Symptom severity of panic disorder associated with impairment in emotion processing of threat-related facial expressions


Correspondence: Kyoung-Uk Lee, MD, PhD, Department of Psychiatry, Uijeongbu St Mary's Hospital, Catholic University of Korea College of Medicine, 65-1 Geumo-dong, Uijeongbu 480-130, Korea. Email:



To compare emotion recognition patterns between patients with panic disorder (PD) and healthy volunteers and to analyze the correlation between the degree of emotion recognition impairment and symptom severity in patients with PD.


Twenty-four patients with PD and 20 healthy controls were tested with a facial emotional expression recognition task involving four basic emotions (i.e. happiness, sadness, anger, and fear). Emotion recognition measures included the recognition threshold, response time, response time of correctly classified emotions (response time_crt), and recognition error. An average of all four emotions for each emotion recognition measure was compared between the two groups and then a comparison of recognition measures for each specific emotion was conducted. The correlations between severity of the State–Trait Anxiety Inventory, Beck Depression Inventory (BDI), and Panic Disorder Severity Scale with emotion recognition indices were also analyzed.


Average recognition threshold was significantly higher in the PD group compared to the control group. In the PD group, there was a non-significant trend of increase in the emotion recognition threshold for fear and the response time for anger compared with the control group. In the correlation analysis, higher trait anxiety was associated with slower response time_crt for anger and a higher BDI score was associated with slower response times and response time_crt for happiness and anger.


This study suggests that symptom severity of PD might be associated with impairment in emotion processing of threat-related facial expressions.

FACIAL EXPRESSIONS ARE one of the most important ways to communicate emotion in social interactions.[1] The ability to accurately recognize others' emotions via facial expressions is a critical social communication skill.[2] Failure to perceive people's facial expressions accurately may have significant adverse consequences.[3] Deficits in interpreting facial expressions of emotion have been demonstrated in many psychiatric disorders including social phobia,[4] generalized anxiety disorder,[5] depressive disorder,[6, 7] and schizophrenia.[8, 9]

Panic disorder (PD) is characterized by sudden and unexpected attacks of fear accompanied by a feeling of severe anxiety and a major change in physical sensations.[10, 11] The cognitive–psychological model emphasizes the importance of biased information processing in explaining both the etiology and maintenance of PD.[12] This model further suggests that the manner in which threat-relevant information is perceived, encoded, and recalled plays an important role. Studies have also indicated that patients with PD tend to direct their attention towards threat stimuli and deduce ambiguous stimuli in a negative way.[13, 14] Accordingly, it is generally assumed that patients with PD may have abnormal patterns and biases in the recognition of facial emotional expressions.

Previous studies have reported that patients with PD have disturbances in their ability to recognize facial emotional expressions compared to healthy individuals,[15] but it is still unclear whether these disturbances are associated with an enhanced or impaired ability to recognize one or more certain emotions. Reinecke et al. have suggested that patients with PD have an enhanced perception of negative (sadness) facial expressions.[16] Similarly, a meta-analysis has reported that patients with anxiety traits or anxiety disorders including PD tend to misclassify neutral expressions as anger expressions.[17] In contrast, Kessler et al. have reported that emotion recognition in patients with PD is generally impaired, especially in recognizing the emotions of sadness and anger.[18] Thus, more studies are needed in order to clarify this controversy. More importantly, studies investigating correlations between symptom severity and the degree of emotion recognition impairment in patients with PD are very scarce. Although Reinecke et al. reported that a higher score on the Panic Disorder Severity Scale (PDSS) is related to higher attentional avoidance of happy faces in patients with PD, 39% (9/23) of patients included in the PD group had other comorbid anxiety or mood disorders.[16] Therefore, the results may have been driven by these comorbidities.

The aims of this study were therefore (i) to compare the emotion recognition patterns between PD patients and healthy volunteers; and (ii) to investigate whether panic symptom severity is correlated with the degree of emotion recognition impairment in patients with PD. We hypothesized that patients with PD would have a bias in emotion recognition of facial expression compared to the control group. We postulated that this dysfunction would be more prominent for negative emotions, especially for threat-related emotions. Finally, we predicted that this observed emotion recognition dysfunction for threat-related emotions would be correlated with panic symptom severity.



