Previous studies have suggested that type 2 diabetic mellitus could lead to learning and memory deficits. We studied cognitive function tests and brain computed tomography (CT) findings in elderly subjects with drug-treated type 2 diabetic patients (n = 9), diet-treated type 2 diabetic patients (n = 8) and nondiabetic subjects (CR, n = 21). A battery of cognitive function tests (Cog-T; WAIS-R's digit span test and symbol test, Stroop Test, ADAS's verbal memory test, and MMSE) was carried out on two occasions, separated by at least 6 months. Brain CT was analyzed by the following 5 variables; 1) Evan's Ratio, 2) Inverse Cella Media Index, 3) maximum width of the third ventricle, 4) maximum width of temporal horn tips on both sides and 5) maximum width of the Sylvian fissure at the insula, bilaterally. The scores of Cog-T did not differ significantly between the groups. On brain CT measurements, maximum width of the temporal horn tips on right side were significantly different in the three groups (ANOVA, P = 0.035). The drug-treated diabetics subjects had wider temporal horn tips on the right side than did the diet treated diabetics and nondiabetic subjects (Fisher's post hoc test, P = 0.030, P = 0.016).