Forty subjects were clinically examined using scales for cerebellar, pyramidal, parkinsonian, and mental status and by quantitative evaluation of neuroimages. The patients were classified into two groups: cerebellar-plus and "pure" cerebellar syndromes. Patients with "pure" cerebellar syndrome were diagnosed as autosomal dominant cerebellar ataxia III (ADCA III) or "pure idiopathic" late-onset cerebellar ataxia (ILOCA) in this series. Patients with cerebellar-plus syndrome were diagnosed as multiple system atrophy (MSA), subclassified as either ILOCA-plus, ADCA I, ADCA II or autosomal recessive LOCA. We have used visual (VEP), brainstem auditory (BAEP) and somatosensory (SEP) evoked potentials in order to establish their diagnostic validity. Cerebellar-plus syndrome and "pure" cerebellar syndrome showed overlapping VEP, BAEP and SEP abnormalities. VEP P100 latency, however, shows a certain ability to differentiate between the two groups (p = 0.08) and appears useful in distinguishing between sporadic cerebellar-plus syndromes (MSA or ILOCA-plus) and "pure" cerebellar syndromes (p < 0.02). The incidence of prolonged N9-N13 latency was significantly higher in the latter subgroup (p < 0.04) as well. Within cerebellar-plus syndromes, VEP, BAEP and SEP abnormalities were more frequent in inherited cases (ADCA I and II, along with autosomal recessive LOCA) than in sporadic ones. The most apparent differences were a higher incidence of abnormal BAEPs at brainstem level (p < 0.002), and of both peripheral and possible central SEP impairment in hereditary cerebellar-plus syndrome than in sporadic cerebellar-plus syndrome (p < 0.03). EP investigation is useful to a certain extent in differentiating between some variants of LOCA.