Objective To investigate cognitive impairment and astrocytes reaction in diabetic rats and the influence of the insulin therapy.Methods 52 health grown-up male Sprague-Dawley (SD) rats ware divided into 2 big groups stochastically: 1-month group (n=26) and 3-month group (n=26); Each group then divided into 3 groups stochastically: diabetic group (n=10), insulin treatment group (n=10) and control group (n=6).Streptozotocin-induced diabetic model rat (60 mg/kg) was established.In the insulin treatment group, the rats were treated with isophane insulin everyday (2-3 U/day) based on blood glucose concentration.Tests of Morris water maze were taken place 1 month and 3 months later respectively, after which slices taken from hippocampus, basal ganglia, the white matter nearby lateral ventricle and frontal lobe were used to immunohistochemistral tests.The expression of GFAP positive cells was observed in diabetic rats.Results (1) Spatial study and memory ability estimate.Compared with control group, the swimming incubation period was prolonged in 3 month insulin treatment group (P<0.05) and diabetic group including both 1-month (P<0.05) and 3-month groups (P<0.05), while that of 1-month insulin treatment group was not prolonged.(2) Compared with control group, the change of GFAP expression in different sites of brain was not significant in 1-month diabetic group and insulin treatment group including 1-month and 3-month groups.While in 3-month diabetic group the number of GFAP positive cells was more than control group in basal ganglia (P<0.05) and white matter nearby lateral ventricle (P<0.05).The number of GFAP positive cells in 3-month insulin treatment group was less than diabetic group in basal ganglia (P<0.05) and white matter nearby lateral ventricle (P<0.05).Conclusion (1) Diabetic rats present learning and memory dysfunction at 1 st month.(2) The astrocytes change doesnt appear with early cognitive impairment in diabetic rats, suggesting cognitive impairment is not caused by astrocytes change.The astrocytes reactive proliferation which has regional selectivity can be found in long-term diabetes (3 months).(3) The early insulin application in diabetic rats can delay the impairment of cognitive function, but with the development of diabetes mellitus, the insulin treatment cant prevent the occurrence of cognitive impairment in diabetic rats.