Abstract Bipolar disorder is a disease that affects 1% of the general population. It is a great social burden due mostly to temporal disability and to the medical care it entrains. The management of bipolar disorder is both curative on relapse phases, and preventive on remission ones. Researchers have insisted on the importance of preventive treatment on the management of bipolar disorders and therefore on the detection of prodromic symptoms and then on the treatment of relapses before they create functional disability. In their opinion, relapses are a result of an imperfect management of remissions. In practice, it is difficult to detect prodromic relapse signs, as the practitioner's diagnosis depends greatly on the patient's insight. Thus, a patient may forget to indicate signs that have a pathologic signification. In such a situation, any instrument detecting the insidious beginning of relapse will be helpful. The aim of this study is to create a scale to detect relapses in bipolar disorders and to proceed to its preliminary validation. The scale was created based on the visual analogical model. It consisted in nine faces showing emotions grading from sadness to extreme joy. The patients were invited to choose the face that corresponded to their best mood in the preceding week. In this logic, detection of relapses is based on detection of pathologic mood variations.The validation was made on a sample of 77 patients suffering mostly from bipolar disorder subtype I. Two thirds were primary school graduates and 84% were men. Patients were evaluated by the Montgomery Asberg Depression Rate Scale (MADRS) for those suffering from depression, and the Young Manic Rate Scale (YMRS) for those suffering from manic episodes. The patients on remitted phases were evaluated by both scales. A total of 94 evaluations were computerized. Twenty four concerning depressed patients, 26 concerning patients in manic episodes and 44 concerning patients on a remission phase. Reliability was tested using the “test–retest” method. The results showed no statistical difference between the different subgroups. The Gold Standard chosen to assess validation was the DSM-IV diagnosis and the concordance rate between the MGRS and the DSM-IV diagnosis was 0,79. The sensibility of the MGRS on diagnosis of “abnormal mood variation” was 94%. Its specificity on the diagnosis of “normal mood variation” was 91,9%. The sensibility to mood variation was tested on a small sample and patients were evaluated at admission and at discharge. The result showed that the MGRS patients were more sensible to mood change than the MADRS patients and less sensible than YMRS patients. The analysis of responses relative to gender showed a significant difference between men and women in the depressed group. That difference may be due to the gender difference in emotion recognition. The analysis of responses regarding educational level showed no statistical differences. The primary validation shows an encouraging result. The authors think, however, that a validation on a larger and more homogenous sample is necessary to confirm the results.