[Objective] This study was conducted to evaluate three training systems for Korla fragrant pear trees by comparing the characteristic parameters of light response curve and CO2 response curve, and to provide some theoretical basis for improving the pruning techniques of Korla fragrant pear trees. [Method] The light response curve and CO2 response curve of the trees trained to three systems were measured by LI-6400 portable photosynthesis system. The SPAD value was measured using SPAD-502 chlorophyll meter, and specific leaf weight was calculated, to evaluate the effects of the three training systems. [Result] The CO2 response curves of the three training systems were basically in agreement with their light response curves, but there were some differences in their characteristic parameters. Among the three training systems, the maximum net photosynthetic rate, apparent quantum yield and light compensation point of espalier trained trees were the highest, while their light saturation point was the lowest. The CO2 saturation point of delayed-open central leader trained trees and open center trained trees were 1 752 and 1 665 μmol/mol, both of which were much higher than that of espalier trained trees. In addition, the carboxylation efficiency and photorespiration rate of espalier trained trees were both higher than those of delayed-open central leader trained trees and open center trained trees, while the CO2 compensation point of espalier trained trees was the lowest. The leaf SPAD value of espalier trained trees was the largest, followed by that of open center trained trees, and the leaf SPAD value of delayed-open central leader trained trees was the smallest. In addition, the leaf area and specific leaf weight of espalier trained trees were both the highest, followed by those of open center trained trees. [Conclusion] Among the training systems for Korla fragrant pear trees, the espalier training system had better ability to capture light, higher photosynthetic productivity and strongest adaptability to light en