Twenty-four patients with PD and 20 healthy control volunteers between 18 and 65 years of age participated in this study. Patients were recruited from the outpatient psychiatry clinic at Catholic University of Korea, Uijeongbu St Mary's Hospital. The patient group was carefully assessed by a psychiatric specialist to confirm the diagnosis of PD with or without agoraphobia. A comprehensive psychiatric interview was completed to rule out patients with psychotic symptoms, cognitive deficits, and neurological conditions. For the control group, participants were screened using the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) semi-structured interview to exclude those with a current or previous history of psychiatric disorders. Both groups completed the Korean version of the State–Trait Anxiety Inventory (STAI) and the Beck Depression Inventory (BDI) to assess anxiety and depressive symptoms. Patients with PD also completed the PDSS. All participants gave their written informed consent. This study was approved by the Institutional Review Board of Catholic University of Korea, Uijeongbu St Mary's Hospital.


The task assessing the ability to recognize emotion in facial expressions addressed four basic emotions (i.e. happiness, sadness, anger, and fear) taken from the ChaeLee Korean Facial Expressions of Emotion.[19] For each of these emotions, morphing software was used to illustrate facial expressions displaying 10% gradations of emotional intensity ranging from 0% emotional intensity (neutral) to 100% emotional intensity. The task consisted of 40 presentations for each emotion, four examples for each gradation of intensity (two male and two female), and 10 presentations of neutral expressions, summing to a total of 170 facial presentations (4 emotions × 10 intensities × 4 examples + 10 neutral expressions = 170 stimuli). Each face was presented for 500 ms in random order and was immediately followed by a blank screen. Stimuli were presented using E-Prime (Psychology Software Tools, Pittsburgh, PA, USA). Participants were instructed to categorize the emotional expression of the face presented into one of the four emotions or neutral by pressing a corresponding key. Participants were told to respond as quickly and as accurately as possible. Accuracy, response speed (response time) and response speed of correct responses (response time_crt) were analyzed. In addition, the recognition threshold was also analyzed based on methods utilized by Harmer et al.[20] Recognition threshold represents the level of intensity required for successful recognition of the emotion. In the present study the recognition threshold was defined as the level of emotional intensity at which participants correctly labeled ≥75% of the emotion of facial expressions for four consecutive presentations. All participants completed the procedure at a similar time of day (late morning to early afternoon), and at least 10 h had passed between the experiment and the last intake of medication.

Statistical analysis

All statistical analysis was performed using SAS/PC version 9.2 (SAS Institute, Cary, NC, USA). Differences in emotional recognition measure scores between the two groups were analyzed using Student's t-test. Correlations between severity of psychopathology-related measures (BDI, PDSS, and STAI) and emotion recognition indices among patients with PD and healthy controls were analyzed using Pearson's correlation. The significance level was set at P < 0.05 (two-tailed).


Participant characteristics

There were no significant differences between the two groups in gender, religion, occupation status, and mean age. The mean State Anxiety Inventory (SAI) score of the PD group was higher than the control group, but this difference was not statistically significant. Mean Trait Anxiety Inventory (TAI) and BDI scores were significantly higher in the PD group than in the control group. All patients in the PD group had been taking medication for at least 1 month before participating in this study. Eight patients were prescribed escitalopram (5–20 mg/day), 11 patients were prescribed paroxetine (10–50 mg/day), and five were prescribed other medications (Table 1).

Table 1. Demographic characteristics

PD group

(n = 24)

Control group

(n = 20)

  1. aAlprazolam, fluoxetine, imipramine, mirtazapine, venlafaxine (n = 1 for each drug). BDI, Beck Depression Inventory; PD, panic disorder; PDSS, Panic Disorder Severity Scale; SAI: State Anxiety Inventory; TAI, Trait Anxiety Inventory.
Gender (male), (%)50 40 0.652
Religion status (yes) (%)29 25 0.757
Occupation status (yes) (%)29 50 0.158
Age (years)47.299.1046.8010.430.868
Disease duration (years)4.154.41   
Duration of medication (years)3.414.17   
Medication type     
Escitalopram, n = 8 (mg/day)13.756.41   
Paroxetine, n = 11 (mg/day)19.5511.83   
Others (n = 5)a     

Emotion recognition patterns

Table 2 shows that the average recognition threshold was significantly higher in the PD group compared to the control group, but that the two groups did not differ on average response time, average response time_crt, and average recognition error. Comparison analysis for each specific emotion between the two groups showed that there was no group difference across all four recognition measures for all four emotions. There was, however, a non-significant trend of increase in recognition threshold for fear and a non-significant trend of increase in response time for anger in the PD group compared to the control group.

