Methanol, ethanol, and propanol have very similar molecular structures, which result in similar near-infrared (NIR) absorption spectra. The similarity of these spectral features challenges the ability to make selective quantitative measurements for one alcohol in the presence of the others. This work explores the issue of measurement selectivity for alcohols based on multivariate analysis of near-infrared spectra. The net analyte signal (NAS) corresponds to the component of the analyte spectrum that is orthogonal to all other sources of spectral variance. The NAS concept can be used to generate a regression vector that is selective for the principal analyte. The extent to which the NAS regression vector is orthogonal to the interferences can be determined by a pure component selectivity analysis(PCSA). In this analysis, inner products between a series of pure component spectra and the NAS regression vector are analyzed. Results indicate that each alcohol can be measured selectively. Partial least squares (PLS) analysis can also generate a selectivity regression vector for each alcohol. In this work, spectra from 79 samples were collected over the first overtone region of the near infrared spectrum. The optimized PLS calibration models provide selective measurements with prediction errors of 5.3, 1.9, and 2.4 mM for methanol, ethanol and propanol, respectively, over concentration ranges of 0 – 500 mM. The mean percent errors for each calibration model are 8.0, 1.0, and 2.7 %, respectively. Comparison between the PLS and NAS regression vectors confirms the selectivity of these measurements.