Ｏｂｊｅｃｔｉｖｅ　To investigate the effects of sevoflurane exposure at different concentrations during late pregnancy on cognitive function and dendritic spines in offspring rats．　Methods　Sixteen pregnant SD rats were randomly divided into 4 groups (n=4): control group (group C), sevoflurane 1 group (group S1), sevoflurane 2 group (group S2), and sevoflurane 3 group (group S3). Rats of group S1, group S2 and group S3 were anesthetized respectively by 1.5%, 2.8%, and 3.3% sevoflurane and maintained under 0.5, 1.0, 1.5 MAC in O2 2 L/min for 4 h successively during 3 d from 18th to 20th day of pregnancy, while group C inhaled oxygen 2 L/min only. The offspring rats were tested for cognitive function through Morris water maze at 21th day after birth. All rats were sacrificed after behavior test at 28th day after birth, and the hipoocampus were then removed to determine the dendritic spine density by Golgi staining, and the expression of caspase-3 mRNA was measured by reverse RT-PCR.　Results　There was no significant differences in the findings of Morris water maze including swimming distances, escape latency, frequency of crossing the original platform, and the time of staying at the original platform quadrant (P>0.05). Compared with group C [(12±3)/10 μm], the density of dendritic spines in hipoocampus were significantly decreased in group S3 [(9±3)/10 μm] (P<0.05). There was no significant differences in the expression of caspase-3 mRNA in hipoocampus among groups (P>0.05).　Conclusions　1.5 MAC sevoflurane exposure for 4 h might reduce the density of dendritic spines in hipoocampus; however, different concentrations of sevoflurane exposure during late pregnancy affect the cognitive function and expression of caspase-3 mRNA in hippocampus in offspring rats.