Emerging evidence suggests that imagery-based techniques may enhance the effectiveness of traditional verbal-linguistic cognitive interventions for emotional disorders. This study extends an earlier pilot study by reporting outcomes from a naturalistic trial of an imagery-enhanced cognitive behavioural group therapy (IE-CBGT, n = 53) protocol for social anxiety disorder (SAD), and comparing outcomes to historical controls who completed a predominantly verbally-based group protocol (n = 129). Patients were consecutive referrals from health professionals to a community clinic specialising in anxiety and mood disorders. Both treatments involved 12, two-hour group sessions plus a one-month follow-up. Analyses evaluated treatment adherence, predictors of dropout, treatment effect sizes, reliable and clinically significant change, and whether self-reported tendencies to use imagery in everyday life and imagery ability predicted symptom change. IE-CBGT patients were substantially more likely to complete treatment than controls (91% vs. 65%). Effect sizes were very large for both treatments, but were significantly larger for IE-CBGT. A higher proportion of the IE-CBGT patients achieved reliable change, and better imagery ability was associated with larger symptom change. Outcomes compared very favourably to published group and individual treatments for SAD, suggesting that IE-CBGT may be a particularly effective and efficient mode of treatment delivery.