Multivariate analysis of soil geochemistry is a powerful tool for differentiating lithological units and detecting geochemical dispersion halos related to mineralization or contamination. While univariate analysis can effectively identify lithological units with pronounced variations, it may fail to differentiate between subtler variations in lithologies. Traditional multivariate techniques such as principal component analysis (PCA) have limitations, including difficulties in understanding the individual contributions of each variable and an inability to work with non-Gaussian data. Independent component analysis (ICA) has emerged as a potential alternative, as it can effectively identify independent components of non-Gaussian data. In this study, we compared the effectiveness of PCA and ICA in relating multivariate soil geochemistry to parent lithology using the Soil Geochemical Atlas of Cyprus and associated digital geological maps. Both PCA and ICA were able to differentiate between the ultramafic units within the Troodos Ophiolite (TO) and the Circum-Troodos Sedimentary Succession (CTSS). However, ICA was more effective than PCA in identifying pillow lavas, providing a clear separation in the scores for IC4 and IC5. Furthermore, both PCA and ICA were able to separate the sheeted dykes from the cumulate mafic units within the TO. The gabbro unit is closely defined by IC2 scores. In contrast, PCA failed to provide factors that effectively delineated the Mamonia Terrane from other units, especially the TO, while ICA was able to provide a distinct separation in IC4 and IC5 scores. Separation between the CTSS and Quaternary units was weakly observed in IC2 scores. These findings demonstrate that there is a difference in the effectiveness of PCA and ICA in identifying different lithological units and emphasize the need for a careful selection of multivariate methods to differentiate between subtle differences in soil geochemistry relating to variations in parent lithology.