AbstractMultivariational analyses were performed to separate the contribution of electronic and steric effects of alkyl groups on the experimental hydrocarbon acidities of four different sets of compounds, consisting of hydrocarbons (SET‐I), acetophenone derivatives (SET‐II), and acetone CH3COCHR1R2 derivatives, where the acidity of the protons on the alkyl substituted site, (SET‐III), and the acidity of the methyl protons are both investigated (SET‐IV). Significant regression was obtained upon backward elimination for the first and last set. Due to the colinearity of the parameters in SETs‐II and ‐III the significance of the correlation is not so high. As expected, the gas phase acidity of the hydrocarbons in SET‐I is very sensitive to the polarizability and electronegativity of the alkyl substituents. The substituent have similar effects on the solution acidities of the ketone derivatives in SETs‐II, ‐III, though suppressed due to solvation, unlike the cases in SET‐II and SET‐III, the effects of the substituents in SET‐IV, where the substituents are relatively far away from the carbanion center generated are strongly reduced and reversed. This different behavior may be attributed to the domination of the solvent effects. Theoretical acidities of these compounds were calculated using the semiempirical methods such as MNDO, AM1 and PM3. The results are, quantitatively and qualitatively, unsatisfactory for the gas phase, as well as for the solution acidities. Multivariational analysis of substituent effects shows the insufficiency of the present parametrization to reflect the substituent effect properly. It seems that all these semiempirical methods are useless in prediction of carbon acidities, without reparametrization using carefully selected sets of compounds.