The aim of the present study is to investigate in detail the near IR (NIR) spectra of the three types of polyethylene, linear low-d. polyethylene (LLDPE), low-d. polyethylene (LDPE) and high-d. polyethylene (HDPE), and to develop calibration models that predict their phys. properties such as d., crystallinity and m.p. The effects of spectral resoln. on the classification and the prediction of d. for the three types of PE have been investigated. Furthermore, the NIR spectral differences among LLDPE, LDPE and HDPE have been explored in more detail using 2 cm-1 resoln. Principal component anal. (PCA) has been performed to differentiate the 18 samples of PE. They are classified into three groups, LLDPE, LDPE and HDPE, by a score plot of the PCA factor 1 vs. 3 based on the NIR spectra pretreated by multiplicative scatter correction (MSC). The 2 cm-1 spectral resoln. yields a slightly better result for the classification. Partial least squares regression has been applied to the NIR spectra after MSC to propose calibration models that predict the d., crystallinity and m.p. of HDPE, LDPE and LLDPE. The correlation coeff. for the d. was calcd. to be 0.9898, 0.9928, 0.9925 and 0.9872 for the spectra obtained at 2, 4, 8 and 16 cm-1 resolns., resp., and the root mean square error of cross validation (RMSECV) was found to be 0.0021, 0.0018, 0.0018 and 0.0023 g cm-3, resp. It has been found that the correlation coeff. and RMSECV for the prediction of the d. and crystallinity change little with the spectral resoln. However, for the prediction of m.p., the higher resolns. (2 and 4 cm-1 resoln.) provide slightly better results than the lower resolns. NIR transmission spectra of thin films of LLDPE, LDPE and HDPE have also been investigated, and calibration models for predicting their d. have been developed for the film spectra.