The purpose of this study is to investigate the feasibility using near-infrared spectroscopy (NIRS) technology for determination of the 1000-grain weight of paddy. Seventy varieties of paddy samples were scanned in the NIR spectral region 4000～10000cm-1. Ten pretreatment methods were used to process the NIR spectrums. NIR calibration models were developed by using the partial least squares (PLS) technique. The best calibration model was determined based on the highest determination coefficient (R2) of cross-validation and the lowest root mean of standard error for cross validation (RMSECV). It was shown that the best model, with the determination coefficient 0.773 and RMSECV 1.67g, could be obtained based on the processed spectral data by the pretreatment min—max normalization in spectral range 11998.9～7497.9＋6101.7～5449.8＋4601.3～4246.5cm-1.The validation results indicated that the best calibration model had correlation coefficient (r) between the NIR predicted and the actual values of 0.945 and standard prediction error (SEP) of 0.76g, showing that the NIR technology is a potential tool to estimate the 1000-grain weight of paddy.