Is facial emotion recognition impairment in schizophrenia.pdf

Available online at www.sciencedirect.com

Schizophrenia Research 99 (2008) 263 – 269

www.elsevier.com/locate/schres

Is facial emotion recognition impairment in schizophrenia identical

for different emotions? A signal detection analysis

Daniel T. Tsoi ⁎ , Kwang-Hyuk Lee, Waqqas A. Khokhar, Nusrat U. Mir,

Jaspal S. Swalli, Kate A. Gee, Graham Pluck, Peter W.R. Woodruff

Sheffield Cognition And Neuroimaging Laboratory (SCANLab), Academic Clinical Psychiatry, Section of Neuroscience, School of Medicine and

Biomedical Sciences, University of Sheffield, Longley Centre, Norwood Grange Drive, Sheffield S5 7JT, United Kingdom

Received 23 May 2007; received in revised form 30 October 2007; accepted 2 November 2007

Available online 3 January 2008

Abstract

Patients with schizophrenia have difficulty recognising the emotion that corresponds to a given facial expression. According to

signal detection theory, two separate processes are involved in facial emotion perception: a sensory process (measured by

sensitivity which is the ability to distinguish one facial emotion from another facial emotion) and a cognitive decision process

(measured by response criterion which is the tendency to judge a facial emotion as a particular emotion). It is uncertain whether

facial emotion recognition deficits in schizophrenia are primarily due to impaired sensitivity or response bias. In this study, we

hypothesised that individuals with schizophrenia would have both diminished sensitivity and different response criteria in facial

emotion recognition across different emotions compared with healthy controls. Twenty-five individuals with a DSM-IV diagnosis

of schizophrenia were compared with age and IQ matched healthy controls. Participants performed a “ yes-no ” task by indicating

whether the 88 Ekman faces shown briefly expressed one of the target emotions in three randomly ordered runs (happy, sad and

fear). Sensitivity and response criteria for facial emotion recognition was calculated as d-prime and In( β ) respectively using signal

detection theory. Patients with schizophrenia showed diminished sensitivity (d-prime) in recognising happy faces, but not faces that

expressed fear or sadness. By contrast, patients exhibited a significantly less strict response criteria (In(

Keywords: Facial emotion recognition; Schizophrenia; Social cognition; Signal detection analysis

1. Introduction

( Feinberg et al., 1986; Archer et al., 1992; Salem et al.,

1996; Addington and Addington, 1998; Kohler et al.,

2003 ). This facial emotion recognition impairment has

been shown to adversely affect the interpersonal and

social functioning of people with schizophrenia ( Mueser

et al., 1996; Hooker and Park, 2002; Kee et al., 2003;

Brekke et al., 2005; Addington et al., 2006 ).

Furthermore, patients with schizophrenia are worse

at recognising negative, as opposed to positive, facial

emotions ( Kohler et al., 2003; Bediou et al., 2005; van't

Substantial evidence supports the suggestion that

individuals with schizophrenia have difficulties recog-

nising the correct emotion of a given facial expression

⁎ Corresponding author. Academic Clinical Psychiatry, Longley

Centre, Norwood Grange Drive, Sheffield S5 7JT, United Kingdom.

Tel.: +44 114 2261512; fax: +44 114 2261522.

E-mail address: t.tsoi@sheffield.ac.uk (D.T. Tsoi).

doi: 10.1016/j.schres.2007.11.006

)) in recognising fearful

and sad faces. Our results suggest that patients with schizophrenia have a specific deficit in recognising happy faces, whereas they

were more inclined to attribute any facial emotion as fearful or sad.

264

D.T. Tsoi et al. / Schizophrenia Research 99 (2008) 263 – 269

Wout et al., 2007 ). This finding raises a question whether

positive and negative facial emotion recognition engage

different processes. Several hypotheses have been put

forward to explain the difference in performance between

positive and negative facial emotion recognition in

schizophrenia. According to the social-cognitive hypoth-

esis, patients avoid stimuli that induce negative emotion

( Walker et al., 1980 ). Phillips et al. (1999) and Gur et al.

(2002) have additionally proposed that amygdala activa-

tion specific to negative facial emotion is reduced in

schizophrenia.

