Objective To determine the serum levels of protein S100β at different time points and cognitive function postoperatively, and investigate their relationship with depth of anesthesia. Methods A total of 102 elderly patients, at American Society of Anesthesiologists (ASA) grade of Ⅱ-Ⅲ, undergoing abdomen surgical treatment with general anesthesia in our hospital from January 2014 to December 2014 were enrolled in this study. According to their bispectral index (BIS), they were divided into the deep anesthesia group (BIS≥45) and light anesthesia group (30＜BIS＜45), with 51 cases in each group. Heart rate (HR), narcotrend index (NTI), mean artery pressure (MAP), and serum S100β level were measured and recorded. Minimental state examination (MMSE), clock drawing test (CDT), and animal fluency test (AFT) were used to assess the occurrence of postoperative cognitive dysfunction (POCD). Results The deep anesthesia group had a obviously higher dose of propofol than the light anesthesia group (P＜0.01). The incidence of POCD was 13.7% in the deep anesthesia group, and 47.1% in the light anesthesia group, with significant difference between them (Chi square＝3.996, P＜0.05). At the time point of finishing operation, the 2nd hour and the 24th hour after operation, the serum levels of S100β were markedly lower in the deep anesthesia group than in the light anesthesia group (P＜0.05). Conclusion Different depths of anesthesia lead to same postoperative pain levels and exert similar anesthesia effect, and both cause nervous system damage. The postoperative serum levels of S100β indicate that the light anesthesia group results in more severe nervous system damage, suggesting that maintaining deep anesthesia reduce the occurrence of POCD.