Neuropsychology Copyright 2004 by the American Psychological Association

2004, Vol. 18, No. 2, 212–218 0894-4105/04/$12.00 DOI: 10.1037/0894-4105.18.2.212

Recognition Accuracy and Response Bias to Happy and Sad Facial

Expressions in Patients With Major Depression
Simon A. Surguladze Andrew W. Young
King’s College London University of York

Carl Senior Gildas Brébion, Michael J. Travis, and

Aston University Mary L. Phillips
King’s College London

Impaired facial expression recognition has been associated with features of major depression, which
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

could underlie some of the difficulties in social interactions in these patients. Patients with major
This document is copyrighted by the American Psychological Association or one of its allied publishers.

depressive disorder and age- and gender-matched healthy volunteers judged the emotion of 100 facial
stimuli displaying different intensities of sadness and happiness and neutral expressions presented for
short (100 ms) and long (2,000 ms) durations. Compared with healthy volunteers, depressed patients
demonstrated subtle impairments in discrimination accuracy and a predominant bias away from the
identification as happy of mildly happy expressions. The authors suggest that, in depressed patients, the
inability to accurately identify subtle changes in facial expression displayed by others in social situations
may underlie the impaired interpersonal functioning.

Defining features of unipolar depression include depressed
mood and anhedonia, which, with increasing severity of illness,
can lead to a restriction of affective range and social dysfunction.
The ability to recognize facial expressions displayed by others is
highly developed in human and nonhuman primates and is ob-
served across many cultures (Ekman & Friesen, 1971). In de-
pressed patients, however, depressed mood has been associated
with specific abnormalities in the identification of facial expres-
sions (Cooley & Nowicki, 1989; Wexler, Levenson, Warrenburg,
& Price, 1994), negative cognitions regarding the self, and dys-
functional appraisal of social events and situations (Beck, 1976).
These abnormalities may lead to impaired interpersonal function-
ing (e.g., Gotlib & Asarnow, 1979).
Previous studies have examined the nature of the impairment in
perception of facial expressions in depressed patients. Some in-
vestigators exploring discrimination accuracy of facial emotional
stimuli in these patients have provided evidence for a generalized
deficit in the recognition of all emotions (and even in the recog-
nition of nonemotional stimuli). Deficits have been demonstrated,
for example, in the recognition of faces expressing happiness
(Jaeger, Borod, & Peselow, 1987), sadness and interest (Rubinow
Simon A. Surguladze, Section of Neuroscience & Emotion, Division of
Psychological Medicine, Institute of Psychiatry, King’s College London,
London; Andrew W. Young, Department of Psychology, University of
York, Heslington, York, United Kingdom; Carl Senior, Neuroscience
Research Institute, Aston University, Birmingham, United Kingdom; Gil-
das Brébion, Michael J. Travis, and Mary L. Phillips, Division of Psycho-
logical Medicine, Institute of Psychiatry, King’s College London.
This study was supported by the James McDonnell Pew Foundation. We
thank S. Rabe-Hesketh for help with data analysis and Chris Andrew for
support in stimulus preparation.
Correspondence concerning this article should be addressed to Simon A.
Surguladze, Division of Psychological Medicine, P.O. Box 63, Institute of
Psychiatry, King’s College London, De Crespigny Park, London SE5 8AF,
United Kingdom. E-mail: sphasis@iop.kcl.ac.uk
& Post, 1992), and fear, anger, surprise, disgust, happiness, sad-
ness, and indifference (Persad & Polivy, 1993). In one study,
however, impairments in performance of both visuospatial and
affective tasks (identification of neutral, happy, sad, fearful, and
angry expressions) were demonstrated in depressed patients (As-
thana, Mandal, Khurana, & Haque-Nizamie, 1998), suggestive of
a general visual perceptive rather than a specific emotion recog-
nition deficit.
In another study, although depressed patients were not signifi-
cantly impaired in tasks requiring matching pictures of emotional
faces, they were impaired in verbal labeling of all emotional
(anger, happiness, sadness, fear, disgust, surprise) and neutral
(Feinberg, Rifkin, Schaffer, & Walker, 1986) faces. These findings
were interpreted as evidence of a specific deficit in the expressive
domain rather than a visual perceptual deficit in these patients.
Conversely, depressed patients compared with healthy volun-
teers have been reported to be significantly impaired in the recog-
nition of both sad and happy, but not neutral, faces (Mikhailova,
Vladimirova, Iznak, Tsusulkovskaya, & Sushko, 1996). In this
study, depressed patients retested in remission did not demonstrate
this deficit in emotion processing, suggesting a state rather than
trait deficit in emotion processing in depression.
Other studies have revealed emotion-specific abnormalities in
depressed patients, with patients demonstrating negative percep-
tual bias (i.e., recognizing significantly more sadness in facial
expressions than healthy volunteers; Bouhuys, Geerts, & Gordijn,
1999; R. C. Gur et al., 1992; Hale, 1998; Matthews & Antes,
1992). There have been additional reports in depressed patients of
impaired recognition of positive facial expressions (Murphy et al.,
1999; Suslow, Junghanns, & Arolt, 2001), diminished emotional
responses to pleasant pictorial stimuli (Sloan, Strauss, Quirk, &
Sajatovic, 1997; Sloan, Strauss, & Wisner, 2001), and increased
response times to happy compared with sad emotional words.
