D., Erlenmeyer-
discriminant validity and diagnostic group comparisons. Am. J.
Kimling, L., 1988. The Continuous Performance Test,
Psychiatry 156, 1951 – 1957.
Identical Pairs Version (CPT-IP): I. New findings about
Hoyle, R., 1995. The structural equation modeling approach. In:
sustained attention in normal families. J. Psychiatr. Res. 26,
Holye, R. (Ed.), Structural Equations Modeling: Concepts,
223 – 238.
Issues, and Applications. Sage Publications, Thousand Oaks,
Dickinson, D., Iannone, V.N., Wilk, C.M., Gold, J.M., 2004.
pp. 1 – 19.
General and specific cognitive deficits in schizophrenia. Biol.
Jolliffe, I.T., 1986. Principal Component Analysis. Springer-
Psychiatry 55, 826 – 833.
Verlag, New York.
Friis, S., Sundet, K., Rund, B.R., Vaglum, P., McGlashan, T.H.,
Kee, K.S., Kern, R.S., Green, M.F., 1998. Perception of emotion
2002. Neurocognitive dimensions characterising patients with
and neurocognitive functioning in schizophrenia: what’s the
first-episode psychosis. Br. J. Psychiatry 181, s85 – s90.
link? Psychiatry Res. 81, 57 – 65.
Gladsjo, J.A., McAdams, L.A., Palmer, B.W., Jeste, D.V., Heaton,
Kee, K.S., Green, M.F., Mintz, J., Brekke, J.S., 2003. Is emotional
R.K., in press. A six-factor model of cognition in
processing a predictor of functional outcome in schizophrenia?
schizophrenia and related psychotic disorders: relationships
Schizophr. Bull. 23, 487 – 497.
with clinical symptoms and functional capacity. Schizophr.
Keefe, R.S.E., 2000. Working memory dysfunction and its
Bull.
relevance to schizophrenia. In: Sharma, T., Harvey, P. (Eds.),
Gold, J.M., 2004. Cognitive deficits as treatment targets in
Cognition in Schizophrenia: Impairments, Importance
schizophrenia. Schizophr. Res. 72, 21–28, this issue.
and Treatment Strategies. Oxford University Press, New York,
Gold, J.M., Carpenter, C., Randolph, C., Goldberg, T.E., Wein-
pp. 16 – 50.
berger, D.R., 1997. Auditory working memory and Wisconsin
Keefe, R.S.E., Goldberg, T.E., Harvey, P.D., Gold, J.M., Poe, M.,
Card Sorting Test performance in schizophrenia. Arch. Gen.
Coughenour, L., 2004. The brief assessment of cognition in
Psychiatry 54, 159 – 165.
schizophrenia: reliability, sensitivity, and comparison with a
Gold, S., Arndt, S., Nopoulos, P., O’Leary, D.S., Andreasen, N.C.,
standard neurocognitive battery. Schizophr. Res. 68, 283 – 297.
1999. Longitudinal study of cognitive function in first-
Kern, R., Green, M., Nuechterlein, K., Deng, B., 2004. NIMH-
episode and recent-onset schizophrenia. Am. J. Psychiatry
MATRICS survey on assessment of neurocognition in schizo-
156, 1342 – 1348.
phrenia. Schizophr. Res. 72, 11 – 19, this issue.
Goldberg, T., Green, M., 2002. Neurocognitive functioning in
patients with schizophrenia: an overview. In: Davis, K.L.,
Kim, J., Glahn, D., Nuechterlein, K., Cannon, T., 2004. Main- Palmer, B.W., Heaton, R.K., Paulsen, J.S., Kuck, J., Braff, D.,
tenance and manipulation of information in schizophrenia: Harris, M.J., Zisook, S., Jeste, D.V., 1997. Is it possible to be
further evidence for impairment in the central executive schizophrenic yet neuropsychologically normal?
component of working memory. Schizophr. Res. 68, 173 – 187. Neuropsychol- ogy 11, 437 – 446.
Kremen, W.S., Seidman, L.J., Faraone, S.V., Pepple, J.R., Tsuang, Park, S., Holzman, P.S., Goldman-Rakic, P.S., 1995. Spatial
M.T., 1992. Attention/information-processing factors in psy- working memory deficits in the relatives of schizophrenic
chotic disorders. Replication and extension of recent neuro- patients. Arch. Gen. Psychiatry 52, 821 – 828.
logical findings. J. of Nerv. Ment. Dis. 180, 89 – 93. Penn, D.L., Corrigan, P.W., Bentall, R.P., Racenstein, J.M., New-
Kurtz, M.M., Ragland, J.D., Bilker, W., Gur, R.C., Gur, R.E., man, L., 1997. Social cognition in schizophrenia. Psychol.
2001. Comparison of the continuous performance test with and Bull. 121, 114 – 132.
without working memory demands in healthy controls and Rosvold, H.E., Mirsky, A.F., Sarason, I., Bransome Jr., E.D.,
patients with schizophrenia. Schizophr. Res. 48, 307 – 316. Beck, L.H., 1956. A continuous performance test of brain
Lancaster, R.S., Evans, J.D., Bond, G.R., Lysaker, P.H., 2003. damage. J. Consult. Psychol. 20, 343 – 350.
