Psychological Medicine, 1994, 24, 829-847.

Copyright © 1994 Cambridge University Press

Cognitive function in depression: its relationship to

the presence and severity of intellectual decline
R. G. BROWN, 1 L. C. SCOTT, C. J. BENCH AND R. J. DOLAN
From the MRC Human Movement and Balance Unit, Institute of Neurology; Academic Department of
Psychiatry, Royal Free School of Medicine; MRC Cyclotron Unit, Hammersmith Hospital; and National
Hospital for Neurology and Neurosurgery, London

SYNOPSIS Cognitive dysfunction is an integral feature of depression, in some cases of sufficient

severity to warrant a diagnosis of dementia. There has been little systematic investigation of whether
cognitive dysfunction is an inevitable consequence of depression, or is specific to a subgroup of
depressed patients. Related to this is the distribution of cognitive dysfunction, whether there is a
continuum of impairment or a distinct demented subgroup. Finally, there is the question of which
aspects of cognitive function are most sensitive to the intellectual decline seen in depression. A study
is described which addresses these issues. The distribution of global cognition was found to be
normally distributed in the sample of 29 patients assessed. Based on this distribution and the scores
of a control sample, the patients were classified as unimpaired, borderline or impaired. Two sets of
independent comparisons were carried out. First, the unimpaired depressed patients were compared
to matched non-depressed controls. Significant deficits were found on a range of neuropsychological
measures covering aspects of language function, memory, both recall and recognition, attention and
behavioural regulation. These same patients were also compared with two groups of matched
depressed patients, with varying degrees of global cognitive impairment. In general, the cognitive
measures showed a gradient of dysfunction across the three patient groups. Significant differences
between the depressed groups were shown on measures of immediate recall, attention and
behavioural regulation. The possible significance of attentional factors for the observed memory
dysfunction is discussed.

iMTDnni i r n r i M Neuropsychological research has addressed a

number of issues relating to cognitive dys-
Depressed individuals frequently complain of function in depression. Some has been concerned
poor memory and concentration. Much effort primarily with clinical questions: whether there
has gone into assessing the nature and severity are differences between clinical subgroups (e.g.
of this impairment. Although focused almost Savard et al. 1980); the effects of treatment
exclusively on memory, a few studies have shown (Sternberg & Jarvik, 1976; Frith et al. 1983) and
deficits in areas such as abstract reasoning (Braff the changes with remission of the depression
& Beck, 1974), simple perceptual discrimination (Savard et al. 1980). Other research has
(Cornell et al. 1984) and verbal fluency addressed the nature of the processes underlying
(Robertson & Taylor, 1985). A simple explan- the deficits. This research has employed various
ation of the observed deficits is that they reflect explanatory frameworks, such as information
poor motivation, or distraction from depressive processing (Hasher & Zaks, 1979), 'effort'
thoughts (see Jorm, 1986). However, there is (Cohen et al. 1982), encoding strategy (Miller &
increasing consensus that cognitive dysfunction Lewis, 1977), processing resources (Watts et al.
is intrinsic to depression and directly related to 1990) or arousal and activation (Weingartner et
the neurobiology of the illness. al. 1981). To date, however, no single frame-
1
work, 'theory' or explanation, has proven to be
Address for correspondence: Dr R. G. Brown, MRC Human Oc
I g ee nn eerradll vv ad ll u e lj nn understanding the nature of
Movement and Balance Unit, Institute of Neurology, Queen Square, 8 Unuerbldnuing me ndiure Ol
London WCIN 3BG. cognitive impairment in depression. The reader
829
830 R. G. Brown and others

is referred to a number of reviews (Miller, 1975; deficits in this group could confidently be
McAllister, 1981; Willner, 1984; Jorm, 1986; attributed to depression, and not the influence
Widlocher & Hardy-Bayle, 1989; Newman & of a subgroup of patients with a possibly
Sweet, 1992). independent dementing disorder. Conversely, if
The literature on the neuropsychology of such patients/a/7 to show any deficits, this would
depression has remained separate from research call into question models which suggest that
concerned with the relationship between de- cognitive dysfunction is an epiphenomenon of
pression and dementia (Cummings & Benson, the depressive symptoms.
1984). Some patients with depression suffer In the present paper, we examine the neuro-
cognitive impairment of sufficient severity to psychological profiles of a sample of depressed
warrant a diagnosis of dementia. The apparent patients classified on the basis of overall in-
reversibility of the dementia with remission of tellectual function. Because of the shortage of
the depressive illness has led to the label data relating to this issue we chose an ex-
'pseudodementia' (Kiloh, 1961) to distinguish it ploratory approach employing a broad assess-
from progressive dementias such as Alzheimer's ment of cognitive function. Our aim was to
disease and Pick's disease. More recently the address three main questions. First, what is the
reversibility, or otherwise, of cognitive impair- distribution of cognitive dysfunction in a sample
ment has become less important in the diagnosis of depressed patients? Is there a continuous
of dementia (Jorm, 1986). As a result the terms distribution, or is there evidence for a distinct
'depressive dementia' or 'dementia syndrome of subgroup of demented patients? Secondly, is
depression' (Cummings, 1989) have become the there evidence of specific cognitive dysfunction
accepted diagnostic labels. even in a sample of depressed patients selected
There are thus two broad strands of research for the absence of global cognitive impairment?
concerned with, first, the nature of the im- Thirdly, within a depressed sample, which
pairment in depressed patients per se, and specific aspects of cognitive function are most
secondly, the nature of depressive dementia, and susceptible to impairment?
its relationship to other dementing disorders
and its neurobiological substrate. To date, little METHOD
attempt has been made to combine these two
issues. However, some important questions arise Subjects
when considering depression and cognitive func- Patients were recruited from district psychiatric
tion in a broader context. Are we dealing with a services in north London and the National
single entity of cognitive impairment in de- Hospital for Neurology and Neurosurgery.
pression, with dementia being the extreme case? Potential patients were administered the Sched-
Is cognitive deficit, whether mild or severe, an ule for Affective Disorders and Schizophrenia
inevitable consequence (or concomitant) of (SADS) (Endicott & Spitzer, 1978). Subjects
depression? who satisfied Research Diagnostic Criteria
Assessing such questions on the basis of (Spitzer et al. 1977) (RDC) for Major Depressive
existing studies is difficult. In the majority of Disorder were further screened with the fol-
studies, patients have been selected on the basis lowing exclusion criteria: past or present history
of their affective disturbance alone. In such a of neurological disease, drug or alcohol abuse,
sample it is possible that only a proportion will any significant past medical illness, a score of
exhibit cognitive deficits. Other patients, with more than 4 on the Hachinski ischemia scale
equally severe depression, may have no obvious (Hachinski et al. 1975). In total, 29 patients
cognitive impairment. The effect of combining entered into the study. These had moderate to
samples of unimpaired depressed with demented severe depression as rated on the 17-item
depressed patients would simply be to show an Hamilton Rating Scale for Depression
average effect which fails to reflect the true (Hamilton, 1960) (mean = 250, S.D. = 4-2,
nature or severity of cognitive disturbance in range = 17-34). The mean age of the sample was
either of the subgroups. One approach is to 580 years (S.D. = 130). The sample comprised
assess a selected group of patients who show 19 males and 10 females. Their mean number of
normal global cognitive function. Specific years of education was 11-3 years (S.D. = 3-0). Of
 Cognitive function in depression 831