In addition to those described, there may be other

explanations for the differences in positive and negative

facial emotion recognition in schizophrenia. Johnston

et al. (2003, 2006) argued that patients with schizo-

phrenia perform poorly at recognising negative facial

emotions because positive facial emotions are generally

easier to recognise than negative emotions (and so less

likely to be adversely affected by the illness). Such

differences in discriminability are not so obvious in

healthy subjects because they usually score at or near the

maximum possible performance level in the standard

facial emotion recognition tests for all emotions (

)) corresponds to the cognitive decision process and

reflects the tendency of individuals to make a certain

decision with the evidence they have received from the

sensory process.

In this study, we applied signal detection theory to

investigate the underlying psychological processes of

facial affect recognition for different emotions in

schizophrenia. As a preliminary exploratory study, we

focused on positive emotion (happy), negative non-

threatening emotion (sad) and negative threatening

emotion (fear). These emotions have been shown to be

significantly disturbed in schizophrenia in the literature

( Kohler et al., 2003; Bediou et al., 2005; van't Wout

et al., 2007 ). We examined whether patients with

schizophrenia differed from controls in both the sensory

and the cognitive decision processes for facial emotion

recognition across positive and negative emotions. The

relationship between symptoms and facial emotion

recognition was also evaluated. Similar analyses have

been used to study facial emotion recognition in healthy

volunteers ( Goos and Silverman, 2002; Grimshaw et al.,

2004 ). Schneider et al. (2006) has also recently examined

the specific psychometric characteristics of facial emo-

tion recognition in schizophrenia, with a stimulus

exposure time of three seconds. In this study, we

deliberately used a very short stimulus exposure time to

reduce the overall accuracy in the controls in order to

minimise the possible

“

ceiling

). This argument is strengthened by the fact that

previous studies of facial emotion recognition in

schizophrenia used relatively long stimulus exposure

times that ranged from 500 ms ( Edwards et al., 2001 )to

15 s ( Addington et al., 2006 ). One potentially useful

method to avoid the ceiling effect in healthy control

subjects is to manipulate stimulus presentation duration

( Kirouac and Dore, 1984; Ogawa and Suzuki, 1999 ). For

example, Grimshow et al. (2004) presented face stimuli

for 30 ms and they were able to observe sex differences in

emotion recognition. The use of such rapid visual

presentation would increase task difficulty, during

which subjects might be in a high degree of uncertainty

and thus more liable to make errors.

The process of facial emotion recognition can be

conceptualised using psychophysical methods such as

signal detection theory. This theory has been widely

applied to the study of perception. Signal detection

theory attempts to explain how individuals make

decisions based on the evidence they have received.

Most decisions a person makes contains a degree of

uncertainty ( McNicol, 1972 ). There are two broad

psychological processes involved in decision making:

the sensory process and the cognitive decision process

( Krantz, 1969 ). The sensory process refers to the

transformation of physical stimuli into internal percep-

tion. The cognitive decision process involves deciding

how to respond based on the output of the sensory

process. Signal detection theory provides separate

”

and to introduce a

relatively high degree of uncertainty. To our knowledge,

the current study is the first that employs signal detection

theory to investigate facial emotion recognition in

patients with schizophrenia. We hypothesised that

individuals with schizophrenia, compared with healthy

controls, would have diminished sensitivity and different

response criteria in facial emotion recognition across

different emotions.

“

ceiling effect

”

2. Method

2.1. Subjects

Twenty-five patients with a DSM-IV diagnosis of

schizophrenia ( American Psychiatric Association, 1994 )

were recruited from in-patient wards (17 patients) and from

the community (8 patients). Twenty-five healthy volun-

teers were recruited from the community as controls. The

demographics of the subjects are summarised in Tabl e 1 .

measures of performance in decision making to reflect

these two processes. The sensory process is measured by

sensitivity (d-prime), which determines how well the

observer is able to select the correct stimuli while

avoiding the incorrect ones. The response criterion (In

(

effect

D.T. Tsoi et al. / Schizophrenia Research 99 (2008) 263 – 269

265

Table 1

Demographics and clinical characteristics of patients and controls (SD in parentheses)

Schizophrenia patients (n=25)

Healthy controls (n=25)

p value

Mean age (years)

37.8 (9.5, range 20 to 54)

40.5 (9.0, range 21 to 55)

0.31

Sex (male/female)

21/4

17/8

0.19

Mean years of education

12.5 (2.4)

11.9 (2.3)

0.40

Mean IQ by NART

105 (13)

106 (8)

0.65

Mean SAPS total score

25 (15)

Mean SANS total score

11 (9)

Mean CDSS total score

4 (4)

The two groups did not differ significantly in age, male/

female ratio, number of years of education received and IQ

estimated with the National Adult Reading Test [NART]

( Nelson and Willison, 1991 ). Exclusion criteria for both

groups included the presence of a history of neurological

disorders or learning disability, or a current diagnosis of

alcohol or drug dependence. Both patients and controls

gave written informed consent before their participation.

