Objective To investigate the learning curve of mouse cervical heart allograft model. Methods A retrospective analysis was conducted of 60 cases of mouse cervical heart allograft. The total transplantation time, recipient preparation time, donor preparation time, arteriovenous cannulation time, and donor cardiovascular connection time were recorded during the operation. Cumulative sum （CUSUM） analysis method and the best fitting curve were used to obtain the minimum number of animals required to pass the learning period in the learning curve of each surgery stage. The mice were divided into a learning stage group and a mature stage group with the minimum number of cases for the surgery stage with the longest learning curve as cut-off point, and the differences in operation time and failure rate were compared between the two groups. Results Sixty mice completed neck heart allograft. The total transplantation time was （119.80±43.17） min, the recipient preparation time was （76.43±26.46） min, the donor preparation time was （29.85±9.82） min, the arteriovenous cannulation time was （32.10±21.62） min, and the donor cardiovascular connection time was （17.47±8.43） min. A cubic CUSUM learning curve was best fit to the data. The minimum number of surgical cases corresponding to the learning curve of each surgery stage was 25 cases for recipient surgery, 19 cases for donor surgery, 29 cases for arteriovenous cannulation, 24 cases for donor cardiovascular connection, and 25 cases for the whole operation. Taking the 29 cases for the surgery stage with the longest learning curve （arteriovenous cannulation） as cut-off point, compared with the learning stage group, the total transplantation time [（156.80±32.80） min vs. （85.06±9.45） min], recipient preparation time [（95.41±18.89） min vs. （50.94±7.45） min], donor preparation time [（37.03±9.51） min vs. （23.13±2.95） min], arteriovenous cannulation time [（50.41±16.65） min vs. （14.97±5.74） min], and donor cardiovascular connection time [（24.38±7.16） min vs. （11.00±1.55） min] were significantly reduced in the mature stage group （P<0.01）. The surgical failure rate of the mature stage group was significantly lower than that of the learning stage group （48.28% vs. 9.68%, P=0.0013）. Conclusion The learning curve of the cervical heart allograft model is slightly different in each operation stage, and the surgical technique is basically mature when the sample number reaches 29 cases. After the surgical technique is mature, the operation time and the mature success rate will be significantly improved.