the 29 patients, 14 were taking antidepressant Word-Pair Associate Learning Test (PALT)
medication at the time of assessment. Most of (Wechsler, 1945); Rey Auditory Verbal Learning
the patients (26) were classified as having Test (RAVLT) (Taylor, 1959). Recall was
unipolar depression, with only three having assessed immediately and after delays of 1 and
bipolar depression. 24 h. The effects of list organization was assessed
Twenty healthy control subjects were also using three, 12-word lists comprising (A) un-
assessed. They comprised 6 males and 14 females, related words, (B) words from 3 categories
with a mean age of 58-4 years (S.D. = 14-5), and randomly presented and (C) words from 3
mean of 12-4 years of education (S.D. = 1-8). The categories clustered by category. Immediate free
same exclusion criteria were applied as with the recall was assessed, followed, for lists B and C,
patients. All subjects gave informed consent. by recall in which the subject was cued with the
category names. After cued recall, subjects were
Psychiatric assessment given a recognition test in which the 12 words in
Previously (Bench et al. 1993), the SADS results each list were randomly mixed with 12 new
of a larger sample, including the present subjects, items. For the lists B and C, the distractors
were subjected to a principal component analysis belonged to the same semantic categories as the
with varimax rotation. The results revealed original words. Subjects were shown the words
one factor 'Anxiety/Somatism', and another one at a time and asked whether or not they
labelled 'Mood/Retardation'. For each indi- recognized them. Short-term memory was
vidual in the present study two factor scores assessed using the Brown-Peterson (BP) test
were derived and used in subsequent correla- (Brown, 1958; Peterson & Peterson, 1959). A
tional analyses. three-consonant trigram was read out to the
subject which they had to repeat immediately or
Neuropsychological assessment after a delay of 5, 10, 20 or 30 s. Rehearsal was
The neuropsychological assessment comprised prevented by asking the subjects to count
two sections. The first was a basic screening of backwards from a 3-digit number, presented
cognitive function employing the CAMCOG immediately after the trigram. A number of tests
(Roth et al. 1988). This measure comprises an related to language and 'executive' function
expanded version of the Mini-Mental State were also administered. Verbal fluency was
Examination (MMSE) (Folstein et al. 1975), assessed using 3 conditions: 'free' (any word),
and provides a total score (maximum 107) as 'category' (boys' names) and 'letter' (any word
well as subscale scores for different aspects of beginning with the letter' s'). Subjects performed
cognition (see Table 1). A MMSE score can also each task for 60 s. Separate scores were given for
be calculated (maximum score 30). A cut-off of the number of words generated in the first and
80 on the CAMCOG was found by Roth et al. second 30 s of each test. Language compre-
(1986) to be useful to indicate the presence of hension was assessed using an abbreviated
significant cognitive impairment. The second version of the Token Test (Spreen & Benton,
part of the investigation comprised a battery of 1969). Subjects were given 16 commands to
neuropsychological tests, administered over two carry out. A ' strict' score was derived from the
sessions 24 h apart. Each subject was assessed at subjects performance on a single reading of the
the same time of day on the two sessions. The command. If the subject made an error, the
order of test presentation was standardized as command was given again, and a separate,
far as possible. Subjects were administered 5 'lenient' score derived. Conceptual ability was
subtests of the Wechsler Adult Intelligence assessed using the Weigl Test (Weigl, 1941).
Scale (WAIS-R) (Wechsler, 1986): Vocabulary,
Similarities, Comprehension, Arithmetic and Statistics and organization of results
Digit Span, from which a pro-rated Verbal IQ The main methods were multivariate analysis of
was calculated. Reading ability was assessed variance (MANOVA) and repeated measures
using the Schonell Graded Word Reading Test analysis of variance (ANOVA). Planned
(Schonell, 1942; Nelson & McKenna, 1975). contrasts were employed where specific effects
Verbal memory and learning were assessed were to be tested on an a priori basis. Where
using: Wechsler Logical Memory (LM) and
significant omnibus effects and interactions were
832 R. G. Brown and others

Table 1. Details (mean and standard deviation) of the control and three depressed groups, for age, years
of education, Hamilton score, Schonnel score, and results from the CAMCOG and WAIS (with scaled
scores). The bracketedfiguresin italics following the measure name refer to the maximum score on the
test or subtest (see also subsequent tables)
Depressed groups
fYintrnl
group Unimpaired Borderline Impaired
(N= 16) (N= 10) (Af= 10) (yV = 9)
Measure Mean (s.D.) Mean (s.D.) Mean (s.D.) Mean (s.D.)

Age 62-6(141) 61-9(13-4) 53-5(31-5) 58-7(11-1)

Years of education 12-4(1-8) 11-9(3-4) 11-2(3-7) 10-7(1-9)
Hamilton Depression Scale score — 24-4 (4-2) 25-6 (5-8) 25 1 (3-8)
Schonnel Reading Test score (100) 94-5 (6-5) 92 1 (13-8) 80-5(18-4) 63-2 (25-2)
CAMCOG total score (107) 97-7 (2-3) 96-1 (3-8) 86-7 (2-5) 711 (10-7)
Orientation (10) 9-9 (0-3) 9-6(1-9) 9-3 (0-5) 6-3 (2-2)
Language(30) 28-6(1-1) 26-3(1-8) 25-0(1-4) 22-9 (2-3)
Memory (27) 22-4(1-9) 22-8(1-8) 19-9(2-8) 15-6(4-8)
Attention (7) 6-8 (0-5) 6-6(1-0) 51 (1-4) 360 (2-5)
Praxis (12) 11-2(0-8) 11-4(0-8) 10-2(1-2) 8-4 (2-2)
Calculation (2) 20 (00) 20 (00) 20 (00) 1-4(0-5)
Abstraction (8) 7-7 (0-9) 7-3 (0-8) 6-4(1-4) 5-3 (2-3)
Perception (//) 10-3(0-8) 10-2 (0-8) 8-9(1-6) 7-8(1-8)
MMSE total score (30) 29-2(1-0) 28-9(1-2) 26-2 (2-6) 19-7 (4-2)
WAIS Verbal IQ 113-6(12-5) 110-3(19-3) 90-1 (7-2) 86-7(16-2)
Vocabulary 12-3 (30) 12-0(4-1) 7-9 (1-7) 6-7 (2-6)
Comprehension 12-5(2-5) 11-4(2-9) 91 (21) 7-7 (2-6)
Similarities 10-9(1-9) 9-7 (3-3) 6-6(1-4) 6-3(1-5)
Digit Span 111 (3-4) 8-9 (3-8) 8-5 (3-2) 6-3(1-5)
Arithmetic 11-2(2-1) 10-6(3-1) 6-9(1-7) 5-6(2-4)

obtained, further comparisons were carried out. Of the patients, 10 (34-5%) scored 92 or more,
Conventional significance levels < 0-05, < 0-01 with an upper score of 101. These depressed
and < 0-001 are adopted. P values of > 0-10 are subjects will be considered 'unimpaired' de-
considered 'not significant'. Those between 005 pressed (UD). The remaining 19, divide into two
and 0-10 (P < 0-10) are reported as' approaching groups: 10 (34-5%) 'borderline impaired' (BD)
significance'. scored less than 92 but more than 81, while the
The first part of the results section is a remaining 9 (31 %) scored 81 or less and formed
classification of the depressed sample into three the 'impaired' group (ID). Of the 20 control
subgroups based on their total CAMCOG scores subjects, 4 achieved higher CAMCOG scores
in relation to the control sample. Analysis of the than any depressed patient. To improve
four groups (one control and three depressed) matching, these 4 subjects were excluded from
were then carried out in two independent sets of subsequent analysis, leaving afinalcontrol group
comparisons. First, between a group of patients of 16 subjects.
defined a priori a s ' unimpaired' and the control
Comparison of unimpaired depressed and
group. Second, between the three subgroups of
control groups
depressed patients, classified according to the
total CAMCOG scores. Age, sex, years of education, reading ability
and CAMCOG score {Table 1)
Age, years of education, Schonell reading test
RESULTS
score and total CAMCOG score, were entered
Total CAMCOG scores - classification of into a MANOVA. The two groups did not differ
patients on the set of variables. Univariate comparisons
Fig. 1 shows the distributions of total CAMCOG confirmed that the two groups were matched on
scores. The scores of the 20 control subjects each measure. There was a significant difference
ranged from 92-104 out of a maximum of 107. in the sex ratios of the two groups (P < 0-05).
 Cognitive function in depression 833

3 r

45 50 55 60 65 70 75 80 85 90 95 100 105
CAMCOG Score
FIG. 1. Distribution of total CAMCOG scores in the controls group (N = 20) ( Q ) and depressed group (N = 29) ( I ).

However, comparison of the male and female subscales (P > 010). A MANOVA with all of
subjects in the two groups failed to reveal the subscale scores, revealed no significant
any consistent difference in performance on the overall difference between the control and UD
various neuropsychological tests between the groups. Univariate F tests revealed that the two
two sexes. Consequently, subject's sex was not groups were matched on all subscales.
considered further.
Memory
CAMCOG subscales (Table 1) Immediate free recall (Table 2)
The MANOVA of subscale scores failed to The UD group showed a mixed pattern of
reveal any significant overall group difference performance on the measures of immediate free
(the Arithmetic subscale score was excluded recall. For the prose passage of the LM test,
from this analysis as all subjects scored at their performance was lower than controls but
ceiling). Despite the non-significant omnibus-F, the difference was not significant. Similarly, the
univariate F tests demonstrated significant group difference for recall of the first pres-
differences between the two groups on the two entation of the 15 word list from the RAVLT
language tests, Comprehension (P < 0-05) and only approached significance (P < 010). Signifi-
Expression (P < 0001), with the UD group cant differences, however, were found for recall
performing less well than the controls. In of the three word lists (P < 001). Across subjects
addition to the subscale analysis, one test item, recall of the unrelated list (A) was inferior to the
serial sevens, was analysed individually because two categorized lists (B and C). Of these, recall
of its utility as a clinical measure of attention. of the unclustered list (B) was superior to that
There was no difference between the perform- for the clustered list (C). This pattern of recall by
ance of the two groups (UD group, mean = 4-8, list type was the same in the two groups. All of
S.D. = 0-4; control group, mean = 4-9, s.D. = the above measures involved supraspan material.
0-3). Mean memory span as measured by the digit
span subtest of the WAIS-R was 6-6 (S.D. = 21)
WAIS- Verbal IQ and Verbal Subscales in the UD group and 71 (S.D. = 1-6) in the
{Table 1) controls. The difference was not significant.
There was no significant difference between the Overall, therefore, immediate recall of supra-
groups on Verbal IQ pro-rated from the 5 span verbal material appeared to be inferior in
834 R. G. Brown and others

Table 2. Results (mean and standard deviation) for control and the three depressed groups on
Logical Memory, Paired Associate Learning and Rey Auditory Verbal Learning Test
Depressed groups
font ml
group Unimpaired Borderline Impaired
(W=16) (N= 10) (N = 10) (N = 9)
Measure Mean (s.D.) Mean (s.D.) Mean (s.D.) Mean (s.D.)

