A calibration set of 140 soybeans from seven different varieties, ranging in moisture from 5 to 22%, wet basis, was used to calibrate a spectrophotometer for predicting moisture of single soybean seeds. Near-infrared absorbance (A) of individual soybean seeds was measured over the spectral region from 800 to 1100 nm by 0.5 nm increments. The axis of the soybean seed parallel to the incident light beam was measured as an estimate of optical pathlength. Three mathematical techniques were used to develop calibration equations: linear correlation with a difference in absorbance (AA), stepwise multiple linear regression (MLR), and partial least squares (PLS). A validation set contained 100 soybeans from 5 different varieties, ranging in moisture from 5 to 20%, wet basis. The standard error of prediction (SEP) for equations using absorbance data only was 0.88% for AA, 0.82% for MLR, and 0.81% for PLS. The SEP for equations using pathlength and absorbance data was 0.73% for AA, 0.69% for MLR, and 0.65% for PLS.