Findings to date from studies examining emotion processing in
depressed patients are, therefore, inconsistent, with evidence for
and against the presence of a general visual perceptual deficit,

212
 RECOGNITION AND RESPONSE BIAS IN DEPRESSION
evidence for impaired identification of all categories of emotion,
and evidence for impairments in the identification of specific
emotions.
There are several potential explanations for the discrepant find-
ings. First, there were differences in the types of patient popula-
tions recruited and examined in the studies. In some studies,
patients were restricted to those with unipolar depression (e.g.,
Feinberg et al., 1986; Jaeger et al., 1987; Persad & Polivy, 1993),
whereas in others, both unipolar and bipolar depressed patients
were examined (R. C. Gur et al., 1992; Rubinow & Post, 1992).
Second, there were differences across the studies in the types of
stimuli used. In some of the studies, facial expressions from a
standardized series (Ekman & Friesen, 1976) were used (Feinberg
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
et al., 1986; Persad & Polivy, 1993), whereas other investigators
This document is copyrighted by the American Psychological Association or one of its allied publishers.
used novel series of photographs (Asthana et al., 1998; R. C. Gur
et al., 1992) or schematic faces (Bouhuys et al., 1999; Suslow et
al., 2001). Furthermore, different categories of emotional expres-
sion (negative or positive) have been used in the different studies.
Third, the ability of depressed patients to identify facial expres-
sions of milder intensity or those presented for shorter durations
remains unexplored. In everyday life, humans process a wide
range of emotional stimuli displayed by others, including signals
less intense than prototypical facial expressions from standardized
series or those displayed for brief rather than long durations.
Studies using event-related potentials and rapid presentation of
facial expressions have demonstrated that healthy participants
were able to discriminate between emotional and neutral facial
expressions presented for durations as brief as 100 to 200 ms
(Junghoefer, Bradley, Elbert, & Lang, 2001; Sato, Kochiyama,
Yoshikawa, & Matsumura, 2001). It has been argued that the
recognition of rapid presentations of emotional expressions might
be impaired in depressed patients because of general slowing of
cognitive processes (Cooley & Nowicki, 1989). The ability of
depressed patients to identify rapid presentations of facial expres-
sions remains unexamined, however.
Previous findings in healthy participants have also demonstrated
that accuracy in recognition of facial expressions decreases as the
intensity of emotional expression displayed in the face is reduced
(e.g., Calder, Young, Rowland, & Perrett, 1997). It is possible that
specific impairments in the identification of subtle changes in
emotional expression, rather than in the identification of prototyp-
ical displays of emotion, may exist in depressed patients. As noted,
intact matching of prototypical displays of emotion has been
demonstrated in these patients (Feinberg et al., 1986). There is,
therefore, a rationale for the examination of the ability of de-
pressed patients to identify facial expressions presented for short
durations or depicting milder intensities of emotion.
Finally, although there is evidence from previous studies for the
presence of impaired recognition accuracy and a negative response
bias to facial expressions in depressed patients, the nature of the
relationship between these processes in depressed patients has not
been examined.
We aimed to examine the recognition accuracy of and response
bias toward positive and negative facial expressions in patients
with unipolar depression and healthy volunteers. We chose happy
and sad emotional expressions, reasoning that these emotional
displays might be of particular relevance to the negative schemata
demonstrated by depressed patients, in particular, the negative
cognitions regarding the abilities of the self compared with others
213
(Beck, 1976). It is, therefore, possible that depressed patients
identify with displays of sadness, but not happiness, in others. We
used happy, neutral, and sad facial expressions from a standardized
series (Ekman & Friesen, 1976) presented at different durations
and transformed with computer software (Facial Expressions of
Emotion: Stimuli and Tests; Young, Perrett, Calder, Sprengel-
meyer, & Ekman, 2002) to depict different intensities of emotion.
We were interested in examining measures of recognition ac-
curacy and response bias. There are two commonly used ap-
proaches to derive these measures: signal-detection theory, with
the discrimination measure d⬘ and the bias measure C (e.g., Stanis-
law & Todorov, 1999), and the two-high threshold (2HT) theory,
with the discrimination accuracy measure Pr and the response bias
measure Br (Corwin, 1994). The latter approach is especially
useful when numbers of criteria items (targets) and distractors
differ, as was the case in the current study. We, therefore, used the
2HT theory to compute measures of recognition or discrimination
accuracy and response bias to happy and sad facial expressions in
both groups. In this study, the discrimination accuracy measure
represented the ability to discriminate among neutral, happy, and
sad expressions. The response bias measure reflected the tendency
of participants, when uncertain about the category to which a facial
expression should belong, to categorize the expression as emo-
tional (happy or sad) rather than neutral.
Findings from previous studies enabled us to make several
predictions in depressed patients.
Hypothesis 1: We predicted a negative response bias toward
facial expressions such that (a) depressed patients would
demonstrate a smaller response bias to happy expressions
(i.e., compared with healthy volunteers, they would less fre-
quently identify happy and neutral faces as happy) and (b)
depressed patients would have a greater response bias to sad
expressions (i.e., compared with healthy volunteers, they
would more frequently identify sad and neutral faces as sad).
