Physiological and biochemical deterioration of naturally aged seed is very common as enzymatic activity is changed over time. Seed vigor is also reduced due to this degradation. The main purpose of this study was to investigate the discrimination between viable and non-viable naturally aged pepper seeds. In this study, Fourier transform near-infrared (FT-NIR) spectroscopy was used for spectral data acquisition in the spectral range of 4000-10000 cm-1 at the interval of 4 cm-1 spectral resolution with total 32 successful scans. After applying different preprocessing methods, spectral angle mapper (SAM) technique was employed as a blind source separation (BSS) tool. This technique investigated the spectral difference between a viable and non-viable seed. Afterwards, the partial least square discriminant analysis (PLS-DA) model was developed using multivariate data analysis. All the samples were set to a germination test to identify the viable and non-viable seed and the result was used to make calibration and validation data set. The PLS-DA model provided a good result with performance accuracy of 91% for calibration and 85% for prediction. The SAM technique and the beta coefficient derived from PLS-DA showed some spectral difference between viable and non-viable seeds that might be a reason to vary the germination ability of the seed. The results demonstrate the possibility of using the FT-NIR spectroscopic technique with chemometrics to separate seeds based on their viability, which is very important in the seed industry.