Objective To explore the differences in clinical features between bipolar depression and unipolar depression from the clinical phenomenology. Methods The clinical data of 200 patients with bipolar depression (bipolar depression group) and 563 patients with recurrent unipolar depression (recurrent unipolar depression group) were compared, and multiple stepwise Logistic regression was employed to identify the clinical features related to bipolar disorder. Results The age at onset in bipolar depression group was younger than that in recurrent unipolar depression group (P<0.001), the percent of patients with age at onset <25 years in bipolar depression group was significantly higher than that in recurrent unipolar depression group (P<0.01). Hypersexuality, which was one of the atypical depressive symptoms, were more common in bipolar depression group than in recurrent unipolar depression group (P<0.01). The percents of patients with psychiatric symptoms, psychomotor retardation, mood instability and duration of every depressive episode <3 months in bipolar depression group were much higher than those in recurrent unipolar depression group (P<0.05). The score of cognitive impairment factor, one of the factors of 17-item Hamilton Rating Scale for Depression (HAMD-17), was significantly higher in bipolar depression group than in recurrent unipolar depression group (P<0.05). Logistic regression analysis revealed that age at onset, age at onset <25 years, family history of bipolar disorder, hypersexuality, psychiatric symptoms, psychomotor retardation, mood instability, duration of every depressive episode <3 months and score of cognitive impairment factor were independent factors for bipolar disorder, and the odds ratios were 1.54, 1.50, 3.25, 1.99, 1.89, 1.48, 1.63, 1.63 and 1.42, respectively. Conclusion Bipolar depression and unipolar depression are different disorders, and distinct clinical features may be potential predictors of bipolar disorder.