Hypothesis 2: We predicted a reduced recognition accuracy
of happy and sad facial expressions overall compared with
healthy volunteers.
Hypothesis 3: We predicted a significantly greater negative
response bias and impaired recognition accuracy in response
to expressions presented for shorter rather than longer dura-
tions and those expressions depicting milder rather than more
severe intensities of emotion.
Hypothesis 4: We predicted a significant positive correlation
between the magnitude of the negative response bias and
impairment in recognition accuracy and the severity of de-
pression measured by standardized rating scales.
Method
Participants
Twenty-seven patients meeting Diagnostic and Statistical Manual of
Mental Disorders (fourth edition; American Psychiatric Association, 1994)
criteria for major depressive disorder were recruited from inpatient and
outpatient services of the Bethlem Royal and Maudsley Hospitals. All
patients had a chronic, recurrent depressive disorder, and none was tested
in the first episode of illness. The mean duration of illness (since first
 214 SURGULADZE ET AL.

episode) in patients was 10.1 years (median duration of illness ⫽ 9 years).
As a control group, 29 healthy volunteers without a history of depression
were recruited from the local community and from a pool of employees at
the Institute of Psychiatry. They were matched with patients for age,
gender, and educational achievement (Table 1). Ethical approval was
obtained from the Ethical Committee of the South London and Maudsley
Trust and Institute of Psychiatry. All participants were right-handed (Old-
field, 1971). Patients were excluded from the study if they reported a
history of mania, psychosis, alcoholism, or organic brain syndrome or if
they scored less than 24 on the Mini-Mental State Exam (Folstein, Folstein,
& McHugh, 1975). All participants completed the Recognition Memory
Test–Faces (Warrington, 1984) as a measure of nonemotional face percep-
tion. Depressed patients did not differ significantly from controls in per-
formance on this task, t(36.1) ⫽ 1.7, p ⬎ .05 (see Table 1).
All participants completed the Beck Depression Inventory (BDI; Beck,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
identities from a standardized series (Young et al., 2002), each displaying
expressions of happiness, sadness, or neutral expressions. Both happy and
sad faces were presented within the same task to avoid task-dependent
systematic errors and allow for a direct comparison of responses to happy
and sad faces. Each emotional facial expression was morphed using com-
puter software with the facial expression of the same individual to depict
two different intensities (50% and 100%) of the emotion, and each was
then presented twice during the task: for 100 ms and 2,000 ms. In the same
task, 10 neutral faces were presented at two durations each: 100 ms
and 2,000 ms. Participants, therefore, viewed 100 stimuli during the
experiment: 10 facial identities, each displaying two emotions, happy and
sad, at two different intensities of emotion and for two different durations.
Each face was presented individually, with an interstimulus interval
of 1,500 ms, during the first 500 ms of which was displayed a fixation
cross. Participants were instructed that they would view either emotional

Ward, Mendelson, Mock, & Erbaugh, 1961). Controls with a score higher (sad or happy) or neutral faces and were requested to label each facial
This document is copyrighted by the American Psychological Association or one of its allied publishers.

than 9 on the BDI or with a history of depression were excluded from the
study. The mean BDI score of the controls was significantly lower than that
of the patients, t(31.5) ⫽ ⫺12.2, p ⬍ .01. The Hamilton Depression Rating
Scale (HAMD; Hamilton, 1960) was also completed for each patient (see
Table 1). To ensure that the experimenter was unaware of the depression
severity rating of patients during task performance, these ratings were
calculated after patients participated in the task. All patients were taking
some type of antidepressant medication: selective serotonin reuptake in-
hibitors (n ⫽ 9), serotonin and noradrenalin reuptake inhibitors (n ⫽ 8),
monoamine oxidase inhibitors (n ⫽ 4), tricyclic antidepressants (n ⫽ 3),
noradrenalin reuptake inhibitors (n ⫽ 2), mirtazapine (n ⫽1), lithium (n ⫽
5), and other mood stabilizers, including carbamazepine and lamotrigine
(n ⫽ 4).
Procedure
A facial expression recognition task was developed to present facial
expressions at different intensities and for different durations. In this task,
participants viewed randomized pictures on a computer screen of 10 facial
expression as happy, sad, or neutral by moving a computer joystick
accordingly in one of the three directions. Before testing, all participants
performed a practice trial to ensure they were able to perform the task.
Statistical Analysis
Raw data were transformed into measures of accuracy and response bias
according to the 2HT model (Corwin, 1994). Discrimination accuracy was
computed for the two separate subsets of targets (i.e., either sad or happy
faces [targets] vs. neutral faces [distractors]): Pr ⫽ (number of hits ⫹ 0.5/
number of targets ⫹ 1) ⫺ (number of false alarms ⫹ 0.5/number of
distractors ⫹ 1). Response bias was computed according to false-alarm
scores (tendency to label a neutral face as happy or sad) in two separate
subsets of sad or happy faces versus neutral faces (distractors): Br ⫽
(number of false alarms ⫹ 0.5/number of distractors ⫹ 1)/(1 – Pr).
Discrimination accuracy and response bias values were, therefore, com-
puted with regard to measures of recognition accuracy of neutral as well as
emotional facial expressions. High accuracy values would indicate an
ability to discriminate accurately among sad, happy, and neutral expres-
sions. Higher response bias scores would indicate a tendency to misidentify
neutral faces as emotional (either sad or happy).

