Light permeability of tree affects the accumulation of sugar in fruits, and there is a difference in quality of fruit between different types of trees. The effect of tree shapes on prediction of sugar content of Huanghua pears in harvest time by visible/near infrared spectroscopy (VIS/NIR) was evaluated. A total of 480 pears were picked from opening canopy trees and trellistrained trees in six batches, there was significant difference in fruit sugar content between two types of fruit trees (p<0.05). The root mean square error of prediction (RMSEP) and residual prediction deviation (the ratio of standard deviation to RMSEP, RPD) of models of single tree shape fruits for predicting similar fruits ranged from 0.69°Brix to 0.64°Brix, and from 2.08 to 1.97, while for all fruits of different tree shapes, they ranged from 1.31°Brix to 1.07°Brix, and from 1.14 to 1.36, respectively. The RMSEP of the combined model was 0.59°Brix, and RPD was 2.38. The results showed that the model established by single tree shape fruits had poor performance, while a model established by using combined shapes fruits would obtain ideal robustness and good accuracy. The results would help to further utilize VIS/NIR to test quality of fruits in harvest time and predict the optimal harvest time.