Social cognition and neurocognitive deficits in schizophrenia. Saykin, A.J., Gur, R.C., Gur, R.E., Mozley, P.D., Mozley, L.H.,
J. of Nerv. Ment. Dis. 191, 295 – 299. Resnick, S.M., Kester, D.B., Stafiniak, P., 1991. Neuropsycho-
Lezak, M.D., 1995. Neuropsychological Assessment. 3rd ed. logical function in schizophrenia: selective impairment in
Oxford University Press, New York. memory and learning. Arch. Gen. Psychiatry 48, 618 – 624.
Mirsky, A.F., 1987. Behavioral and psychophysiological Saykin, A.J., Shtasel, D.L., Gur, R.E., Kester, D.B., Mozley, L.H.,
markers of disordered attention. Environ. Health Perspect. 74, Stafiniak, P., Gur, R.C., 1994. Neuropsychological deficits in
191 – 199. neuroleptic naive patients with first episode schizophrenia.
Mirsky, A.F., Anthony, B.J., Duncan, C.C., Ahearn, M.B., Kellam, Arch. Gen. Psychiatry 51, 124 – 131.
S.G., 1991. Analysis of the elements of attention: a neuro- Seidman, L.J., Yurgelun-Todd, D., Kremen, W.S., Woods, B.T.,
psychological approach. Neuropsychol. Rev. 2, 109 – 145. Goldstein, J.M., Faraone, S.V., Tsuang, M.T., 1994. Relation-
Mueser, K.T., Doonan, R., Penn, D.L., Blanchard, J.J., Bellack, ship of prefrontal and temporal lobe MRI measures to neuro-
A.S., Nishith, P., DeLeon, J., 1996. Emotion recognition and psychological performance in chronic schizophrenia. Biol.
social competence in chronic schizophrenia. J. Abnorm. Psychiatry 35, 235 – 246.
Psychology 105, 271 – 275. Spaulding, W., Garbin, C.P., Crinean, W.J., 1989. Die logischen
Nayak, G.V., Palmer, B.W., Jeste, D.V., Heaton, R.K., 2004. und psychometrischen Voraussetzungen fur eine kognitive
Sensitivity of visual and verbal memory tests to cognitive Therapie der Schizophrenie. In: Boker, W., Brenner, H.D.
impairment in schizophrenia. Annual Meeting of the Interna- (Eds.), Schizophrenie als systemische Storung. Verlag Hans
tional Neuropsychology Society, Baltimore, Maryland. Huber, Bern, Switzerland, pp. 157 – 169.
Nestor, P.G., Shenton, M.E., McCarley, R.W., Haimson, J., Smith, Strauss, M.E., Summerfelt, A., 2003. The neuropsychological
R.S., O’Donnell, B., Kimble, M., Kikinis, R., Jolesz, F.A., 1993. study of schizophrenia: a methodological perspective. In:
Neuropsychological correlates of MRI temporal lobe Lenzenweger, M.F., Hooley, J.M. (Eds.), Principles of Exper-
abnormal- ities in schizophrenia. Am. J. Psychiatry 150, 1849 imental Psychopathology: Essays in Honor of Brendan A.
– 1855. Maher. American Psychological Association, Washington, DC,
Nuechterlein, K.H., 1983. Signal detection in vigilance tasks and pp. 119 – 134.
behavioral attributes among offspring of schizophrenic Tek, C., Gold, J., Blaxton, T., Wilk, C., McMahon, R.P.,
mothers and among hyperactive children. J. Abnorm. Psychol- Buchanan, R.W., 2002. Visual perceptual and working
ogy 92, 4 – 28. memory impair- ments in schizophrenia. Arch. Gen. Psychiatry
Nuechterlein, K.H., Dawson, M.E., 1984. Information processing 59, 146 – 153.
and attentional functioning in the developmental course of Tulsky, D.S., Price, L.R., 2003. The joint WAIS-III and WMS-III
schizophrenia disorders. Schizophr. Bull. 10, 160 – 203. factor structure: development and cross-validation of a six-factor
Nuechterlein, K.H., Dawson, M.E., Gitlin, M., Ventura, J., model of cognitive functioning. Psychol. Assess. 15, 149 – 162.
Goldstein, M.J., Snyder, K.S., Yee, C.M., Mintz, J., 1992. Wechsler, D., 1981. WAIS-R Manual: Wechsler Adult Intelligence
Developmental processes in schizophrenic disorders: longitu- Scale—Revised. Psychological Corporation, San Antonio, TX.
dinal studies of vulnerability and stress. Schizophr. Bull. 18, Wechsler, D., 1987. Wechsler Memory Scale—Revised, Manual.
387 – 425. Psychological Corporation, New York.
Nuechterlein, K.H., Subotnik, K.L., Green, M.F., Yee-Bradbury, Wechsler, D., 1995. Manual for the Wechsler Adult Intelligence
C.M., Asarnow, R.F., Ventura, J., Kelly, K., Gitlin, M., Siegel, Scale III. Psychological Corporation, New York.
B., Mintz, J., 2003. Neurocognitive predictors of work Weickert, T.W., Goldberg, T.E., Gold, J.M., Bigelow, L.B., Egan,
outcome in the initial course of schizophrenia. Annual Meeting M.F., Weinberger, D.R., 2000. Cognitive impairments in
of the International Neuropsychology Society, Honolulu, patients with schizophrenia displaying preserved and compro-
Hawaii. mised intellect. Arch. Gen. Psychiatry 57, 907 – 913.
Nunnally, J., 1978. Psychometric Theory. McGraw Hill, New
York.