This study was approved by the local Research Ethics

Committee.

All patients were judged as clinically stable by their

psychiatrists at the time of assessment. All patients

except one were on antipsychotic medication. The mean

daily dose in chlorpromazine equivalence was 408.6 mg

(SD=253.0 mg). Twenty-one patients were on atypical

antipsychotic medications and the other three patients

were on typical antipsychotics. The mean duration of

illness was 13.5 years (SD=9.3 years, range from

6 months to 35 years).

press a button on a mouse connected to the computer as

quickly as possible to indicate whether the face shown

was displaying the target emotion. The participants were

asked to press a different button to indicate that the face

shown was not expressing the target emotion. There

were 88 faces shown for each block and participants

were asked to identify a specific target emotion in each

block. Therefore, there were three different blocks for

three target emotions and the sequence of these blocks

was randomised. Each stimulus was shown for 50 ms,

and then followed by a 1300 ms central-fixation cross,

during which a response was to be made. There were 28

faces in each block with the target emotion (7 faces from

2 males and 2 females). The other 60 faces with non-

target emotions included the other five emotions from

the 2 males and 2 females (3 faces for each person

expressing one non-target emotion). Participants were

given a five-minute break between each block.

Before the experimental procedure, subjects viewed

sample photographs from the POFA (including both male

and female but these photographs were different from

those used in the actual experiment) exhibiting the six

different emotions. All participants had 12 practice trials

for each target emotion, using the sample photographs.

After the facial emotion recognition task, the patient

group were assessed with: (1) the Scale for the

Assessment of Positive Symptoms (SAPS) ( Andreasen,

1984 ) and Negative Symptoms (SANS) ( Andreasen,

1983 ) for symptom severity; (2) the Calgary Depression

Scale for Schizophrenia (CDSS) ( Addington et al.,

1990 ) for depressive symptoms.

2.2. Experimental procedure and task

task in a quiet well-

lit room. The stimuli of the task consisted of twenty-four

black and white photographs from the Pictures of Facial

Affect (POFA) ( Ekman and Friesen, 1976 ). These

photographs feature two males and two females exhibit-

ing each of the following six facial expressions:

happiness, sadness, fear, disgust, surprise and anger.

Hair features and clothing were occluded from all these

faces ( Phillips et al., 1999 ). These stimuli were presented

with the Presentation software package (Neurobehavioral

Systems Inc., San Francisco) on a laptop computer (screen

size: 28.8 cm×21.3 cm, stimuli size: 14.4 cm×21.3 cm at

the centre of screen). Participants sat approximately 60 cm

from the computer screen.

Participants were instructed to decide whether the

faces shown on the screen expressed one of three target

emotions: happiness, sadness or fear. They were told to

“

yes-no

2.3. Statistical analysis

)

respectively, according to signal detection theory. The

calculation of d-prime and In(

) was based on the

formulareportedinthepaperby Macmillan and

Creelman (1990) . Adjusted Hit Rate (HR) and False

Alarm Rate (FAR) according to Corwin (1994) were

On the day of testing, all participants were interviewed

to obtain demographic information. They then performed

a facial emotion recognition

”

Sensitivity and response criterion for facial emotion

recognition were calculated as d-prime and In(

266

D.T. Tsoi et al. / Schizophrenia Research 99 (2008) 263 – 269

). The sensitivity measure d-prime refers to the

sensory dimension of facial emotion recognition, i.e. the

ability to distinguish target facial emotion from non-

target emotions. A larger d-prime means a better ability

to differentiate the target emotion from other emotions.

The response criterion In(

further analysed using signal detection theory in order to

take account of the possibility that subjects may be more

inclined to indicate that every face was displaying the

target emotion and hence could lead to higher hit rate for

at the expense of false alarms. Furthermore, as indicated

by the significant interaction effect (group by emotion)

[F(2,96) =4.15, p=0.02, power=0.72] in false alarm

rate, any such bias produced could induce differential

effects on different facial emotions in the patient and

control groups.