Table 1
Results
Demographic and Clinical Data
Variable Patients (n ⫽ 27) Controls (n ⫽ 29) Response Accuracy
Because the data were not normally distributed, they were
Mean age (years)
M 46.9 43.0 log-transformed. Log-transformed scores of response accuracy
SD 11.2 11.9 were entered in 2 ⫻ 2 ⫻ 2 repeated measures analysis of variance
Gender (ANOVA) with emotion (sad, happy), intensity (50% and 100%),
Female 15 17 and duration (100, 2,000 ms) as within-subject factors and group
Male 12 12
Education
(patients, controls) as the between-subject factor. There were sev-
M 12.5 13.8 eral significant main effects: emotion, F(1, 42) ⫽ 176.1, p ⬍ .01;
SD 3.2 2.1 intensity, F(1, 42) ⫽ 197.7, p ⬍ .01; duration, F(1, 42) ⫽ 69.2,
Illness duration (years) p ⬍ .01; and group, F(1, 42) ⫽ 26.2, p ⬍ .01. The degrees of
M 10.1 freedom are smaller than expected (54) because of some missing
SD 8.3
RMT data.
M 38.7 41.6 Comparison of the mean values for discrimination accuracy
SD 6.9 4.2 indicated that in both groups, these significant effects were the
HAMD result of sad expressions being recognized less accurately than
M 16.9
SD 5.5
happy ones, expressions with higher intensity being recognized
BDI more accurately than those with lower intensity, and stimuli with
M 33.0 3.1 longer durations being recognized more accurately than those with
SD 9.9 3.5* shorter durations. Patients were less accurate overall compared
Note. RMT ⫽ Recognition Memory Test–Faces; HAMD ⫽ Hamilton
with controls.
Depression Rating Scale; BDI ⫽ Beck Depression Inventory. Significant interactions were observed for the following: Emo-
* p ⬍ .001. tion ⫻ Group, F(1, 42) ⫽ 7.9, p ⬍ .01; Duration ⫻ Group, F(1,
 RECOGNITION AND RESPONSE BIAS IN DEPRESSION
42) ⫽ 19.8, p ⬍ .01; and Emotion ⫻ Duration ⫻ Group, F(1,
42) ⫽ 5.0, p ⬍ .05. Comparison of mean values revealed that the
Emotion ⫻ Group interaction reflected the significantly greater
impairment in patients in the discrimination of sad rather than
happy expressions. To examine which of the duration conditions
contributed most to the interaction of Emotion ⫻ Group, a post hoc
ANOVA was performed for each presentation duration. A signif-
icant Emotion ⫻ Group interaction was observed for the 100 ms,
F(1, 42) ⫽ 9.8, p ⬍ .01, but not for the 2,000 ms, F(1, 50) ⫽ 0.06,
p ⫽ .80, presentation duration condition. In other words, the
accuracy of facial emotion discrimination in patients did not differ
from that of controls for facial expressions presented for 2,000 ms
but did differ for facial expressions presented for 100 ms.
Because there was no additional interaction of Emotion ⫻
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Group ⫻ Duration with intensity, mean measures of discrimination
This document is copyrighted by the American Psychological Association or one of its allied publishers.
accuracy for both intensities (50% and 100%) of each emotional
expression (sad and happy) were computed for the 100-ms condi-
tion in both groups. Between-group t tests of these mean measures
revealed that patients were significantly impaired compared with
controls in the discrimination of both happy and sad facial expres-
sions at 100-ms duration but significantly more impaired in the
discrimination of sad expressions, t(51) ⫽ 5.6, p ⬍ .01, and
t(54) ⫽ 4.4, p ⬍ .01, for sad and happy expressions, respectively
(Table 2).
Overall, therefore, depressed patients were less accurate com-
pared with controls in the discrimination of happy and sad facial
expressions of either intensity presented rapidly (100 ms). This
impairment was more pronounced for sad expressions. Patients
were not significantly different from controls in their ability to
accurately discriminate happy and sad facial expressions of either
intensity presented for longer (2,000 ms) durations.
Response Bias
Log-transformed measures of response bias were entered into a
2 ⫻ 2 ⫻ 2 repeated measures ANOVA with emotion (sad, happy),
intensity (50%, 100%), and duration (100, 2,000 ms) as within-
subject variables and group (patients, controls) as the between-
subject variable. There were significant main effects of emotion,
F(1, 54) ⫽ 10.3, p ⬍ .01; intensity, F(1, 54) ⫽ 158.6, p ⬍ .01; and
group, F(1, 54) ⫽ 7.8, p ⬍ .01. Comparison of mean values for the
Table 2
Log-Transformed Data of Discrimination Accuracy Scores
Patients
Condition M SD
100% happy, 100 ms ⫺0.53 0.49
50% happy, 100 ms ⫺0.91 0.53
50% sad, 100 ms ⫺1.75 0.59
100% sad, 100 ms ⫺1.18 0.65
100% happy, 2,000 ms ⫺0.33 0.49
50% happy, 2,000 ms ⫺0.55 0.49
50% sad, 2,000 ms ⫺1.16 0.56
100% sad, 2,000 ms ⫺0.57 0.37
a
Between-group post hoc t tests after analysis of variance for main effects of emotion, intensity, duration, and
diagnostic group. In these comparisons, the mean values for discrimination accuracy for happy and sad facial
expressions of both intensities (50% and 100%) were used.
