BACKGROUND: A previous factor analysis of pooled data demonstrated that the Liebowitz Social Anxiety Scale (LSAS) can be divided into six subscales. This paper examines data from a fixed-dose trial of escitalopram versus paroxetine, in order to determine the differential effects of these agents on symptom dimensions in social anxiety disorder (SAD). METHODS: Data from a 24-week randomised, placebo-controlled, comparative study of fixed doses of escitalopram (5 mg, 10 mg, 20 mg) versus paroxetine (20 mg) in SAD were examined. The six factors identified in a previous factor analysis of baseline data from escitalopram studies on the primary efficacy scale, the LSAS, were used to compute subscale scores. These were analysed using analysis of covariance (ANCOVA), and standardised effect sizes were calculated. RESULTS: The combined escitalopram data and the paroxetine data both demonstrated significant superiority to placebo on each of the 6 LSAS factors at week 24 (OC analysis). Escitalopram doses of 5 mg, 10 mg, and 20 mg were generally more effective than placebo for each of the factors. Escitalopram 20 mg was significantly more effective than paroxetine 20 mg on 5 of the 6 symptom dimensions. CONCLUSION: Factor analysis of the LSAS allows for useful secondary analyses that support and extend the primary efficacy analysis of this instrument. The analysis here indicates that different escitalopram doses are effective across the various symptom dimensions of SAD.