A method to discriminate textiles was proposed based on terahertz time-domain spectroscopy (THz-TDS) and clustering analysis, and some typical cotton textiles were investigated to prove its feasibility. Their time domain waveforms were measured using THz-TDS system and then their absorption spectra were obtained. Principal component analysis (PCA) was applied to extract features of the data, and then Mahalanobis distance discriminant method was employed to classify these materials. The results show that this method can classify these five textiles accurately. It indicates that the method to classify textiles is feasible which combines PCA and Mahalanobis distance discriminant method based on their THz absorption spectra. The proposed method has a potential for identifying textiles of similar composition.