215
response bias indicated that the main effect of emotion reflected
the greater bias in both groups toward labeling neutral expressions
as sad rather than happy. The main effect of intensity indicated the
propensity for a greater bias in both groups toward labeling more
intense expressions as emotional regardless of their valence (sad or
happy). The main effect of group reflected a generally more
conservative (smaller) response bias in patients compared with
controls. In other words, patients had a greater than normal ten-
dency to label any expression as neutral rather than emotional.
A significant Emotion ⫻ Intensity ⫻ Group interaction was
observed, F(1, 54) ⫽ 10.3, p ⬍ .01. To further investigate this
interaction, a post hoc ANOVA was performed for each intensity
(50% and 100%) of facial expression. There was no significant
Emotion ⫻ Group interaction for facial expressions of 100%
intensity of either emotion, F(1, 54) ⫽ 0.1, p ⫽ .80. For facial
expressions of 50% intensity of emotion, the Emotion ⫻ Group
interaction was significant, F(1, 54) ⫽ 6.1, p ⬍ .05. Subsequently,
separate analyses were performed for each presentation duration
(100 ms and 2,000 ms) for both emotions presented at 50%
intensity. There was a significant Group ⫻ Emotion interaction for
the 2,000-ms condition, F(1, 54) ⫽ 8.6, p ⬍ .01, but not for the
100-ms condition, F(1, 54) ⫽ 1.1, p ⫽ .30. Thus, a significant
interaction of Emotion ⫻ Group on response bias occurred for
facial expressions of 50% intensity presented for 2,000 ms.
To explore this interaction further, we performed between-
groups post hoc t tests. These demonstrated that controls had a
greater bias toward happy expressions of 50% intensity presented
for 2,000 ms compared with patients, t(54) ⫽ 3.8, p ⬍ .01.
Because the response bias variable was computed from two mea-
sures (correct recognitions and false-positive responses), the
greater response bias in controls reflected their tendency to more
frequently label the neutral faces as happy as well as the tendency
to more frequently label the happy expressions as happy. On the
other hand, the smaller response bias in patients toward the happy
expressions suggest a greater tendency of patients compared with
controls to label these happy and neutral faces as neutral rather
than happy. In the same condition (facial expressions of 50%
intensity of emotion presented for 2,000 ms), there was no signif-
icant difference between groups in response bias toward sad ex-
pressions (Figure 1).
Controls
M SD Significance of difference
⫺0.22 0.12 t(54) ⫽ 4.4, p ⬍ .001a
⫺0.41 0.24 t(54) ⫽ 4.4, p ⬍ .001a
⫺1.02 0.39 t(51) ⫽ 5.6, p ⬍ .001a
⫺0.53 0.36 t(51) ⫽ 5.6, p ⬍ .001a
⫺0.15 0.12 ns
⫺0.26 0.16 ns
⫺1.01 0.59 ns
⫺0.40 0.35 ns
 216 SURGULADZE ET AL.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
Figure 1. Response bias in the 100-ms (a) and 2,000-ms (b) conditions.
Graphs demonstrate the log-transformed scores for response bias to the
different intensities of happy and sad facial expressions. **After analysis
of variance, between-group t tests demonstrated that healthy volunteers,
compared with patients, had a significantly greater response bias to happy
expressions of 50% intensity presented for 2,000 ms, t(54) ⫽ 3.8, p ⬍ .01.
Correlations Between Response Accuracy and Bias Scores
and Depression Severity Ratings
In patients, significant negative correlations were demonstrated
between HAMD scores and discrimination accuracies for 50%
intensity sad expressions presented for 100 ms (Spearman’s ␳ ⫽
⫺.44, p ⬍ .05) and for 50% intensity sad expressions presented
for 2,000 ms (␳ ⫽ ⫺.48, p ⬍ .05). There was no significant
correlation between HAMD and response bias scores.
BDI measures correlated significantly and negatively with dis-
crimination accuracy for 50% intensity sad faces at 2,000-ms
duration (Spearman’s ␳ ⫽ ⫺.49, p ⬍ .05). BDI also correlated
negatively and significantly with response bias toward 100%
happy faces presented at 100 ms (Spearman’s ␳ ⫽ ⫺.52, p ⬍ .01).
Effect of Antidepressant Medication
To examine the possible effect of antidepressant medication on
patients’ performance, all medications were coded in terms of dose
by levels from 1 to 4, according to those proposed by Sackeim
(2001). For example, venlafaxine at a dose less than 75 mg/day
for 4 weeks or more is considered as Level 1; 75–224 mg/day,
Level 2; 225–374 mg/day, Level 3; and more than 375 mg/day
for 4 weeks, Level 4. Patients were, therefore, divided into sub-
groups with low (Levels 1 and 2) and high (Levels 3 and 4)
medication levels. Accuracy and response bias were compared
between these two subgroups of patients using independent-sam-
ple t tests. Recognition accuracy in the high-dosage subgroup was
significantly reduced in only one of the eight conditions: for sad
faces of 50% intensity presented for 100 ms, t(26) ⫽ 2.2, p ⬍ .01.
