A comprehensive study on the quantification of sulfathiazole polymorphs by near infrared (NIR) spectroscopy combined with multivariate analysis is reported. Calibration models were developed for each of the sulfathiazole polymorphs that can be obtained as polymorphically pure samples, sulfathiazole forms I, III and V, in binary mixtures in the 0-15% concentration range. Partial least squares regression (PLS) was employed and different spectral pre-treatment algorithms including, multiplicative scatter correction (MSC), standard normal variate (SNV) transformation, 1st derivative calculations and SNV combined with 1st derivative were applied to the data. The NIR method was compared with X-ray powder diffraction (XRPD) and was found to give more accurate results in all cases except one where XPRD was more accurate. For the NIR method, limits of detection were < 0.5% except for form V in I/V mixtures where the limit of detection was 1.5%. Limits of quantification ranged from 1.1% to 4.6%. When quantities of forms II and IV were added to samples which were analysed using the model developed for form III in I, their presence was detected by very large prediction errors. A comparison of multivariate and univariate analysis of the NIR data demonstrated that, in general, multivariate methods were superior.