Wechsler Logical Memory

Immediate recall score (46) 24-4(5-1) 20-9 (8-8) 14-6(4-9) 10-4(5-6)
1 h delayed recall score 20-6 (5-6) 14-4(9-6) 8-1(5-3) 5-4 (6-4)
1 h delayed recall (% immediate) 83 9 ( 1 1 2 ) 63-6(28-1) 41-8(32-0) 40-7 (39-8)
24 h delayed recall score 19-4(6-2) 11-8(9-9) 71 (5-2) 5-8 (6-5)
24 h delayed recall (% immediate) 82-4(14-1) 50-5 (32-7) 44-1 (34-9) 47-2 (46-9)

Paired Associate Learning Task

Easy items Trial 1 (4) 3-4 (0-7) 3-3 (0-8) 3-4 (0-7) 2-7(1-2)
Trial 2 3-8 (0-6) 3-8 (0-6) 3-4 (0-7) 2-8(1-3)
Trial 3 40 (00) 3-9 (0-3) 3-6 (0-5) 2-8(1-5)
Trial 4 40 (00) 3-9 (0-3) 3-8 (0-4) 2-7(1-6)
Trial 5 40 (00) 3-9 (0-3) 3-8 (0-4) 2-6(1-7)
Trial 6 4 0 (00) 3-8 (0-6) 3-9 (0-3) 3-0(1-3)
Total (24) 23-2(10) 22-6(1-2) 22-0(1-3) 16-7(81)

Hard items Trial 1 (4) 1-0(0-9) 1-0(1-3) 0-5 (0-7) 0-1(0-3)

Trial 2 2-6(1-3) 1-5(1-2) 1-3(0-8) 0-3 (0-5)
Trial 3 3-3(1-0) 2-4(1-1) 1-8(1-2) 0-7 (0-9)
Trial 4 3-6 (0-7) 30 (0-9) 2-5 (0-5) 1-8(1-1)
Trial 5 3-6 (0-9) 2-8(1-4) 2-5(1-2) 1-4(1-4)
Trial 6 3-8 (0-8) 31 (1-2) 2-9(1-5) 1-1(1-3)
Total (24) 17-9(4-3) 13-8(6-3) 11-2(3-9) 5-2 (4-6)

1 h delayed recall
Easy Items 4 0 (00) 3-8 (0-4) 40 (00) 2-4(1-3)
Hard Items 2-0(1-2) 1-9(1-3) 1-8(1-4) 0-7 (0-7)

24 h delayed recall
Easy Items 4 0 (00) 3-8 (0-4) 3-7 (0-5) 2-4(1-2)
Hard Items 1-9(1-3) 0-9(1-0) 1-1(1-3) 0-2 (0-4)
Rey Auditory Verbal Learning Test
List A Trial 1 (15) 5-5(1-5) 41 (2-1) 4-4(1-1) 2-9(1-5)
Trial 2 91 (1-9) 7-0(2-5) 6-5(1-3) 4-9(2-1)
Trial 3 10-8 (3-0) 8-7 (3-6) 7-2 (20) 5-7 (2-9)
Trial 4 12-3(21) 10-4(2-2) 9-7 (3-8) 5-9(3-1)
Trial 5 13-1(1-7) 11-8(2-6) 102 (3-3) 7-0(2-7)
List B (15) 5-6(1-9) 3-8(1-4) 4-2 (2-4) 2-7(1-0)
List A Trial 6 10-8(1-9) 8-5(5-1) 7-2 (3-8) 3-1 (2-9)
1 h delayed recall 7-8 (4-4) 3-9(4-1) 3-7 (3-4) 10(2-4)
24 h delayed recall 8-3 (2-9) 2-8 (4-0) 4-2 (3-5) 0-7 (2-0)

the UD group for a variety of materials. The both groups by trial 2. The rate of learning, as
most sensitive measures, however, tended to be indicated by the polynomial trends in the trial
recall of word lists, whether organized or not. data, did not differ between the two groups. On
Recall of structured prose material, in contrast, the RAVLT, list B could potentially interfere
was relatively intact. with the list A material learned over the
preceding 5 trials. Subsequent free recall of list
Learning supraspan material {Table 2) A revealed significant interference (P < 0-001).
The ability to learn supraspan material was The size of the effect, however, did not differ in
assessed by the RAVLT and the PALT. In both the two groups.
cases, the two groups showed clear evidence of
learning (P < 0001) with the exception of the Delayed recall (Table 2)
PALT easy items which approached ceiling in Delayed recall after 1 h and 24 h was assessed
 Cognitive function in depression 835

Table 3. Results (mean and standard deviation) for the control and three depressed groups on
immediate recall, cued recall and recognition of the three word lists A (uncategorized), B (categorized -
unorganized) and C (categorized - organized)
Depressed groups
(""nntrol
group Unimpaired Borderline Impaired
(W= 16) (N= 10) (N= 10) (N = &)
Measure Mean (s.D.) Mean (s.D.) Mean (s.D.) Mean (s.D.)

List A (uncalegorized word list)

Free Recall score (12) 6-6(1-9) 4-1 (1-5) 4-3(1-5) 3-2(1-7)
Recognition
Total correct (24) 21-6(1-6) 19-2(1-8) 19-9(1-9) 16-9(2-1)
True positive responses (12) 10-6(1-2) 7-7(1-9) 9-2(1-4) 7-8 (2-4)
True negative responses (12) 11-6(0-6) 11-5(10) 10-7(1-3) 91 (2-3)
False positive response (12) 0-4 (0-6) 0-5(1-1) 1-3(1-3) 2-8 (2-2)
False negative response (12) 1-9(1-2) 4-3(1-9) 2-8(1-4) 4-2 (2-4)
List B (categorized unclustered
word list)
Free Recall score (12) 7-2 (2-3) 5-8 (2-3) 51 (1-6) 3-5(1-3)
Cued Recall Score (12) 7-9(1-2) 7-3 (2-5) 60 (2-0) 41 (2-6)
Recognition
Total correct (24) 19-6(1-7) 19-0(2-4) 171 (2-9) 16-0(2-6)
True positive responses (12) 101 (1-2) 101 (1-5) 9-0(1-5) 81 (2-5)
True negative responses (12) 9-5(1-6) 8-9(1-7) 8-1 (2-4) 7-9(1-9)
False negative response (12) 2-5(1-6) 3-1(1-7) 3-9 (2-4) 41 (1-9)
False negative response (12) 1-9(1-2) 1-9(1-5) 3-0(1-5) 3-9 (2-5)
List C (categorized clustered word
list)
Free Recall score (12) 6-6(1-5) 4-8(1-6) 5-5(1-8) 3-8(1-4)
Cued Recall Score (12) 7-1 (1-3) 5-8(1-2) 6-0(1-8) 4-4 (2-3)
Recognition
Total correct (24) 21-0(1-5) 19-2(1-9) 18-6(1-4) 17-9(1-6)
True positive responses (12) 10-5(1-3) 9-3(1-5) 9-0(1-7) 8-5 (2-3)
True negative responses (12) 10-5(1-2) 9-9(1-6) 9-6(1-9) 9-4(1-3)
False positive response (12) 1-5(1-4) 21 (1-6) 2-4(1-9) 2-6(1-3)
False negative response (12) 1-5(1-3) 2-7(1-5) 3-0(1-7) 3-5 (2-8)

for the LM, RAVLT and PALT. Because of 24 h (P < 005). Finally, for the RAVLT, the
differences in overall performance of the groups UD group again showed differential effect of
prior to the delay, delayed recall was considered delay (P < 001) with a tendency for patients but
in relation to immediate recall (LM) or recall not the controls to continue to forget items over
after trial 5 of the two learning tests (RAVLT the 24 h delay period.
and PALT). With the exception of the PALT
easy items both groups recalled significantly less Cued recall and recognition (Table 3 and
material after delay. For the PALT Hard items, Figs 2a, b)
the control group showed no further forgetting Cued recall was assessed for the categorized lists
between 1 h and 24 h. The patients, however, A and B immediately after free recall. Overall
tended to continue forgetting with increasing cueing led to a significant increase in the number
delay (P<0-\0). A similar pattern was seen of items recalled, regardless of list and in both
with the LM test. Although the differential effect groups. There were no differential effects of
of delay on the raw recall scores failed to reach either list or group on the cueing effect.
significance, the groups showed a clear effect Recognition performance was assessed for all
when percentage recall scores were considered. three lists, A, B and C. Analysis of the
The delayed recall of the controls remained at recognition data was restricted to the' hits' (true
over 80 % of immediate recall levels, even after positive responses) and 'false alarms' (false
24 h. The UD group's performance, however, negative responses) as, with an equal number of
dropped to 63-6% after 1 h and 50-5% after target items and distracters, these provide a
836 R. G. Brown and others

