A study on statistical classification of aphasia by multivariate analysis methods was attempted. The author applied the SLTA (Standard Language Test for Aphasia) to several groups of aphasic patients who had been clinically diagnosed as Wernicke's aphasia (23 cases), Broca's aphasia (39), amnestic aphasia (17), conduction aphasia (4) and global aphasia (36). The SLTA scores were analyzed by means of discriminant, multiple discriminant and cluster analysis.     The results were as follows :     (1) The correct identification rate (CIR) between statistical classification and clinical classification averaged 88.7% (range 66.7% to 100%).     (2) The CIR was lower in analysis between Wernicke's and Broca's aphasia and Broca's and amnestic aphasia.     (3) Global aphasia was well discriminated from other types.     (4) In multiple discriminant analysis, cases of Wernicke's aphasia and Broca's aphasia overlapped.     (5) In cluster analysis, cases of Wernicke's aphasia and Broca's aphasia were not well separated.     The results suggested that it was to some extent possible to classify aphasia types using only the SLTA score, but discrimination between Wernicke's and Broca's aphasia is not always clear. A possible major cause for this is that items for evaluating fluency, prosody and phonetic construction are not contained in SLTA.