A systematic search for computational strategies able to accurately predict pKa values of protonated amines (HA+) has been performed. Thermodynamic cycles have been used in conjunction with the cluster‐continuum model, including up to eight explicit water molecules per system. The calculations have been performed with nine different methods within the density functional theory (DFT), and with MP2. It was found that including one explicit water molecule in the vicinity of the protonated site is highly important for obtaining accurate results. We recommend the F2 reaction scheme [HA+(H2O) + 3H2O = A + H3O+(3H2O)], combined with M05‐2X/6‐311++G(d,p) calculations. This particular combination produces mean unsigned errors (MUE) equal to 0.54 for all the tested amines, and 0.51 pKa units for the studied neurotransmitters. Moreover it was found to be the only one that systematically leads to errors lower than, or very close to, 1 unit of pKa for every one of the studied species, with the maximum error equal to 1.09 pKa units. This scheme has the additional advantages of being computationally feasible for amines of relative large size, independent of experimental values, and free of fitting‐based corrections. In addition the high influence of the reference acid of choice, in the proton exchange scheme (D), on the quality of the pKa values of amines is discussed. © 2012 Wiley Periodicals, Inc.