) reflects the cognitive

dimension of facial emotion recognition and it measures

the tendency for a subject to judge a facial emotion as a

specific emotion. If the In(

) is positive, the subject has

adopted a strict criterion and is biased towards judging that

any facial emotion is a non-target emotion. The more

positive the In(

3.1. Sensitivity for different emotions (d-prime)

), the stricter criterion the subject has

adopted. On the other hand, a negative In(

There was a significant interaction effect (group by

emotion) [F(2, 96)=4.08, p=0.02, power=0.71].

( Table 2 ). The interaction was explained by the fact that

d-prime differences between the patient and control group

were statistically significant for happy faces [F(1, 48) =

7.52, p

) indicates that

a subject has adopted a lax criterion and is biased towards

judging that any facial emotion is the target emotion. A

zero In(

) would suggest an unbiased judgement.

D-prime, In(

) and reaction time (RT) were exam-

ined as dependent variables and these variables were

entered into a 2×3 multivariate analysis of variance

(MANOVA), with a between-subject variable of group

(schizophrenia or controls) and within-subject variables

of emotion (fear, happiness or sadness). Pearson's

correlation analysis was used to examine associations

between dependent variables and clinical characteristics.

Statistical significance was set at 0.05 and all statistical

tests were two-tailed.

0.01, power=0.77] but not for faces

showing sadness [F(1, 48) =3.11, p=0.08, power=0.41]

and fear [F(1, 48) =0.76, p=0.39,power=0.14].Themain

effect of emotion on d-prime [F(2, 96) =112.64, p

0.001,

power=1.00] was significant, suggesting that d-prime was

highest for happy faces and lowest for faces showing fear in

both groups. There was also a significant main effect of

group [F(1, 48) =5.22, p=0.03, power=0.61] and this

indicates patients exhibited a lower d-prime compared with

controls across different emotions.

3. Results

3.2. Response criteria for different emotions (In(

))

Patients with schizophrenia showed a significant

lower Hit Rate (HR) in recognising happy faces [F

(1,48) =6.55, p=0.01, power=0.71], but not in recog-

nising faces with fear or sadness ( Table 2 ). False Alarm

Rate (FAR) was higher in the patient group for all three

emotions [fear: F(1,48) =5.03, p=0.03, power=0.59;

happy: F(1,48) =5.42, p=0.02, power=0.63; sad: F

(1,48) =10.07, p

0.01, power=0.85]

( Table 2 ). The interaction analysis showed that there were

significant between-group differences in recognising faces

with fear [F(1, 48) =4.67, p=0.04, power=0.56] and

sadness [F(1, 48) =12.37, p

)[F(2,96) =5.58, p

0.01, power=0.93] but not

for happy faces [F(1, 48) =2.23, p=0.14, power=0.31].

Within each group, there was significant difference

0.01, power=0.88]. These data were

Table 2

Performance indicators (mean values with SD in parentheses) of facial emotion recognition tasks for three target emotions

Hit rate (unadjusted) False alarm rate (unadjusted) D-prime

In(

)

Reaction time (milliseconds)

Fear

Schizophrenia (n=25) 0.61 (0.19)

0.28 (0.18)

0.94 (0.51) 0.18 (0.50) 653.31 (145.13)

Controls (n=25)

0.54 (0.16)

0.18 (0.12)

1.08 (0.55) 0.47 (0.44) 574.49 (85.74)

Happy

Schizophrenia (n=25) 0.75 (0.17)

0.12 (0.12)

2.07 (0.95) 0.52 (0.99) 608.37 (137.48)

Controls (n=25)

0.85 (0.12)

0.06 (0.08)

2.76 (0.81) 0.87 (0.64) 532.60 (81.19)

Sad

Schizophrenia (n=25) 0.65 (0.28)

0.31 (0.31)

1.18 (0.74) 0.26 (0.99) 621.05 (134.92)

Controls (n=25)

0.52 (0.18)

0.10 (0.14)

