MR is generally known to be more sensitive but less specific than CT in the detection of brain lesions. In our opinion multiple echo sequences can markedly improve MR specificity in the diagnosis of tumors. We reviewed a series of 343 intracranial tumors studied with MR using multiple echo sequences and histologically verified. On the basis of the different signal patterns we divided brain tumors into 5 classes. Class 1: the signal intensity of the tumor increases progressively in T2 WI (100% of craniopharyngiomas, 21/21; 100% of epidermoid tumors, 12/12; 81% of astrocytomas (grades I to III), 64/79; 65% of neurinomas, 30/46). Class 2: the signal intensity of the lesion decreases progressively in T2 WI: A) the tumor has higher signal intensity than the parenchyma in all echoes (100% of medulloblastomas, 14/14; 53% of pituitary adenomas, 15/28); B) the tumor has the same signal intensity as the parenchyma in late echo acquisitions (100% of ependymal tumors, 12/12; 60% of meningiomas, 25/41). Class 3: the tumor has the same signal intensity as the parenchyma in all echoes (34% of meningiomas, 14/41). Class 4: glioblastoma model: one or more cysts of high signal intensity in T2 WI and slightly hyperintense nodules and/or rings and hyperintense peritumoral edema (73% of glioblastomas, 35/48; 72% of metastases, 18/25). Class 5: oligodendroglioma model: mixed hyper/hypointense pattern; cyst, calcifications and edema are very difficult to recognize within the lesion (95% of oligodendrogliomas, 18/19). The signal pattern was sometimes characteristic but never pathognomonic. Nevertheless, this classification proved to be an useful criterion to restrict the number of possible diagnoses. The study of T1 and T2 values seems to be less useful.