false alarm results interpretable as sensitivity

and response bias. A dL' of zero is equivalent to
chance responding. Increasingly positive values
correspond to increasing ability to discriminate
previous items from novel items, whereas nega-
tive values correspond to worse than chance
performance. For C,, a value of zero implies
no bias. Increasing positive values imply an
increasingly conservative response bias (saying
that the item is novel) while increasing negative
values imply an increasingly liberal bias (saying
that the item is one presented previously).
The UD group showed a significantly lower
dL' (Fig. 2a) (P < 005). Across the two groups,
Unrelated Categorized Categorized
words unclustered clustered dL' differed significantly between lists (P < 001),
with a significantly higher sensitivity for the
unrelated list than for the categorized lists
(<001). The group by list interaction was
not significant. In contrast to the effect for
sensitivity, response bias (CJ (Fig. 2b) did not
differ between the two groups, although the data
in Fig. 2b suggests an increase in bias for the
unrelated list A. Across groups, there was a
strong effect of list on bias (P < 0-001).
Univariate analyses revealed that the bias was
significant and positive for List A (P > 0-001),
non-significant (i.e. zero bias) for list C, and
with a trend towards a negative bias for List B
(P < 0-10). However, this pattern of bias was the
same in the two groups. Thus, the recognition
-0-5 performance of the UD patients was charac-
Unrelated Categorized Categorized terized by a significant decrease in sensitivity
words unclustered clustered
(i.e. ability to discriminate previously presented
FIG. 2. Mean (and s.E.) signal detection parameters dL' (a) and CL (b) items from novel items), but with no abnormality
for the three word lists, measured in the control group (D) and the in response bias.
three depressed groups (H, UD; H, BD; • . ID).

Short-term memory (Brown-Peterson task)

complete description of recognition perform- (Fig. 3)
ance. Considering the two response types sep- The UD group recalled significantly fewer
arately revealed a significant difference between trigrams overall (P < 0-05), with a significant
the groups for hits (P < 001) but not for false group by delay interval interaction (P < 005).
alarms, with UD patients identifying fewer items. Post hoc analysis revealed that the two groups
In addition, for the hits, there was a significant did not differ for immediate recall (0 s delay),
group by list interaction (P < 005). Planned with both groups achieving near perfect per-
contrasts revealed that the source of the inter- formance (maximum score = 15). Significant
action lay in the difference between the groups differences (P < 0-001) between the groups,
for unrelated versus categorized lists (P < 0-05), however, were found at all other delay intervals.
rather than between the clustered and Nun- Analysis of intervals 5-30 s confirmed the
clustered categorized lists. significant main effects of group and interval.
Employing the procedure recommended by The interaction between these two effects,
Corwin et al. (1990), two independent para- however, was no longer significant. Thus, the
meters dL' and C, were derived from the hit and patients showed a deficit in recall after delay
 Cognitive function in depression 837

15
On the Token test, the control subjects made
relatively few errors, even with the strict scoring
(one attempt). Half (8/16) made 1 error, and 1
subject made 2 errors. The UD group made
significantly more errors (Mann-Whitney U test,
=3 10 P < 001) with 5/10 making 1 or 2 errors and
3/10 making 3 or more errors (maximum 9).
With the lenient scoring criterion 2/16 controls
made a single error, while 1 patient made a
single error and 2 made 2 errors. Performance
by this criterion did not distinguish the groups.
No significant deficit was found in the UD
group on the Weigl sorting test. All of the
control subjects and 9/10 of the patients were
5 10 20 30 able to sort the tokens successfully according to
Delay (seconds) both methods of classification (shape and
FIG. 3. Mean (and S.E.) recall score (maximum = 15) for the 5 delay colour).
intervals of the Brown-Peterson test, measured in the control group
( • ) and the three depressed groups (A, UD; • . BD; • . ID). Comparison of three depressed samples
Age, sex, years of education and reading
compared to the no-delay condition, but the ability and level of depression (Table 1)
degree of impairment was unrelated to the The three depressed groups (UD, BD and ID)
duration of the delay interval. did not differ in mean age, years of education,
Hamilton score or sex ratios. Two of the 3
Language and executive function (Table 4) bipolar patients were in the UD group and 1 in
In the verbal fluency tests the patients generated the BD group. A significant difference was
fewer words overall (P < 0-01). In both groups found between the groups for the Schonell
and to a similar extent, most words were reading test (P < 0-05). Post hoc comparisons
generated in the free condition and least in the revealed only that the UD group scored higher
letter ' s ' condition. All subjects generated more than the ID group.
words in the first 30 s of each condition than in
the second (P < 0-001). However, this effect was CAMCOG and MMSE (Table I)
the same in the two groups. There was a highly significant difference between

Table 4. Results (mean and standard deviation) for the control and three depressed groups on the
verbalfluencytasks
Depressed groups
Control
group Unimpaired Borderline Impaired
(JV= 16) (N= 10) (N= 10) (JV = 9)
Measure Mean (S.D.) Mean (S.D.) Mean (S.D.) Mean (S.D.)

Free condition
First 30 s 15-1 (2-2) 14-5(2-2) 13-4(5-5) 10 3 (3-5)
Second 30 s 12-9(30) 11-2(21) 10-9(2-8) 6-8 (3-0)
Total (60 s) 28-0 (3-5) 25-7 (3-3) 24-3 (6-9) 17-1 (5-3)
Boy's names
First 30 s 12-2(2-6) 9-9 (2-4) 10-0(30) 6-9 (20)
Second 30 s 8-3 (3-2) 6-3 (20) 5-9(1-4) 3-7 (2-2)
Total (60 s) 20-5 (5-2) 16-2(3-2) 15-9(3-9) 10-6(2-8)
Letter S
First 30 s 11-5(2-5) 9-4 (2-5) 6-9 (3-4) 5-7 (2-7)
Second 30 s 7-9 (3-0) 6-4(1-6) 5-5 (4-4) 3-0(1-9)
Total (60 s) 19-4(4-9) 15-8(3-6) 12-4(7-4) 8-7 (4-2)
838 R. G. Brown and others

the groups in total CAMCOG score (P < 0001), mean = 8-7, S.D. = 2-9; ID mean = 60, S.D. -
with each group differing significantly from each 2-1), and the recall of two of the separate word
other. For the MMSE total, however, the UD lists (A and C). No significant group differences
and BD groups did not differ significantly, a were found, however, except for the List (B)
reflection, perhaps, of the lower ceiling of the with the UD group recalling more items than the
MMSE and decreased sensitivity to identify ID group. Thus, the recall of prose material but
mild levels of impairment. Univariate statistics not word lists showed a clear gradient of
revealed that the three groups differed signifi- performance with increasing overall cognitive
cantly on all CAMCOG subscales. Post hoc impairment.
comparisons revealed that the ID group per-
formed significantly worse than the UD group Learning supraspan material {Table 2)
on all scales, and worse than the BD group for In general, the three groups could not be
Orientation, Memory and Calculation. In con- distinguished statistically for rate of learning on
trast, the UD and BD groups did not differ either the RAVLT and PALT tests. Clear
significantly on any scale. The groups differed evidence of learning was demonstrated in all
on the serial sevens item (UD group mean = 4-8, cases apart from the easy items on the PALT.
S.D. = 0-4; BD group mean = 3-5, S.D. = 1 1 ; ID Although not significant, the UD group tended
group mean = 22, S.D. = 20) (P < 001), with to improve performance most with practice in
the ID group significantly worse than the UD the learning tests and the ID group least.
group.
Delayed recall (Table 2)
WA IS- Verbal IQ and subscale scores Overall the tests failed to reveal any differential
(Table 1) ' effect between the groups of delay recall per-
There was a significant difference between the formance relative to immediate recall. For the
pro-rated Verbal IQ of the three groups (P < LM test, all groups showed a decrease in recall
001), with the UD group having a higher verbal from immediate to 1 h delay. Overall, however,
IQ than the others. The BD and ID groups did there was no significant decrease over the next
not differ. MA'NOVA with all subscale scores, 24 h. The effect of delay on recall did not differ
revealed a significant overall effect of group (P < between the groups. Considering delayed recall
001). Univariate tests revealed that the groups as a percentage of immediate recall, there was no
differed significantly on all subscales (P < 001) significant differences in performance between
except digit span. Post hoc comparisons between groups. For the PALT the effect of delay on
the groups on the scales with significant F recall was examined for the easy and hard items
statistics revealed that the UD and ID groups together. Across trial type (easy or hard) and
differed significantly in each case. Furthermore, recall delay, there was a significant difference
the UD group performed significantly better between the groups ( P < 0-001), with the ID
than the BD group on three of the tests: patients recalling less than the other two groups.
Vocabulary, Similarities and Arithmetic, but However, none of the two- or three-way inter-
not on Comprehension. action involving group were significant. Finally,
for the RAVLT there was a significant de-
Memory crease in recall from trial 5 (immediate recall)
Immediate recall {Table 2) to 1 h delayed recall (P < 0001), but not from
On the recall of the LM passages, a significant 1 h to 24 h (P > 010). Once again there was no
difference was shown between the groups (P < significant difference in the performance of the
005), with the ID group recalling significantly three groups across the two delay intervals.
less than the UD group. On the first trial of the
RAVLT, no clear or significant group differences Cued recall and recognition (Table 3 and
were found, although in this instance, there was Figs 2 a, b)
a tendency for the BD group to recall more than Comparing the free and cued recall performance
the UD group. A similar pattern was shown for of lists B and C for the three groups showed a
digit span (UD mean = 8-2, S.D. = 3-7; BD significant effect of cueing (P < 005) with more
 Cognitive function in depression 839