1.54 (0.70) 1.20 (0.90) 546.67 (85.89)

used in the calculation to avoid problems of division by

zero or result in infinite values of the calculated d-prime

and In(

There was a significant interaction effect (group by

emotion) on In(

D.T. Tsoi et al. / Schizophrenia Research 99 (2008) 263 – 269

267

) values for faces with fear and happiness

(p=0.02 for patients; p

4. Discussion

0.01 for controls). The difference

) values for faces with fear and sadness

was significant in the control group (p

Using signal detection theory, we found that

individuals with schizophrenia, when compared with

controls, demonstrated a diminished sensitivity (d-

prime) in recognising happy faces, although the main

effect of emotion (highest d-prime for happy faces

recognition) suggested that happy faces were more

easily recognised than other facial emotions. On the

other hand, individuals with schizophrenia adopted a

significantly less strict response criteria (In(

0.001) but not in

the patient group (p=0.58). The difference between the In

(

) values for faces with happiness and sadness was also

only significant in the control group (p=0.05) but not in

the patient group (p=0.12). The main effect of group [F

(1,48)=8.42, p

0.01, power=0.81] indicated that value

) was lower in the patient group than the controls

across all three target emotions. The main effect of

emotion [F(2,96)=8.83, p

)) than the

controls in recognising fear and sad facial emotions.

This suggests that individuals with schizophrenia, when

compared to the controls, were more inclined to attribute

any facial emotion as fearful or sad.

We used a short stimulus exposure time (50 ms) in

our experimental paradigm to avoid the potential

“

0.001, power=0.97] was

also statistically significant. This suggests that the In(

)

value was lowest (i.e. less positive) for recognising faces

displaying fear in both groups.

3.3. Reaction time for different emotions

in other facial emotion recognition

tasks. The highest mean hit rate was 0.85 (controls in

recognising happy faces) and it was well below the

possible maximum hit rate of 1.0. There is substantial

evidence suggesting that perception of facial expres-

sions can occur automatically ( Hansen and Hansen,

1994; Stenberg et al., 1998 ) and without conscious

awareness ( Morris et al., 1998 ) in healthy volunteers.

When a negative facial emotion was presented for 20 ms

as prime before a neutral face, individuals with

schizophrenia judged that neural facial emotion as

more unpleasant than controls ( Hoschel and Irle, 2001 ).

Taken together, the results of these studies suggest that

patients with schizophrenia can perceive facial expres-

sions even at very short stimulus exposure times.

A facial affect recognition deficit in schizophrenia may

reflect a specific facial affect processing deficit ( Borod

et al., 1993 ). Alternatively, it mayalsobesecondarytoa

generalised perceptual impairment ( Archer et al., 1992 ).

Previous studies reported that normal individuals had

greatest accuracy in recognising happy faces from a range

of facial emotion expressions ( Ekman et al., 1972; Johnston

et al., 2003 ). Hence, the differences in performance of the

task between patient and control group in recognising facial

emotion should be smallest for happy faces ( Chapman and

Chapman, 1978 ). Contrary to what Chapman and Chap-

man (1978) suggested, we found that recognition of happy

faces was more impaired than recognition of sad or fearful

faces in our patient sample. Our results suggest that the

facial emotion recognition deficit in schizophrenia is more

complex than previously thought. We have shown that

there are at least two potentially different recognition

processes involved in the observed deficit, namely

impaired sensitivity and response bias. The underlying

process of this recognition deficit is different for each

”

The reaction time was significantly delayed in the

patient group compared to the control group across all

three emotions as shown by the significant main effect

of group [F(1, 48)=7.19, p=0.01, power=0.75]

( Table 2 ). However, there was no significant interaction

effect [F(2, 96)=0.01, p=0.98, power=0.05]. The

main effect of emotion on the reaction time was

significant [F(2,96) =5.28, p

0.01) and

for faces displaying sadness (p=0.01). There was no

statistically significant difference in the reaction time of

recognising happy and sad faces (p=0.34).

3.4. Relationship between facial emotion recognition

parameters and clinical characteristics in the patient group

There was a significant positive correlation between

the SAPS delusion total score and d-prime in task of

recognising both fear (r=0.40, p=0.05) as well as sad

faces (r=0.48, p=0.02). D-prime (fear) had also a

significant positive correlation with the CDSS total

score (r=0.42, p=0.04). In(

0.01). NART IQ,

duration of illness and dosage of antipsychotic medica-

tion (in chlorpromazine equivalence) did not have

significant correlations with either d-prime or In(

)in

each emotion. There were no significant correlations

between scores of the clinical measures and In(

) for

the emotions of fear and sadness. There were also no

significant correlations between d-prime (happy) and

clinical ratings.

between the In(

of In(

ceiling effect

0.01, power=0.83]. The

reaction time for recognising faces with fear was

significantly delayed in both groups, compared to the

reaction time for recognising happy faces (p

) in task recognising

happy faces had a significant positive correlation with

the SANS total score (r=0.56, p

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: