The photosynthetic rate (Pn) of winter wheat (Triticum aestivum L.) is directly affected by water conditions, which can be effectively estimated by coupling vegetation indices (VIs). Establishing sound empirical models between VIs and Pn of winter wheat under different water conditions therefore is a sound approach for predicting Pn for broad applications. A rainout‐shelter field experiment was conducted to investigate the changes in Pn and VIs (normalized difference vegetation index [NDVI], enhanced vegetation index [EVI], ratio vegetation index [RVI], photochemical reflectance index [PRI]) under five water stress conditions. We found that: (a) heavy and moderate drought stress inhibited Pn, with their accumulated effects became more significant with the duration of treatments while the light drought stress and slight water logging produced minor effects on Pn. Heavy and moderate drought stress also caused the peak time of Pn preceded during the growth period; (b) NDVI showed low sensitivity to vegetation coverage while EVI explained more seasonal variation than other VIs when leaf area index (LAI) was >3; (c) the linear models between Pn and EVI under heavy (normalPnormaln=2.0449EVI−1.2906) and moderate (normalPnormaln=1.7742EVI−1.7021) water stress were selected as the best model for Pn, while a power function for Pn‐NDVI performed better under sufficient water supply (Pn = 37.982NDVI3.0101) and slight water logging (Pn = 28.024NDVI2.5646). Our results demonstrated that Pn–VI relationships varied by water stress for a winter wheat over the growing season. These results could be used for the prediction of winter wheat Pn under water stress at a large spatial scale.