The distinction between the depressive troubles according to their inclusion in bipolar disorders or in recurrent depressive disorders offers an evident practical interest. In fact, the curative and mainly the preventive treatment of these troubles are different. So it is necessary to identify the predictive factors of bipolar development in case of inaugural depressive episode. In 1983, Akiskal was the first who identified those factors: pharmacological hypomania, puerperal depression, onset at early age (<25 years), presence of psychotic characteristics, hypersomnia and psychomotor inhibition. Through this study, the authors try to compare the epidemiological, clinical and evolution characteristics of major depression in bipolar disorders to recurrent depressive disorders in order to indicate the correlated factors with bipolarity. It is a retrospective and comparative study based on about 155 inpatients for major depressive episode during the period between January 1994 and December 1998. These patients were divided into two groups according the DSM IV criteria: bipolar group (96 patients) and recurrent depressive group (59 patients). Both groups were compared according to socio-demographic data, life events in childhood, personal and family history, clinical and evolution characteristics of the index depressive episode. The predictive factors proposed by Akiskal were systematically examined. It was found out that the following factors were correlated with bipolarity: high rate of separation and divorce (17.7% versus 5.1%; p=0.02), family history of psychiatric disorders (56.3% versus 35.6%; p=0.012) especially bipolar ones (29.2% versus 3.4%; p=0,00008), onset at early age (mean age of onset: 24.8 8.2 years versus 34.1 12.6 years; p=0.000004), number of affective episode significantly more frequent (mean 3.6 versus 2.5; p=0.03), sudden onset of depressive episode (44.8% versus 15.9%; p=0.0003) and presence of psychotic characteristics (69.8% versus 16.7%; p=0.0001) catatonic characteristics (37.3% versus 20.3%; p=0.03), hypersomnia (51% versus 20.3%; p=0.03) and psychomotor inhibition (83.3% versus 42.4%; p=0.00007). Negatively correlated factors of bipolar depression were: somatic comorbidity such as diabetes, hypertension and rhumatismal diseases (12.5% versus 28.8%; p=0.012) and association with dysthymic disorders (2.2% versus 12.1%; p=0.029). No correlation was found between bipolarity and life events in childhood, seasonal character, alcoholic dependence and suicide attempt. Concerning the validity of predictive factors of bipolarity proposed by Akiskal, we found: history of bipolar disorders (Sensibility: 29.2%, specificity: 96.6%, Positive Predictive Value (PPV): 93%), hypersomnia (Sensibility: 51%, specificity: 80%, PPV: 80%), onset before the age of 25 years (Sensibility: 62.5%, specificity: 70%, PPV: 77%), psychomotor inhibition (Sensibility: 83.3%, specificity 58%, PPV: 76%), and psychotic characteristics (Sensibility: 69.8%, specificity: 62.7%, PPV: 75%). In spite of methodological differences, our results tallied with the other studies. We focus on the importance of the bipolar family history criterion, which has the highest PPV, and the limits of psychotic characteristics criterion which has the lowest PPV. This may be explained by the frequency of these characteristics of affective disorders in our cultural context. The association of the hypersomnia and psychomotor inhibition in one criterion in order to increase their diagnostic power. Our study helps us to identify the factors that would predict the bipolar evolution of a depressive episode allowing the use of specific treatment and ensuring the improvement of prognostic.