An extensive data base which comprises the retention data of a total of 2106 peptides has been established and used to derive individual amino acid group retention coefficients. A multiple linear regression matrix approach was employed for solving the numerical value of the coefficients from the multivariate structure-retention dependencies. Statistical analysis of the retention data revealed that a minimum of 100 peptides is required to provide consistent values of the amino acid coefficients. Categorisation of all peptides allowed the influence of various chromatographic parameters on the coefficients to be evaluated. In particular, a decrease in the alkyl chain length of the chemically modified n-alkylsilica from octadecyl to butyl did not generally coincide with a decrease in the value of the group retention coefficients of individual amino acids. This study has established a detailed computational basis for characterising peptide retention behaviour and provides further insight into the mechanism of the interaction of peptides with immobilised hydrocarbonaceous ligands.