The Chinese Naoshekang Recipe (CNR) extract used to treat diabetes is a traditional Chinese prescription composed of 9 crude drugs: Astragalus Membranaceus, Prepared Rehmannia Root, Solomonseal Rhizome, Szechwan Lovage Rhizome, Barbary Wolfberry Fruit, Epimedium Leaf, Atractylodes Lancea, Kudzuvine Root, and Coptis chinensis Franch. In our previous study, the CNR and its constitutive extracts showed a significant improvement in electrophysiological loss of diabetic patients or cerebral ischemia-reperfusion rabbits. This study was designed to investigate the effect of the CNR extract on learning and memory ability of rats with diabetes mellitus, and to study its neuroprotective mechanism. Thirty-four male Wistar rats were randomly divided into 3 groups: normal control group (NC group, n=10), diabetic control group (DC group, n=12), diabetic rats treated with the CNR extract (DT group, n=12). Diabetes was induced by a single intraperitoneal injection of STZ (60 mg/kg). Three days after the confirmation of diabetes, the rats of DT group were orally administered a dose of the CNR extract 600mg/kg body weight/d (prepared into 5% water solution with distilled water) by a stomach tube. NC and DC groups were only administered an equal volume of distilled water. All operations continued for six weeks. At the sixth weekend, we detected fasting blood glucoses (FBG), the learning and memory ability by Y-maze, and HE staining, immunohistochemistry and imageanalysis of brain tissue. The results indicated that (1) FBGs in DC and DT groups are higher than that in the NC group with a significant difference (all p<0.01), and FBG in DT group is significantly lower than that in DC group (p<0.01); (2) the scores of Y-maze in DC group is significantly higher than that in NC group (both p<0.05), while the scores in DT group is significantly higher than those in DC and NC groups (all p<0.001); (3) the hippocampal neuron counting in DC group is significantly lower than that in NC group (p<0.001), while that in DT group is significantly lower than that in NC group, significantly higher than that in DC group (both p<0.001); (4) the Bcl-2 expression in DT group is significantly higher than that in DC group, significantly lower than that in NC group, while the Bax expression in DT group is significantly lower than that in DC group, significantly higher than that in NC group(all p<0.05). We conclude that (1) the CNR extract improves the learning and memory ability of diabetic rats; (2) its mechanism may regulate the expression of apoptosis protein Bax and anti-apoptosis protein Bcl-2, and inhibit apoptosis of hippocampal neurons.