This study was undertaken to test the ability of the molecular connectivity model to predict retention indices using both statistical correlation coefficients and correctly predicted elution sequences as criteria of fit. The test was performed on three groups of chloroalkanes. Regression analyses show that the molecular connectivity model successfully predicts the retention indices of chlorinated alkanes on polar and non-polar stationary phases. However, first-order molecular connectivity indices alone are not sufficient, higher order indices are demonstrated to be necessary. The results also indicate that different structural features determine the retention index values of mono- and dichlorides. For monochlorides the major factor is the size of the alkyl chain, while for dichlorides the major factor is the topological relation between the two chlorine atoms. The comparison of the results obtained with the molecular connectivity model and the empirical additive scheme reveals several important advantages of the molecular connectivity approach.