Table 2. Facial recognition task: emotion recognition measurements

PD group

(n = 24)

Control group

(n = 20)

  1. Student's t-test; average of all four emotions. PD, panic disorder; response time_crt, response time for correctly recognized emotions.
Average recognition threshold51.049.8345.387.530.041
Average response time (ms)1514.08753.791254.60410.020.156
Average response time_crt (ms)1559.53786.471249.87426.460.245
Average recognition error 4.882.854.092.590.348
Recognition threshold (each emotion)
Response time (each emotion, ms)
Response time_crt (each emotion, ms)
Recognition error (each emotion)

Correlation analysis

In both groups, there were no significant correlations between measures of symptom severity with recognition threshold and with recognition errors across all four emotional expressions. In the PD group, TAI severity was positively correlated with response time_crt for anger. BDI severity was positively correlated with both response time and response time_crt for happiness and anger. In the control group, TAI severity was positively correlated with both response time and response time_crt for sadness, and BDI severity was positively correlated with response time for fear (Table 3).

Table 3. Correlation between psychopathology-related measures and emotion recognitiona
  1. aPearson correlation coefficients. BDI, Beck Depression Inventory; PD, panic disorder; PDSS, Panic Disorder Severity Scale; response time_crt, response time for correctly recognized emotions; SAI, State Anxiety Inventory; TAI, Trait Anxiety Inventory.
Recognition threshold        
PD group        
Control group        
Happiness0.2892170.0001.000  −0.1480.533
Sadness0.0230.9250.0830.728  0.1490.531
Anger0.3030.1940.1790.450  0.1230.607
Fear0.2700.250−0.0160.946  0.0080.972
Response time        
PD group        
Control group        
Happiness0.2790.2340.3650.114  0.2930.210
Sadness0.3470.1340.4650.039  0.2430.302
Anger0.3040.1920.3900.089  0.3220.167
Fear0.3950.0850.3780.101  0.4780.033
Response time_crt        
PD group        
Control group        
Happiness0.3890.0900.4080.074  0.3950.085
Sadness0.3610.1170.4570.043  0.2810.230
Anger0.1820.4410.2560.276  0.2700.250
Fear0.3620.1160.2800.231  0.2780.236
Recognition error        
PD group        
Control group        
Happiness0.0940.6950.0980.681  −0.3160.174
Sadness−0.0210.9290.1070.652  −0.0860.719
Anger0.0910.7030.1710.471  0.3260.160
Fear−0.2870.220−0.1740.464  0.0020.993

Effects of medication

In the PD group, the escitalopram- and paroxetine-treated groups did not differ in emotion recognition threshold, response time, response time_crt, or recognition error for all four emotions. There were also no significant correlations between medication dosage (escitalopram and paroxetine), duration of illness, and duration of medication with emotion recognition measures.


Confirming the hypothesis, the present results show that patients with PD have a general deficit in emotion recognition as evidenced by a higher average recognition threshold compared to healthy controls. In terms of individual emotions, although trend only, the recognition threshold for fear and the response time for anger were higher in the PD group. This partially matches with results from a study by Kessler et al.,[18] but this is in direct contrast to results from a recent study by Reinecke et al.[16] One possible reason for these contradictory results might be the treatment status of the PD status. All of the present PD participants and the Kessler et al. PD participants were under medication,[18] whereas all PD patients in the Reinecke et al. study were untreated.[16] Evidence indicates that serotonin re-uptake inhibitors or serotonin norepinephrine re-uptake inhibitors can either reduce or modify emotional information processing biases.[21-24] Therefore, medication may have stabilized biased emotional processing in patients with PD in the present and the Kessler et al. studies. The fact that the mean PDSS score for the present PD patients was within the mildly ill level (8.0) might support this explanation.[25] Although the analysis of medication and emotion recognition measures showed no significant medication effects, it is still unclear whether and how medication could have affected emotion recognition processing. We believe that subsequent studies comparing drug-naïve patients and patients under medication may help resolve such inconsistencies regarding the recognition of facial expressions in PD.

