Fourier-transform (FT) Raman spectra were measured for 7 kinds of different polypropylene copolymers and the assignments of all Raman bands were made through a detailed analysis. Ultimately, by choosing 2 700~3 100 cm-1 as the characteristic spectral, partial least squares (PLS) regression models were set up for xylene solubles content, ethylene content and ethylene content of xylene solubles in polypropylene copolymers. The values of correlation coefficient (r) between predicted results and actual values were all greater than 0.94, and the values of average relative error (ARE) were all less than 5%. The second PLS factor was also analyzed in this paper, and it was concluded that the original information of the sample was accurately extracted by using PLS regression. This paper offered a kind of possibility for on-line measurement of xylene solubles content, ethylene content and ethylene content of xylene solubles in polypropylene copolymers.