being recalled in the cued condition than in the

free condition. However, the effect of cueing was Short-term memory (Brown-Peterson task)
similar in the three groups and for the two lists.
For the recognition data attention was limited While there was a significant effect of delay on
to hits and false alarms. For the hits, there was recall (P < 0001), there was no overall difference
no significant difference, overall, between the between the groups and no group by delay
three groups. The effect of list (A, B, C) interaction. Thus, all subjects recalled less with
approached significance (P < 010) with a lower increasing recall delay.
hit rate for the unrelated list (A) than for the
categorized lists (B and C) (P < 0-05), but Language and executive function (Table 4)
with no difference between the latter two lists. Across the three verbal fluency conditions, there
The group by list interaction also approached was a significant difference between the groups
significance (P < 010). Subsequent one-way (P < 001), with the main difference between the
ANOVAs revealed no significant differences ID and the other two groups. However, this was
between the three groups for lists A or C, while the only significant difference between the
the result for list B approached significance groups. All groups generated most words in the
(P < 010) with the UD group tending to per- 'free' condition relative to the other two (P <
form best and the ID group worse. Of the three 0001), and in the first 30s of each condition
groups only the UD group showed a difference relative to the second (P < 0001).
in hits rates between the three lists ( P < 0 1 0 ) On the token test only one subject (in the
with the effect attributable to a lower hit rate in impaired group), made more than two errors
the unrelated list A than in the two categorized with the lenient scoring criterion. Errors with
lists (P < 005). For the false alarms there was a the strict criterion, however, were shown by
trend, across all three lists for the UD group subjects in each of the groups. Adopting the
to make the fewest number of false positive definition of impairment derived from the
responses and the ID group the most. However, control group data (2 or more errors in the strict
neither the group effect nor the group by list scoring criterion), three of the UD depressed
interaction were significant. and BD depressed groups were impaired on the
Of the signal detection parameters, there was token test. In contrast, almost all (8/9) of the ID
a significant difference in the sensitivity (d,') group, had an abnormal performance on this
between the groups (P < 001) (Fig. 2a), with the test (P < 0-05).
UD and ID groups differing significantly (P < The ID groups also tended to have problems
0001). Overall there was a significant effect of on the Weigl test. Of the UD group, 9/10 were
list on sensitivity (P < 005), but this list effect able to sort by two categories successfully, as
were similar in the three groups. In contrast to were 8/10 of the BD group. However, only 4/9
sensitivity, bias (C,) did not differ between the of the ID group were able to sort the shapes
groups when averaging across the three lists by more than one category. This difference
(Fig. 2b). A highly significant effect of list was approached significance (P < 0-10).
shown ( f < 0-001), with significant differences
between both random and categorized lists (P < Associations between clinical measures and
0001) and between the two categorized lists neuropsychological measures
(P < 005). In addition, the group by list inter- Because of the large number of neuro-
action was significant (P < 005). Planned con- psychological measures, the data were first
trasts revealed that this interaction was due to the subjected to a principle components analysis
pattern of response by the groups to the random with varimax rotation. A set of variables were
v. categorized lists (P < 005) rather than be- chosen after examining the correlation matrix to
tween the two categorized lists (P < 010). The eliminate redundant variables. The final set of
main finding, seen in Fig. 2b, was the higher variables produced a 5 factor solution account-
positive bias of the UD group for the random ing for 72% of the variance. Factor scores were
list, while the three groups did not differ for lists correlated with total Hamilton score, the two
B and C. factor scores derived from the SADS, and the
840 R. G. Brown and others

Table 5. Summary of previous studies on cognitive function in depressed patients

Significant No significant
Task impairment shown impairment shown Note

Immediate free recall

Prose passages Breslow et at. 1980
Hart et at. 1987 a, b
Watts & Cooper, 1989
Word lists, paired associates Sternberg & Jarvik, 1976
Weingartner el al. 1981
Calev el al. 1986
Roy-Byrne el al. 1986
Wolfe et al. 1987
Golinkoff& Sweeney, 1989
Watts el al. 1990
Effect of semantic organization Watts el al. 1990
Weingartner el al. 1981
Learning supraspan material Wolfe et al. 1987
Golinkoff & Sweeney, 1989
Delayed recall Sternberg & Jarvik, 1976 Weingartner & Silberman. 1982
Kopelman, 1986
Wolfe el al. 1987
Recognition memory
Hit rate (true +ve) Miller & Lewis, 1977 Dunbar & Lishman, 1984* 'Impairment only for words of
Dunbar & Lishman, 1984* positive hedonic tone
Cole & Zarit, 1984
Calev & Erwin, 1985
Watts et al. 1987
Wolfe et al. 1987
Golinkoff & Sweeney, 1989
False alarm rate (false -ve) Frith el al. 1983 Miller & Lewis, 1977 •Increase in vocalization
Watts et al. 1987* Dunbar & Lishman, 1984 condition, decrease in silent
Wolfe et al. 1987f Calev & Erwin, 1985 reading condition
Watts et al. 1987* •(•Impairment in unipolar patients
Wolfe et al. 1987f but not in bipolar
Sensitivity Dunbar & Lishman 1984* Miller & Lewis, 1977 •Normal for neutral words,
Watts et al. 1987 Dunbar & Lishman, 1984* increased for negative words and
Corwin et al. 1990f Corwin et al. 1990f decreased for positive words
timpaired only in more severely
depressed patients
Response bias Watts et al. 1987 Miller & Lewis, 1977* •All showed more positive
Dunbar & Lishman, 1984* (conservative) bias
Corwin et al. 1990*
Short-term memory
Digit span Breslow et al. 1980
Brown-Peterson paradigm Cohen et al. 1982* Cohen et al. 1982* •Impairment only in 'severely
depressed' patients
Language and executive function
Conceptual ability Savard et al. 1980* Savard et al. 1980* *Halstead Categories Test -
Hart et al. I987fff impaired only in bipolar patients
fWisconsin Card Sorting Test
Verbal fluency Robertson & Taylor, 1985 Wolfe etal. 1987* •Impairment only in bipolar
Hart et al. 1987 a patients
Wolfe et al. 1987*

individual SADS subscale scores. Of the 70 half (14/29) were taking antidepressant medi-
paired associations, only one was significant at cation at the time of assessment. This same ratio
the 0-01 level, close to chance level. was represented in each of the three depressed
subgroups (UD: 5/10; BD: 5/10; ID: 4/10).
Effect of antidepressant medication on Overall the medicated (N = 14) and unmedicated
cognitive function patients (N = 15) did not differ on the set of
Of the total sample of depressed patients, almost variables age, years of education, Hamilton
 Cognitive function in depression 841

score or CAMCOG total. To determine whether impairment of sufficient severity to warrant a

