Near-Infrared (785 nm) excitation was used to obtain Raman spectra from a series of 33 solid mixtures containing cocaine, caffeine, and glucose (9.8-80.6 % by weight cocaine) which were then analysed using chemometric methods. Principal Component Analysis of the data was employed to ascertain what factors influenced the spectral variation across the concentration range. It was found that 98 % of the spectral variation was accounted for by three principal components. Analysis of the score and loadings plots for these components showed that the samples can be clearly classified on the basis of cocaine concentration. Discrimination on the basis of caffeine and glucose concentrations was also possible. Quantitative calibration models were generated using Partial Least Squares (PLS) algorithms which predicted the concentration of cocaine in the solid mixtures containing caffeine and glucose from the Raman spectrum with a root mean standard error of prediction (RMSEP) of 4.1 %. Caffeine and glucose concentrations were estimated with RMSEPs of 5.2 % and 6.6 % respectively. These measurements demonstrate the feasibility of using near-IR Raman spectroscopy for rapid quantitative characterisation of illegal narcotics.