The NIR spectra of 502 domestic raw cottons were analyzed to confirm the ability of NIR technology to predict simultaneously multiple fundamental properties of cotton fibers. The precision of several NIR multivariate calibration techniques were compared to the predictive ability of traditionally applied multiple linear regression using the criteria of Standard Error of Prediction. The potential misuse of the R2 criteria in fitting the prediction equations was also examined. The statistical analyses demonstrated that the precision of multivariate algorithms in predicting fiber properties matched the precision of the arealometer reference method. The NIR spectra and arealometer data were further analyzed to determine what potential improvements in the methodology would be necessary to better predict cotton maturity and fineness.