Major depressive disorder has been associated with volumetric abnormality in the amygdala. In this meta-analysis we examine results from magnetic resonance imaging volumetry studies of the amygdala in depression in order to assess both the nature of the relationship between depression and amygdala volume as well as the influence of extraexperimental factors that may account for significant variability in reported findings. We searched PubMed and ISI Web of Knowledge databases for articles published from 1985 to 2008 that used the wildcard terms 'Depress*' and 'Amygdal*' in the title, keywords or abstract. From the 13 studies that met inclusion criteria for our meta-analysis, we calculated aggregate effect size and heterogeneity estimates from amygdala volumetric data; we then used meta-regression to determine whether variability in specific extraexperimental factors accounted for variability in findings. The lack of a reliable difference in amygdala volume between depressed and never-depressed individuals was accounted for by a positive correlation between amygdala volume differences and the proportion of medicated depressed persons in study samples: whereas the aggregate effect size calculated from studies that included only medicated individuals indicated that amygdala volume was significantly increased in depressed relative to healthy persons, studies with only unmedicated depressed individuals showed a reliable decrease in amygdala volume in depression. These findings are consistent with a formulation in which an antidepressant-mediated increase in levels of brain-derived neurotrophic factor promotes neurogenesis and protects against glucocorticoid toxicity in the amygdala in medicated but not in unmedicated depression.