The high-dosage subgroup also had a significantly greater conser-
vative response bias toward all sad faces ( p ⬍ .05) but did not
differ from the low subgroup in the response bias for happy faces.
Thus, higher dosages of antidepressant medication in the depressed
patients were associated with a bias away from labeling sad
expressions as sad but not with a bias away from labeling happy
expressions as happy.
Discussion
Our findings indicate impaired recognition accuracy and a con-
servative response bias particularly to happy emotional facial
expressions in depressed patients. Depressed patients were not
impaired in a recognition memory test for faces (Warrington,
1984), an indirect measure of the perception of nonemotional
faces. Patients were significantly impaired in discrimination accu-
racy compared with controls for sad and, to a lesser extent, happy
expressions presented for shorter (100 ms) durations, but did not
significantly differ from controls in discrimination accuracy for
facial expressions presented for 2,000 ms. Patients also demon-
strated a significantly lower than normal tendency to label happy
faces of 50% intensity and neutral faces presented at 2,000 ms as
happy. These findings support those of previous studies demon-
strating impaired facial expression recognition (Asthana et al.,
1998; Jaeger et al., 1987; Mikhailova et al., 1996; Persad & Polivy,
1993; Rubinow & Post, 1992) and also those demonstrating ab-
normal bias in response to facial expressions (R. C. Gur et al.,
1992; Matthews & Antes, 1992; Hale, 1998; Suslow et al., 2001)
in depressed patients. These findings support our Hypothesis 1a,
but not Hypothesis 1b, suggesting that a bias away from labeling
facial expressions as happy is not associated with a bias toward
labeling expressions as sad. The finding in depressed patients of an
impaired discrimination for facial expressions presented for
shorter durations also supports Hypothesis 2.
Importantly, we found, in patients compared with healthy vol-
unteers, a discrimination impairment and an emotion-specific re-
sponse bias. These two abnormalities were demonstrated in dif-
ferent conditions, however: Discrimination accuracy in patients
was more impaired for facial expressions, particularly sad expres-
sions, presented for short durations (100 ms), whereas abnormal
response bias was demonstrated away from happy facial expres-
sions of 50% intensity presented for longer durations (2,000 ms).
Thus, response bias abnormalities were demonstrated by depressed
patients in conditions in which discrimination accuracy was not
impaired and vice versa. Happy expressions were more accurately
recognized than sad expressions in both groups, however. These
results suggest that discrimination accuracy impairments alone did
not contribute to the abnormal response bias away from mildly
happy expressions demonstrated by depressed patients. Our find-
 RECOGNITION AND RESPONSE BIAS IN DEPRESSION
ings further suggest that discrimination accuracy and response bias
can be considered separate processes underlying facial expression
recognition.
Depression has been associated with impairments in various
aspects of interpersonal functioning, for example, interpersonal
problem-solving performance (Gotlib & Asarnow, 1979), social
competence (Fisher-Beckfield & McFall, 1982), and marital inter-
actions (Gotlib & Whiffen, 1989). We suggest that abnormal
processing of emotional expressions may underlie some of those
difficulties. Furthermore, the emotion-specific bias demonstrated
by depressed patients in our study may be responsible for the
tendency of depressed patients to judge social interactions and
appraise social situations more negatively or less positively.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Depressed patients demonstrated abnormalities in discrimina-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
tion accuracy predominantly to expressions presented for shorter
durations and abnormalities in response bias to happy expressions
of milder intensity presented for longer durations, supporting Hy-
pothesis 3. We did not find significant impairments in discrimina-
tion accuracy and response bias in depressed patients to prototyp-
ical expressions presented for longer durations, nor did we dem-
onstrate between-group differences on these measures to milder
expressions presented for shorter durations. Although the former
stimuli may be relatively easy to identify, the latter stimuli may be
difficult to categorize in both normal and depressed populations.
Our findings indicate that the use of milder expressions and shorter
presentation durations allows the detection of more subtle impair-
ments in measures of recognition accuracy and response bias in
depressed patients. The impaired social function and interaction
apparent in depressed patients may, therefore, be associated with
specific abnormalities in the identification of and bias toward these
more subtle displays of facial emotion.
Further analyses revealed that depression severity was posi-
tively correlated with the magnitude of the discrimination accuracy
impairment for sad facial expressions of 50% intensity and with
the magnitude of the response bias away from labeling as happy
facial expressions of 100% intensity happiness presented for 100
ms. These findings partially support Hypothesis 4 and suggest that
the magnitude of the discrimination accuracy impairment for sad
facial expressions and magnitude of the response bias away from
labeling happy facial expressions as happy may be greater in more
severely depressed patients. It is unlikely, however, that depression
severity was the sole determinant of the magnitude of observed
abnormalities in depressed patients because patients were also
impaired in the discrimination of higher intensity sad and happy
facial expressions presented for shorter durations, and demon-
strated a reduced response bias away from labeling happy facial
expressions of milder intensity presented for longer durations.