Results from the correlation analyses suggest that symptom severity of PD might be associated with impairment in emotion processing of facial expression. In particular, PD patients with higher trait anxiety and a higher PDSS score responded more slowly to accurately recognize anger (trend only for PDSS). The present study also found that patients with a higher BDI score were slower to respond and to accurately recognize happiness and anger. These results suggest that PD patients with a higher symptom severity (trait anxiety, depressive symptoms, and PDSS) experience more difficulty in recognizing threat-related emotions. Previous studies have indicated that individuals with PD have greater difficulty in identifying and describing emotions[26, 27] and they especially have a greater tendency to suppress and constrict experience and expression of negative emotions.[28] One possible explanation might be because PD patients have a tendency to avoid threat-related situations and to use defensive withdrawal rather than an orienting response to threatening stimuli.[29, 30] Furthermore, individuals with PD not only lack emotional acceptance and clarity, but also tend to use avoidant regulation strategies in response to negative-related stimuli.[31] The relationship between higher symptom severity and decreased ability to recognize anger observed in the present study might also be explained with similar reasons, because it is well documented that higher panic symptom severity is associated with higher degree of avoidance.[32]

The correlation of BDI, STAI, PDSS scores with the response time, but not with recognition threshold in all four emotional expressions might also be caused by the medication effect. In patients with PD, symptoms may have initially altered recognition of threat-related emotions and slowed or inhibited responses to it.[33] As stated earlier, medication may have stabilized these emotional information-processing biases for threat-related emotions, but not affected response speed. A previous study in patients with depressive disorder supports this speculation in that it showed that use of antidepressants improved emotion recognition accuracy for both positive and negative emotions, but did not improve response speed to negative emotions.[20]

It is also of interest that recognition of happiness was associated only with depressive symptom severity in the PD group. Impairment in the recognition of happiness has been widely reported in depression.[7, 34] Hence, the depressive component of PD may have primarily influenced the recognition of happiness, although no patients with PD met the diagnostic criteria for a depressive disorder. A study by Kessler et al. supports this hypothesis and reports that depression might have a strong association with emotion recognition in patients with PD.[18]

In healthy volunteers, higher TAI had an association with slower response time and response time_crt for sadness only. Non-clinical individuals with higher trait anxiety have been reported to have difficulty in recognizing sadness.[35] More importantly, fear and anger rather than sadness are generally acknowledged as threat-related emotions. It is unclear as to why higher BDI scores were significantly correlated with slower response time for fear in healthy controls. This correlation, however, was not consistently observed in the control group, in contrast to the consistently observed correlation between emotion recognition disturbances for threat-related emotion (i.e. anger) across TAI, BDI and PDSS scores (trend only for PDSS); this result needs further investigation.

This study has several strengths. First, healthy controls included in the study were carefully selected. This enabled comparison of emotion recognition patterns between PD and control groups with minimal bias from other cofactors such as differences in demographic background and comorbid psychiatric disorders. Second, the homogeneity of participants in the PD group was also a strength. Specifically, the fact that all participants in the PD group were taking medication minimized effects from differences in treatment status when analyzing correlations between emotion recognition patterns and symptom-related measures.

Several limitations of the study must also be noted. The homogeneity of participants in the PD group, which was the second strength of the study, might also be a limitation. The PD group did not include patients with untreated PD or those receiving psychotherapy with or without medication. Thus, it was not possible to investigate how drug and psychotherapy affect emotion recognition impairment in PD or whether such impairments can be reversed by these therapies. Additionally, the PD group included patients both with and without agoraphobia, but the effect of agoraphobia on emotion recognition impairment was not analyzed. One could argue that the correlation between symptom-related measures and emotion recognition impairment observed in this study may have been driven by the presence and severity of agoraphobia. The relatively small sample size is another important limitation. Thus, it might not be possible to generalize the present results to all PD patients.

Despite these limitations, we believe the present findings provide guidance for future studies regarding facial emotional recognition in patients with PD. The present study provides evidence that patients with PD have a general deficit in the recognition of emotion, which might be more prominent for threat-related emotions. More importantly, the results suggest that symptom severity might be associated with impairment in emotion processing of threat-related facial expressions. This confirms that patients with PD tend to use avoidance rather than orienting responses to threatening stimuli and emotions. Such defensive withdrawal may hinder the effectiveness of exposure therapy.[36] Whether this impaired emotion processing can be reversed is an important unanswered question. We believe that subsequent studies with larger sample sizes and varying treatment status are needed to clarify these issues.


This study was supported by a grant from the Korean Research Foundation (2006–2005152).