In soil science, Visible and Near Infrared (VNIR) Diffuse Reflectance Spectroscopy is being used in an effort to develop proximal sensors to quantify soil inorganic carbon, clay content, clay mineralogy, and many other soil constituents. The common analysis techniques used in soil spectroscopy have included multiple regression, principal component analysis, partial least squares regression (PLS), and boosted regression trees. The technique of PLS has been most prominent in the literature; however, all these techniques are limited in their ability to promote interpretations of the wavebands and particular absorptions that are most important to the prediction models. A new algorithm to incorporate wavelet analysis into VNIR spectroscopy as a preprocessing tool is proposed in this study. The new algorithm was tested on two datasets. Both datasets were VNIR diffuse reflectance measurements made on air-dried ground soils. The resultant wavelet regression models were compared to conventional PLS models with respect to their prediction accuracy, number of regression parameters, and possibility of physical interpretations. The results showed that in the first dataset, the r2 and RMSE of the wavelet model for the soil total clay content were 0.83 and 57 g/kg, respectively, both of which were similar to those of the PLS model (using band-averaging and the partial least square regression). Additionally, because of the multi-resolution capability of the wavelet analysis, the wavelet model had the ability to separate fine and coarse spectral absorptions by distinguishing them into the wavelet regressors at different scales. This allowed the further physical interpretation of the wavelet model to relate wavelet regressors to the true spectral absorptions of soil constituents.