Objective: Patients with major depressive disorder (MDD) have been reported to perform less well in neurocognitive tests than normal control subjects. The author tested the hypotheses that verbal working memory (WM) is predictive of the functional outcome in patients with MDD. Methods: In this naturalistic longitudinal study, the subjects consisted of 22 adult outpatients receiving paroxetine as antidepressant therapy. Functional outcome was rated on a scale of 0 (non-impaired) to 3 (severely impaired). Results: 1) At 12 weeks, nine of the 22 patients currently experiencing MDD exhibited full remission; 2) significantly decreased 7-item Hamilton Rating Scale for Depression (HAM-D7) scores were observed during the 12-week study period, while Digit Sequencing Task (DST) scores increased significantly; 3) at baseline, functional outcomes correlated significantly with HAM-D7 scores, but, at 12 weeks, correlated significantly with both HAM-D7 and DST scores. Furthermore, when looking at only patients in full or partial remission (mild depression), functional outcome correlated more strongly with DST than with HAM-D7 scores. Conclusions: A deficit of verbal WM correlated with the functional outcome after treatment in patients with MDD. Antidepressant therapy with paroxetine might contribute to improvement of verbal WM.