the presence or absence of medication had any diagnosis of dementia. Roth et al. (1986)
effect on neuropsychological function, indepen- recommended a cut-off of 80 as providing
dent of overall level of cognitive impairment, a acceptable levels of both sensitivity and speci-
representative set of 21 neuropsychological ficity for their mixed population of patients. In
measures were entered into a MANOVA (Logi- the present study a cut-off of 81 was employed
cal memory: immediate recall and 1 h delayed instead as this had the practical advantage of
recall as % of immediate recall. Rey AVLT: dividing the remaining patients into two approxi-
total score and 1 h delayed recall; PALT total mately equal halves. Thus, we obtained a sample
easy and total hard; Brown-Peterson recall after of patients classified as 'impaired' who con-
0 and 30 s; free recall and recognition score for stituted 31% of the total depressed sample.
word lists A, B and C and cued recall for lists B Although one might label such patients as
and C; total Verbal fluency score; Token test: 'demented', we will continue with the label
strict and lenient score). The results revealed no 'impaired' to avoid the implication that we have
significant difference between the medicated and provided a prevalence figure for dementia in
unmedicated depressed groups. Independent depression.
univariate F tests revealed no significant Having identified three subgroups of patients
differences between the groups for any measure, we could then turn to the second question. How
with the F score in all but two instances being normal were the unimpaired (UD) patients?
less than 1. Thus, the medication status of the Were there deficits in specific aspects of cog-
patients appeared to have no impact on their nition, despite globally defined intellectual func-
overall level of cognitive function, or on their tion within the normal 'range? The results
performance on the specific neuropsychological revealed clearly that the UD group were
tests. impaired on a wide range I of measures. Those
most sensitive to the presence of depression
included deficits in recall memory, particularly
after a delay, aspects of recognition memory,
short-term memory, verbalfluencyand language
DISCUSSION comprehension. How does this pattern of results
The first empirical question addressed by this compare with that reported elsewhere in the
study was the distribution of overall intellectual literature? The aspects of cognitive function
function in a sample of depressed patients. under consideration are shown in Table 5,
Specifically, was there any evidence for a distinct together with the results of the main published
subgroup of demented patients. The data studies.
revealed that CAMCOG scores were unimodally Considering first long-term memory and
distributed with an extended 'tail' towards the learning. The UD patients showed no significant
impaired end of the continuum (Fig. 1). The impairment on the immediate recall of the 'easy'
classification of patients as 'impaired' or 'un- items of the PALT or for the prose passages of
impaired', therefore, must be based on cut-offs the LM test. While the absence of impairment
derived from a normative sample. for the PALT may be due to ceiling effects, the
Methodologically, the description of the 'un- same explanation cannot account for the results
impaired' group was simply a matter of taking for the LM test. The lack of any significant
the range of values obtained from a normal impairment on this test contrasts with other
control group. Less straightforward was the studies which have found impaired immediate
classification of the remaining patients. As a memory for prose passages in depressed samples.
major concern of the study was the issue of Inadequate power of the statistical test may
dementia in depression it was not sufficient partly account for the lack of a significant result
simply to take the cut-off provided by the in the present case. In addition, however, the
minimum control group score on the CAMCOG. previous studies all used samples unselected for
Although all patients scoring less than this the severity of their overall level of cognitive
would lie outside of the 'normal' range, this is function. Our results suggest that immediate
not the same as saying that they had a cognitive recall of prose material may be a relatively
842 R. G. Brown and others

insensitive measure of memory dysfunction in recognition paradigms in depressed samples is

intellectually intact depressed patients. However, complicated by differences not only in the basic
other measures appear to be more sensitive, and paradigm, but also in the indices of performance
showed significant effects even with the present employed, all of which assess a different aspect
sample size and in globally intact subjects. of recognition memory. However, a majority
Impaired performance was observed in the recall of studies report the hit rate (true positive
of supra-span word lists (RAVLT and the responses), and almost all indicate impaired
various random, organised and categorized word performance in depressed subjects. The present
lists) and for 'hard' paired associates. In these study suggests that the level of structure may be
cases, the findings concur with the majority of important, with a significant decrease in hit rate
previous studies (see Table 5). being observed only for a list of unrelated
Also assessed was the effects on recall of words. Where the lists were made up of items
semantic categorization and clustering of the from a small number of categories, no im-
stimulus material. It was shown that these effects pairment in hit rate was observed.
were normal, a similar result to that shown by The consensus in the literature on hit rate is
Watts et al. (1990). Weingartner et al. (1981) not shared by other recognition memory para-
found categorization and clustering of the meters. Of those studies that report the false
material at input aided patients to the point that alarm rate (false positive responses) some find a
their performance was no longer significantly significant increase, some show a decrease, while
impaired. In the present study, however, per- others, including the present study, find no
formance remained impaired, even with or- significant difference. Only a few studies have
ganisation of the material. employed signal detection analysis as a way of
Distinct from the ability to recall supra-span quantifying recognition performance. Sensitivity
material after a single presentation, is the ability provides the main index of mnestic ability for
to learn that material with repetition. Verbal recognition performance. Across studies the
learning was assessed by the RAVLT and the results appear to depend upon the nature of the
PALT. In both tests, although overall per- material or the level of depression of the sample.
formance was poorer (with the exception of the The present study, however, provided clear
'easy' pairs), the rate of learning was relatively evidence of impairment, unrelated to depression
normal in the UD group. Surprisingly little data severity and in all lists assessed. An equally
exists on this aspect of memory in the literature, important aspect of memory function is response
and that which exists suggests impairment. bias, i.e. their willingness to commit themselves
Although it is difficult to generalize from such a to a decision about whether they have seen a test
small set of findings, the present study suggests item before. A more conservative response bias,
that this impairment in verbal learning with if a general characteristic of performance, might
repetition (as opposed to immediate recall), may cause subjects to perform less well on all memory
not be typical of all depressed patients. tasks not because the memories are less accessible
Although many studies have examined the but because the subjects lack confidence in them.
immediate recall of verbal material, surprisingly Several studies have shown a more significant
few have assessed recall after a delay. Those that positive (i.e. conservative) bias in their depressed
have provide no consistent pattern. In the samples. These findings contrast with those of
present study, however delayed recall, over both the present study, and those of Watts et al.
1 h and 24 h, was impaired for all material with (1987) which found no significant change.
the exception of the 'easy' paired associates. However, from Fig. 1 it can be seen that there
Even LM which showed no significant deficit in was at least a trend for the UD group to show a
immediate recall, showed an impairment in the greater level of positive bias for the list made up
UD patients after a delay. This suggests that of unrelated words.
delayed recall may provide a more sensitive It seems likely that, whatever their theoretical
index of mnestic capacity in depressed subjects. value, the two signal detection parameters of
The final aspect of long-term memory recognition memory performance are highly
examined in the present study was recognition. sensitive to differences in patient sample, ma-
Evaluation of the various studies employing terial and possibly method of testing. Without
 Cognitive function in depression 843

being able to generalize, therefore, we can say Cohen et al. (1982) found normal performance,
only that our sample of'unimpaired' depressed both immediately and after delay, for a group of
subjects showed a decreased sensitivity, with 3 'moderately depressed' patients (mean
increasing impairment across the patient groups Hamilton Depression Rating 21 -9). A group of 5
with increasing overall cognitive dysfunction. In 'severely depressed' patients (mean Hamilton
contrast, response bias was generally normal, at Depression Rating 44-6) showed impaired per-
least for lists comprized of categorized words. formance with greater deterioration with in-
Thisfindingon bias is important if the results of creasing delay interval. These findings contrast
the other memory tests are to be interpreted in with our own, at least with regard to the per-
terms of mnestic ability rather than as an artefact formance of the moderately depressed group
of a more conservative strategy in the sample with depression levels similar to our own UD
being studied. sample. However, the sample size of only 3
The discussion, to this point, has concentrated patients makes the results of Cohen and
on aspects of long-term memory function. In colleagues difficult to evaluate.
contrast to the large number of studies on this Thefinalaspects of cognitive function assessed
facet of memory there has been virtually no were language and executive function. Con-
systematic attention paid to short-term memory, ceptual ability was assessed by the Weigl test,
i.e. the retention of small amounts of information and showed no significant impairment in the
over durations measurable in seconds. At the UD group. Although there are few studies on
simplest level, it has been assessed using this aspect of cognition, problems with con-
measures of memory span for digits. Breslow et ceptual ability does not appear to be charac-
al. (1980) reports a significant decrease in span, teristic of patients with depression. However, a
a result which was not replicated in the present different picture is observed for language func-
sample in the UD group. tion and particularly verbal fluency. In the
A second approach to assessing short-term present study, despite the UD and control groups
memory has been to assess recall after short being matched for overall CAMCOG score, the
delays during which time rehearsal is prevented patients showed significantly impaired function
using the Brown-Peterson paradigm. In the on both the expressive language and compre-
present study the UD group showed an im- hension subscales. Both subscales include a
pairment in recall after a delay relative to number of different aspects of language related
immediate recall. The degree of deficit, however, tests. The comprehension subscale involves
was not delay dependent suggesting that the carrying out motor responses to command (e.g.
main factor was the introduction of a delay per 'Tap each shoulder twice with two fingers
se or the presence of the backward counting task keeping your eyes shut') and answering ques-
employed during the delay period. This latter tions (e.g. 'Was there radio in this country
effect might suggest that the patients have an before television was invented?'). The deficit on
increased susceptibility to interference from a the comprehension subscale is consistent with
second task. The absence of any differential the impairment shown by the UD group on the
interference effect on the RAVLT test seems to Token test. An important result on this latter
argue against this possibility. However, a critical test, however, was that the patients were
difference between the two tasks is that the Rey impaired only when attempting to carry out the
material had been learned, through repetition, command on the first attempt. After repetition
prior to the interference task, whereas in the of the command, no significant deficit was
short-term memory task the interference task apparent. Such a pattern of performance seems
occurred after a single presentation of the to argue against any basic deficit in linguistic
stimulus material and prevented any rehearsal. function. Rather, the poor performance may be
In this respect, the deficit on short-term memory more reasonably attributed to difficulty with
task may be a better indicator of the types of registering the command rather than com-
difficulties that patients may have with prehending it. The deficit, therefore, may be one
registering new information in every-day life. of attention or short-term memory.
Only one other study has utilized the Brown-
Peterson paradigm with a depressed sample. Expressive language on the CAMCOG was
assessed by defining words, naming pictures,
PSM 24
844 R. G. Brown and others

