Selective attention to angry faces in clinical social phobia

Manuscript Critique

Mogg, K., Pierre P., & Bradley, B.P. (2004). Selective attention to angry faces in clinical social phobia. J Abnormal Psych, 113 (1), 160-165.

The present study investigated the time course of attentional biases to emotional facial responses in patients with diagnosed social phobia. The social phobia group showed enhanced vigilance to angry faces, relative to happy and neutral faces, compared to matched controls at 500 ms but not 1250 ms of exposure duration

The results of the present study provide evidence for initial vigilance for angry faces in patients with clinical social anxiety. These data are consistent with several studies related to cognitive bias in anxiety disorders (Mogg & Bradley, 1998). Results from this study suggest that social phobia has a different pattern of attentional bias from other anxiety disorders. Social phobia is characterized by attentional avoidance rather than vigilance for external threat cues.

The finding of attentional bias for angry faces at 500 ms in patients with high-trait social anxiety is consistent with previous research (Mogg & Bradley, 1998). However, several studies have reported avoidance of emotional faces presented at 500 ms (Mansell, Clark, Ehlers, & Chen, 1999). Thus, research in this particular field is inconclusive at best.

This was a prospective, nonrandomized, controlled trial. A convenience sample of 15 test subjects were enrolled based on diagnosis of social phobia. Control subjects had no known history of psychological disorders and were matched for baseline characteristics such as age, gender, and educational status. The control group used in this study was adequate. A stronger study design would have been to randomly select subjects within each group, thus minimizing bias. A potential confounding factor that was not controlled for was reaction time. Thus, reaction times may have been different between groups at study onset. This variable may have potentially been used as a covariate in the analysis of variance model to minimize confounding effects.

There is no indication as to whether a power analysis was conducted before this study. Thus, it is impossible to determine if this study had ample power to detect significant differences if in fact they truly existed. Aside from a lack of psychological disorders and matched baseline variables, selection criteria for the control group were not stated. Thus, selection bias may have been a threat to the internal validity of this study.

The authors readily admit that techniques such as online monitoring of eye movements during attentional tasks may provide a more stable measure of attentional bias. However, the use of common testing mechanisms makes comparisons to previous research findings possible. Of particular interest was that response time data from one test subject was excluded from statistical analyses because of the high variation in response time. Other data were excluded because of high variability, although the authors do not state the quantity that was discarded. Due to these effects observed in a small sample, questions are raised as to the reliability of response time for measuring attentional bias.

All significant outcomes in this study are likely true effects given the small sample size. However, many nonsignificant effects and interactions are likely due to insufficient sample size. Furthermore, one may argue that the social phobia cohort had faster reaction times because they may have participated in similar psychological tests in the past and had experienced a learning effect of the testing.

The following text provides a proposed design to strengthen the study. A power analysis will be conducted a priori to determine an adequate sample size. Test and control subjects will be randomly selected from their respective cohorts. Initial response time to a neutral stimulus will be measured and used as a covariate if necessary. Furthermore, practice tests will be given on three separate days to minimize a learning effect of the testing. The addition of these factors to the current study design would aid in increasing statistical power and minimizing bias.