To assess the value of MRI for meningioma of the posterior cerebral fossa, in correlation with surgical and pathological findings, we retrospectively reviewed 31 cases. The patients (24 females and 6 males ranging in age from 25 to 79 years) were preoperatively studies on a 1,5 T MR imager (GEMS Signa) between july 1989 and november 1993. The protocol included: 1. MR scan with axial sections in T2-weighted spin-echo sequence (3 mm thickness), T1-weighted spin-echo sequence before and after gadolinium injection (3-5 mm thickness), coronal and sagittal T1-weighted sections performed after injection. 2. Surgery reports. 3. Histopathological reports; the predominant histological subtype of each tumor was graded according to the classification scheme of Russel and Rubinstein. We focused on five items: 1. The site of the dural attachment of the meningioma. 2. Tumoral extensions (to the tentorium, to the jugular foramen, to the internal auditory canal). 3. The meningioma signal in T1- and T2-weighted sequence using the same visual scoring system for grading signal intensities as Elster and al. 4. Secondary features (necrosis, cysts, calcifications) within the tumor. 5. Interface between meningioma and encephalic structures. Meningiomas arose from the posterior surface of the petrous bone in 74% of the cases and from the clivus in 9.6%. Meningiomas were bulky at the time of diagnosis as since tumoral arrow overtook 2 cm in 64.5% of the cases. Surgical approach was guided by an anatomo-radiologic classification based on the exact site of tumoral dural attachment. This determination relied on: 1. Osseous reaction noted in 58% of the cases (enostosic spur in 19%, localized osseous thickening in 16%). 2. The trigeminal nerve displacement by the tumor; in case of clival meningioma extended to the petrous apex, this nerve is displaced outside; otherwise, meningioma of the petrous bone extended to the clivus displaced the trigeminal nerve inside. 3. Radiate structure within tumor converging to vascular basal pole of the meningioma noted in 42% of the cases. Tentorial involvement remained a difficult diagnosis on MR images. It was affirmed when the tumor extended on the opposing surface of the tentorium and when focal hypersignal existed through the usual tentorial hyposignal on T2-weighted images and T1-weighted images after gadolinium. On the other hand, tentorial linear dural enhancement adjacent to the tumor was not a reliable sign (error in 15.8% of the predicted cases). The meningothelial (syncitial) type was noted in 67.7% of the cases. (ABSTRACT TRUNCATED AT 450 WORDS)