Abstract Background White matter abnormalities may interfere with limbic–cortical balance and contribute to chronic depressive syndromes in the elderly. This study sought to clarify the relationship of SH to treatment response. We hypothesized that patients who failed to remit during a 12-week controlled treatment trial of escitalopram would exhibit greater SH burden than patients who remitted. Methods The participants were 42 non-demented individuals with non-psychotic major depression and 25 elderly comparison subjects. After a 2-week single blind placebo period, subjects who still had a Hamilton Depression Rating Scale (HDRS) of 18 or greater received escitalopram 10 mg daily for 12 weeks. Remission was defined as a HDRS score of 7 or below for 2 consecutive weeks. FLAIR sequences were acquired on a 1.5 T scanner and total SH were quantified using a semi-automated thresholding method. Results The patient sample consisted of 22 depressed patients who achieved remission during the study and 20 depressed patients who remained symptomatic. ANCOVA, with age and gender as covariates, revealed that depressed subjects had greater total SH burden relative to non-depressed controls. Furthermore, patients who failed to remit following escitalopram treatment had significantly greater SH burden than both patients who remitted and elderly comparison subjects, whereas SH burden did not differ between depressed patients who remitted and elderly comparison subjects. Limitations Patients were treated with a fixed dose of antidepressants and the index of SH is an overall measure that does not permit examination of the relationship of regional SH to treatment remission. Discussion SH may contribute to a “disconnection state” both conferring vulnerability to and perpetuating late-life depression.