Analysis of possible medication effects on discrimination accu-
racy and response bias revealed that patients taking higher doses of
antidepressant medication were more impaired than those taking
lower doses of medication only in the discrimination of sad faces
of 50% intensity presented for 100 ms. Patients taking higher doses
of antidepressant medication also demonstrated a tendency to
identify sad but not happy faces as neutral. The main findings in
our study of an impaired recognition accuracy in all depressed
patients, specifically for happy and sad facial expressions pre-
sented at the shorter duration, and a tendency to identify happy but
not sad faces as neutral are not, therefore, explained simply in
terms of an effect of antidepressant medication dose. It is, there-
217
fore, unlikely that attentional impairments associated with medi-
cation are responsible for our findings.
In conclusion, we have demonstrated impaired recognition ac-
curacy and response bias to emotional facial expressions in de-
pressed patients and suggest that an inability to accurately identify
and respond to the subtle changes in facial expression displayed by
others in social situations may be associated with the social dys-
function apparent in these patients. We have also demonstrated the
importance of using facial expressions of different intensities
presented for different durations in the examination of the nature
of abnormalities in facial expression recognition in clinical popu-
lations. Our findings add to the increasing number of studies
examining abnormalities in attention to and recognition of facial
expressions in clinical populations, including schizophrenia (e.g.,
Borod & Koff, 1989; Edwards, Pattison, Jackson, & Wales, 2001;
R. E. Gur et al., 2002; Kerr & Neale, 1993; Horley et al., 2001;
Whittaker, 2001) and psychopathy (Blair, Colledge, Murray, &
Mitchell, 2001) and indicate the importance of investigating the
relationship between emotion-processing abnormalities and spe-
cific symptoms of psychiatric disorders (Phillips et al., 2003).
Future studies examining depressed patients at different stages of
illness will help to clarify the extent to which these abnormalities
in facial expression identification are state or trait features of the
disorder.
References
American Psychiatric Association. (1994). Diagnostic and statistical man-
ual of mental disorders (4th ed.). Washington, DC: Author.
Asthana, H. S., Mandal, M. K., Khurana, H., & Haque-Nizamie, S. (1998).
Visuospatial and affect recognition deficit in depression. Journal of
Affective Disorders, 48, 57– 62.
Beck, A. (1976). Cognitive therapy and the emotional disorders. Madison,
CT: International Universities Press.
Beck, A. T., Ward, C., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An
inventory for measuring depression. Archives of General Psychiatry, 4,
561–571.
Blair, R. J., Colledge, E., Murray, L., & Mitchell, D. G. (2001). A selective
impairment in the processing of sad and fearful expressions in children
with psychopathic tendencies. Journal of Abnormal Child Psychol-
ogy, 29, 491– 498.
Borod, J. C., & Koff, E. (1989). The neuropsychology of emotion: Evi-
dence from normal, neurological, and psychiatric populations. In E.
Perecman (Ed.), Integrating theory and practice in clinical neuropsy-
chology (pp. 175–215). Hillsdale, NJ: Erlbaum.
Bouhuys, A. L., Geerts, E., & Gordijn, M. C. M. (1999). Depressed
patients’ perceptions of facial emotions in depressed and remitted states
are associated with relapse: A longitudinal study. Journal of Nervous
and Mental Disorder, 187, 595– 602.
Calder, A. J., Young, A. W., Rowland, D., & Perrett, D. I. (1997).
Computer-enhanced emotion in facial expressions. Proceedings of Royal
Society of London, Series B, 264, 919 –925.
Cooley, E. L., & Nowicki, S., Jr. (1989). Discrimination of facial expres-
sions of emotion by depressed subjects. Genetic, Social, & General
Psychology Monographs, 115, 449 – 465.
Corwin, J. (1994). On measuring discrimination and response bias: Un-
equal number of targets and distractors and two classes of distractors.
Neuropsychology, 8, 110 –117.
Edwards, J., Pattison, P. E., Jackson, H. J., & Wales, R. J. (2001). Facial
affect and affective prosody recognition in first-episode schizophrenia.
Schizophrenia Research, 48, 235–253.
 218 SURGULADZE ET AL.
Ekman, P., & Friesen, W. V. (1971). Constants across cultures in the face
and emotion. Journal of Personality and Social Psychology, 17, 124 –
129.
Ekman, P., & Friesen, W. V. (1976). Pictures of facial affect. Palo Alto,
CA: Consulting Psychologists Press.
Feinberg, T. E., Rifkin, A., Schaffer, C., & Walker, E. (1986). Facial
discrimination and emotional recognition in schizophrenia and affective
disorders. Archives of General Psychiatry, 43, 276 –279.
Fisher-Beckfield, D., & McFall, R. M. (1982). Development of a compe-
tence inventory for college men and evaluation of relationships between
competence and depression. Journal of Consulting and Clinical Psy-
chology, 50, 697–705.
Folstein, M. F., Folstein, S. E., & McHugh P. R. (1975). “Mini-Mental
State”: A practical method for grading the cognitive state of patients for
the clinician. Journal of Psychiatric Research, 12, 189 –198.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Gotlib, I. H., & Asarnow, R. F. (1979). Interpersonal and impersonal
This document is copyrighted by the American Psychological Association or one of its allied publishers.
problem-solving skills in mildly and clinically depressed university
students. Journal of Consulting and Clinical Psychology, 47, 86 –95.
Gotlib, I. H., & Whiffen, V. E. (1989). Depression and marital functioning:
An examination of specificity and gender differences. Journal of Abnor-
mal Psychology, 98, 23–30.