verbal fluency (animals in 60 s), sentence rep- research to determine the nature of the processes
etition ('no ifs ands or buts') and writing name underlying these deficits and their relationship
and address. Once again, the task of sentence to the more pervasive impairments in long-term
repetition may be sensitive to deficits in at- memory.
tention. One of the main reasons for the overall While it is important to be able to identify
deficit on the Expressive language subscale, cognitive dysfunction in depressed patients
however, was poor performance of the UD relative to controls, it is equally important to
group on the verbal fluency task, which con- distinguish degrees of dysfunction within a
tributes almost half of the total subscale score. A depressed sample. Specifically, which measures
deficit in verbal fluency was found not only in of cognitive function are most sensitive to
the CAMCOG, but also when tested separately cognitive decline and which remain relatively
for the free generation of words, and for words constant? An answer to this question may have
beginning with the letter 's', and boy's names. important clinical implications for the choice of
Similar findings are reported in other studies. tests to identify demented depressed patients or
The cause of thisfluencydeficit, however, is less to monitor change with treatment or progression
clear. If it represents a deficit in access to of the impairment. Comparisons between the
semantic memory one might expect to find some three depressed groups was made easier by the
variation in the level of impairment dependent fact that they were matched for age and years of
upon the particular demands of the task. In the education, Hamilton score and the proportions
present study, however, the patients appeared to receiving antidepressant medication. A signifi-
show a constant deficit regardless of the precise cant difference was found, however, for reading
nature of the fluency condition. Alternatively, ability as measured by the Schonell reading test.
the slowness in word generation may be a non- The UD group performed best, the ID group
specific feature of the slowness found in patients worse, and the BD group intermediate. This
with depression in their movement and spon- finding raises the possibility that the global
taneous speech and gesture ('psychomotor differences between the groups represent simply
retardation'). Unfortunately, no other time- a difference in pre-morbid cognitive function. A
dependent tasks were administered in the present number of facts, however, argue against this
study, making it difficult to determine the degree possibility, and against the idea that reading age
to which motor or 'cognitive' slowness is a is an unbiased indicator only of premorbid
general feature of the sample of patients studied. intelligence in a patient group. First, even using
To summarize, we have demonstrated cog- the Schonell score as a covariate, there was a
nitive dysfunction on a range of measures in a highly significant difference between the total
sample of depressed patients who show no CAMCOG score of the three groups (P <
global impairment. This provides clear evidence 0001). This probably reflects the fact that the
that cognitive dysfunction is a real feature of CAMCOG is not an IQ test, and that even
depressive illness, and that previous studies have normal subjects with low IQ's will still score
not been reporting averaged group effects biased outside the range characteristic of patients
by the influence of a proportion of patients with with cognitive impairment. Secondly, total
more severe global deterioration. The deficits on CAMCOG score is significantly correlated with
which the UD group were impaired were, with reading ability only in the patient group (r =
only a few exceptions, consistent with the 0-79, P < 0001). No such association exists in
existing literature on cognitive function in the control group (/• = 0-27, P > 010) suggesting
depression. In addition, however, the results that the association in the patients is due to the
point to some important aspects of impairment presence of some additional factor in the patient
not generally investigated, aspects which may group to which both measures are related.
relate more closely to 'real-life' memory Thirdly, one reason for a low reading score on
problems reported by patients. These relate to the Schonell might be a failure of the subjects to
poor performance on language comprehension attend to the precise spelling of the more difficult
tests (following a complex set of instructions), words. As most of these are phonetically regular,
and poor short-term memory when rehearsal is an attentive and motivated subject could pro-
prevented. It would be important for future duce an accurate pronunciation. Consistent with
 Cognitive function in depression 845

such an explanation, reading ability loaded on addition, a significant gradient of impairment

the same factor in the principal components was found for verbal fluency. An important
analysis as other attention related tasks such as question is the degree to which these various
serial sevens, the Token Test as scored by the aspects of cognitive impairment represent the
strict criteria and delayed recall on the Brown- effect of a single common factor in contrast to a
Peterson task. Even if the influence of Verbal IQ number of independent factors. In particular, a
is taken into account, reading ability and serial primary attentional deficit may have important
sevens still have a partial correlation of 0-60 implications for a wide range of tests, including
(P < 0-001). Finally, although the data are not tests of memory, particulary the immediate recall
presented here, a follow-up of the patients after of supraspan material. In contrast, attentional
recovery of their depression showed a significant factors may be expected to have less impact on,
improvement not only in overall cognitive for example, delayed recall where memory
function but also in reading ability. Together function per se may be more important.
thesefindingsindicate that the deficit in reading Unfortunately, the role of attention in cognitive
ability in the ID patients may be considered a function in depression has not been the subject
facet of their cognitive impairment, probably detailed investigation in the literature to date.
related to attention, and not a trait indicator of The present findings, with the relatively crude
lower pre-morbid intelligence. measures employed, can provide only a pre-
The data reported here suggests that the large liminary suggestion of its importance. However,
majority of tests revealed a gradient of function these behavioural data are consistent with the
with the UD group performing best, the ID evidence from Positron Emission Tomography
group worst, and the BD group intermediate. (PET) on the cohort of patients from which the
This pattern confirms the conclusion that cog- present sample was drawn (Bench et al. 1992).
nitive dysfunction in depression is continuously Regional cerebral blood flow (rCBF) in the
distributed with no evidence for a discretely depressed patients was significantly reduced in
defined demented subgroup. There was con- the regions of the anterior cingulate cortex and
siderable variation between the tests, however, left dorsolateral prefrontal cortex compared to
in the degree to which the three groups could controls, brain areas implicated more with
be distinguished statistically. This may relate attention and behavioural regulation than with
to the tests themselves (e.g. ceiling effects) or the memory function. These findings suggest that a
variability in performance within the depressed thorough investigation of attentional function
sample. In any event, it is probably unwise to in depression is justified, both in its own right
draw firm conclusions from the specific pattern and to determine its relationship to memory
of tests which show significant group differences impairment.
and those which do not. Finally, one unexpected finding from the
With the present sample sizes the three patient present study was the absence of significant
groups could be distinguished on the immediate relationship between cognitive function and any
recall of supraspan material, particularly the index of depression relating to severity, symp-
prose passage of the LM test. In contrast, on tomatology or treatment. The three depressed
this and other tests, the results failed to reveal groups, with global levels of impairment ranging
any systematic and significant difference between from normal to 'demented', had identical mean
the three groups for the rate of new learning, the Hamilton scores. This provides powerful evi-
effect of cueing on recall, and effect of delay on dence that cognitive dysfunction cannot be
recall. Of the various recognition memory considered as an epiphenomenon of depressive
parameters, sensitivity showed the clearest group symptomatology as has been suggested by some
differences. Importantly, significant group investigators. It further suggests that we need to
effects were non limited to memory function. look for different neurobiological explanations
The depressed groups could also be distinguished for the affective and the cognitive features of
on a number of tests which have a large depression (Dolan et al. 1992).
attentional component. These included serial
sevens, the ability accurately to follow a complex
verbal command, and the Weigl sorting test. In L.C.S. and C.J. B. were supported by a grant from
the Wellcome Trust.
29-2
846 R. G. Brown and others

Rate of forgetting in dementia and depression. Journal of

REFERENCES Consulting and Clinical Psychology 55, 101—105.
Hart, R. P., Kwentus, J. A., Wade, J. B. & Hamer, R. M. (19876).
Bench, C. J., Friston, K. J., Brown, R. G., Scott, L. C , Frackowiak, Digit symbol performance in mild dementia and depression.
R. S. J. & Dolan, R. J. (1992). The anatomy of melancholia - focal Journal of Consulting and Clinical Psychology 55, 236-238.
abnormalities of cerebral blood flow in major depression. Hasher, L. & Zacks, R. T. (1979). Automatic and effortful processes
Psychological Medicine 22, 607-615. in memory. Journal of Experimental Psychology 108, 356-388.
Bench, C. J., Friston, K. J., Brown, R. G., Frackowiak, R. S. J. & Jorm, A. F. (1986). Cognitive deficits in the depressed elderly: a
Dolan, R. J. (1993). Regional cerebral blood flow in depression review of some basic unresolved issues. Australian and New
measured by Positron Emission Tomography: the relationship Zealand Journal of Psychiatry 20, 11-22.
with clinical dimensions. Psychological Medicine 23, 579-590. Kiloh, L. G. (1961). Pseudo-dementia. Ada Psychiatrica
BrafT, D. L. & Beck, A. T. (1974). Thinking disorder in depression. Scandinavica 37, 336-351.
Archives of General Psychiatry 31, 456-459. Kopelman, M. D. (1986). Clinical tests of memory. British Journal of
Breslow, R., Kocsis, J. & Belkin, B. (1980). Memory deficits in Psychiatry 148, 517-525.
depression: evidence utilizing the Wechsler Memory Scale. McAllister, T. W. (1981). Cognitive functioning in the affective
Perceptual and Motor Skills 51, 541-542. disorders. Comprehensive Psychiatry 22, 572-586.
Brown, J. (1958). Some tests of the decay theory of immediate Miller, W. R. (1975). Psychological deficits in depression. Psycho-
memory. Quarterly Journal of Experimental Psychology 10, 12-21. logical Bulletin 82, 238-260.
Calev, A. & Erwin, P. G. (1985). Recall and recognition in Miller, E. & Lewis, P. (1977). Recognition memory in elderly patients
depressives: use of matched tests. British Journal of Clinical with depression and dementia: a signal detection analysis. Journal
Psychology 24, 127-128. of Abnormal Psychology 86, 84-86.
Calev, A., Korin, Y, Shapira, B., Kugelmass, S. & Lerer, B. (1986). Nelson, H. E. & McKenna, P. (1975). The use of current reading
Verbal and non-verbal recall by depressed and euthymic affective ability in the assessment of dementia. British Journal of Social and
patients. Psychological Medicine 16, 789—794. Clinical Psychology 14, 259-267.
Cohen, R. M., Weingartner, H., Smallberg, S. A., Pickar, D. & Newman, P. J. & Sweet, J.J. (1992). Depressive Disorders. In
Murphy, D. L. (1982). Effort and cognition in depression. Archives Handbook of Neuropsychological Assessment: A Biopsychosoeial
of General Psychiatry 39, 593-597. Perspective (ed. A. E. Puente and R. J. McCaffrey), pp. 263-307.
Cole, K. D. & Zarit, S. H. (1984). Psychological deficits in depressed Plenum Press: New York.
medical patients. Journal of Nervous and Mental Diseases 172, Peterson, L. R. & Peterson, M. J. (1959). Short-term retention of
150-155. individual items. Journal of Experimental Psychology 58, 193-198.
Cornell, D. G., Suarez, R. & Berent, S. (1984). Psychomotor Robertson, G. & Taylor, P. J. (1985). Some cognitive correlates of
retardation in melancholic and non-melancholic depression: affective disorders. Psychological Medicine 15, 297-309.
cognitive and motor components. Journal of Abnormal Psychology Roth, M., Tym, E., Mountjoy, C. Q., Huppert, F. A., Hendrie, H.,
93, 152-157. Verma, S. & Goddard, R. (1986). CAMDEX: a standardised
Corwin, J., Peselow, E., Feenan, K., Rotrosen, J. & Fieve, R. (1990). instrument for the diagnosis in the elderly with special reference to
Disorders of decision in affective disease: an effect of/?-adrenergic the early detection of dementia. British Journal of Psychiatry 149,
dysfunction. Biological Psychiatry 27, 813-833. 698-709.
Cummings, J. L. (1986). Subcortical dementia. Neuropsychology, Roth, M., Huppert, F. A., Tym, E. & Mountjoy, C. Q. (1988).
neuropsychiatryand pathophysiology. Brif/sA Journal of Psychiatry CAMDEX: the Cambridge Examination for Mental Disorders of
149, 682-697. the Elderly. Cambridge University Press: Cambridge.
Cummings, J. L. (1989). Dementia and depression: an evolving Roy-Byrne, P. P., Weingartner, H., Bierer, L. M., Thompson, K. &
enigma. Journal of Neuropsychiatry 1, 236-242. Post, R. M. (1986). Effortful and automatic cognitive processes in
Cummings, J. L. & Benson, D. F. (1984). Subcortical dementia. depression. Archives of General Psychiatry 43, 265-267.
Review of an emerging concept. Archives of Neurology 41,874-879. Savard, R. J., Rey, A. C. & Post, R. M. (1980). Halstead-Reitan
Dolan, R. J., Bench, C. J., Brown, R. G., Scott, L. C , Friston, K. J. Category Test in bipolar and unipolar affective disorders:
& Frackowiak, R. S. J. (1992). Regional cerebral blood flow relationship to age and phase of illness. Journal of Nervous and
abnormalities in depressed patients with cognitive impairment. Mental Diseases 168, 297-304.
Journal of Neurology, Neurosurgery and Psychiatry 55, 768-773. Schonell, F. (1942). Backwardness in the Basic Subjects. Oliver &
Dunbar, G. C. & Lishman, W. A. (1984). Depression, recognition- Boyd: London.
memory and hedonic tone. A signal detection analysis. British Spitzer, R. L., Endicott, J. & Robins, E. (1977). Research Diagnostic
Journal of Psychiatry 144, 376-382. Criteria for a Selected Group of Functional Disorders. Biometrics
Endicott, J. & Spitzer, R. L. (1978). A diagnostic interview. The Research Division, New York State Psychiatric Institute: New
York.
Schedule for Affective Disorders and Schizophrenia. Archives of
Spreen, O. & Benton, A. L. (1969). Neurosensory Centre Com-
General Psychiatry 35, 837-844.
prehensive Examination for Aphasia. Neuropsychology Labora-
Folstein, M. F., Folstein, S. E. & McHugh, P. R. (1975). 'Mini-
tory, Department of Psychology, University of Victoria: Victoria,
mental state': a practical method for grading the cognitive state of
BC, Canada.
patients for the clinician. Journal of Psychiatric Research 12,
Sternberg, D. E. & Jarvik, M. E. (1976). Memory function in
189-198.
depression: improvement with antidepressant medication. Archives
Frith, C. D., Stevens, M., Johnstone, E. V., Deakin, J. F. W., Lawler,
of General Psychiatry 33, 219-224.
P. & Crow, T. J. (1983). Effects of ECT and depression on various Taylor, E. M. (1959). Psychological Appraisal of Children with
aspects of memory. British Journal of Psychiatry 142, 610-617. Cerebral Defects. Harvard University Press: Cambridge, Mass.
Golinkoff, M. & Sweeney, J. A. (1989). Cognitive impairments in Watts, F. N. & Cooper, Z. (1989). The effects of depression on
depression. Journal of Affective Disorders 17, 105-112. structural aspects of the recall of prose. Journal of Abnormal
Hachinski, V. C , Iliff, L. D., Zilkha, E., Du Boulay, G. H., Psychology 98, 150-153.
McAllister, V. L., Marshall, J., Ross Rissell, R. W. & Symon, L. Watts, F. N., Morris, L. & MacLeod, A. K. (1987). Recognition
(1975). Cerebral blood flow in dementia. Archives of Neurology 32, memory in depression. Journal of Abnormal Psychology 96,
632-637. 273-275.
Hamilton, M. (1960). A rating scale for depression. Journal of Watts, F. N., Dalgleish, T., Bourke, P. & Healy, D. (1990). Memory
Neurology, Neurosurgery and Psychiatry 23, 56-62. deficits in clinical depression: processing resources and the structure
Hart, R. P., Kwentus, J. A., Taylor, J. R. & Harkins, S. W. (1987a). of materials. Psychological Medicine 20, 345-349.
 Cognitive function in depression 847

Wechsler, D. (1945). A standardized memory scale for clinical use. Gerdt, C. (1981). Cognitive processes in memory. Archives of
Journal of Psychology 19, 87-95. General Psychiatry 38, 42-47.
Wechsler, D. (1986). The Wechsler Adult Intelligence Scale - Revised. Widlocher, D. & Hardy-Bayle, M-C. (1989). Cognition and control
Incorporating British Amendments to the Test Administration of action in psychopathology. European Bulletin of Cognitive
Section. The Psychological Corporation Ltd.: New York. Psychology 9, 1-33.
Weigl, E. (1941). On the psychology of so-called processes of Willner, P. (1984). Cognitive functioning in depression: a review of
abstraction. Journal of Abnormal and Social Psychology 36, 3-33. theory and research. Psychological Medicine 14, 807-823.
Weingartner, H. & Silberman, E. (1982). Models of cognitive Wolfe, J., Granholm, E., Butters, N., Saunders, E. & Janowsky, D.
impairment: cognitive changes in depression. Psychopharmaco- (1987). Verbal memory deficits associated with major affective
logical Bulletin 18, 27-42. disorders: a comparison of unipolar and bipolar patients. Journal
Weingartner, H., Cohen. R. M., Murphy, D. L., Martello, 3. & of Affective Disorders 13, 83-92.