Gur, R. C., Erwin, R. J., Gur, R. E., Zwil, A. S., Heimberg, C., & Kraemer,
H. C. (1992). Facial emotion discrimination: II. Behavioral findings in
depression. Psychiatry Research, 42, 241–251.
Gur, R. E., McGrath, C., Chan, R. M., Schroeder, L., Turner, T., Turetsky,
B. I., et al. (2002). An fMRI study of facial emotion processing in
patients with schizophrenia. American Journal of Psychiatry, 159, 1992–
1999.
Hale, W. W. (1998). Judgment of facial expressions and depression per-
sistence. Psychiatry Research, 80, 265–274.
Hamilton, M. (1960). A rating scale for depression. Journal of Neurology,
Neurosurgery and Psychiatry, 23, 56 – 62.
Horley, K., Gonsalvez, C., Williams, L., Lazzaro, I., Bahramali, H., &
Gordon, E. (2001). Event-related potentials to threat-related faces in
schizophrenia. International Journal of Neuroscience, 107, 113–130.
Jaeger, J., Borod, J., & Peselow, E. (1987). Depressed patients have
atypical hemispace biases in the perception of emotional chimeric faces.
Journal of Abnormal Psychology, 96, 321–324.
Junghoefer, M., Bradley, M. M., Elbert, T. R., & Lang, P. J. (2001).
Fleeting images: A new look at early emotion discrimination. Psycho-
physiology, 38, 175–178.
Kerr, S. L., & Neale, J. M. (1993). Emotion perception in schizophrenia:
Specific deficit or further evidence of generalized poor performance?
Journal of Abnormal Psychology, 102, 312–318.
Matthews, G. R., & Antes, J. R. (1992). Visual attention and depression:
Cognitive biases in the eye fixation of the dysphoric and non-depressed.
Cognitive Therapy and Research, 16, 359 –371.
Mikhailova, E. S., Vladimirova, T. V., Iznak, A. F., Tsusulkovskaya, E. J.,
& Sushko, N. V. (1996). Abnormal recognition of facial expression of
emotions in depressed patients with major depression disorder and
schizotypal personality disorder. Biological Psychiatry, 40, 697–705.
Murphy, F. C., Sahakian, B. J., Rubinsztein, J. S., Michael, A., Rogers,
R. D., Robbins, T. W., & Paykel, E. S. (1999). Emotional bias and
inhibitory control processes in mania and depression. Psychological
Medicine, 29, 1307–1321.
Oldfield, R. C. (1971). The assessment and analysis of handedness: The
Edinburgh inventory. Neuropsychologia, 9, 97–113.
Persad, S., & Polivy, J. (1993). Differences between depressed and non-
depressed individuals in the recognition of and response to facial cues.
Journal of Abnormal Psychology, 102, 358 –368.
Phillips, M. L., Drevets, W. C., Rauch, S. L., & Lane, R. D (2003). The
neurobiology of emotion perception: II. Implications for understanding
the neural basis of emotion perceptual abnormalities in schizophrenia
and affective disorders. Biological Psychiatry, 54, 515–528.
Rubinow, D. R., & Post, R. M. (1992). Impaired recognition of affect in
facial expression in depressed patients. Biological Psychiatry, 31, 947–
953.
Sackeim, H. A. (2001). The definition and meaning of treatment-resistant
depression. Journal of Clinical Psychiatry, 62(Suppl. 16), 1–17.
Sato, W., Kochiyama, T., Yoshikawa, S., & Matsumura, M. (2001). Emo-
tional expression boosts early visual processing of the face: ERP record-
ing and its decomposition by independent component analysis. Neuro-
Report, 12, 709 –714.
Sloan, D. M., Strauss, M. E., Quirk, S. W., & Sajatovic, M. (1997).
Subjective and expressive emotional responses in depression. Journal of
Affective Disorders, 46, 135–141.
Sloan, D. M., Strauss, M. E., & Wisner, K. L. (2001). Diminished response
to pleasant stimuli by depressed women. Journal of Abnormal Psychol-
ogy, 110, 488 – 493.
Stanislaw, H., & Todorov, N. (1999). Calculation of signal detection theory
measures. Behaviour Research Methods, Instruments, & Computers, 31,
137–149.
Suslow, T., Junghanns, K., & Arolt, V. (2001). Detection of facial expres-
sions of emotions in depression. Perceptual and Motor Skills, 92,
857– 868.
Warrington, E. K. (1984). Recognition Memory Test. Windsor, England:
NFER-NELSON.
Wexler, B. E., Levenson, L., Warrenburg, S., & Price, L. H. (1994).
Decreased perceptual sensitivity to emotion-evoking stimuli in depres-
sion. Psychiatry Research, 51, 127–138.
Whittaker, J. F. (2001). Face processing in schizophrenia: defining the
deficit. Psychological Medicine, 31, 499 –507.
Young, A. W., Perrett, D. I., Calder, A. J., Sprengelmeyer, R., & Ekman,
P. (2002). Facial expressions of emotion: Stimuli and tests (FEEST)
[Computer software]. Bury St. Edmunds, England: Thames Valley Test
Company.
Received September 20, 2002
Revision received April 25, 2003
Accepted May 5, 2003 䡲