Three case studies comparing molecular ideal gas heat capacity estimations using empirical group additive methods and statistical mechanical methods using ab initio vibrational frequencies are reported. In the first study, results from several ab initio levels of theory are compared with the experimental heat capacities of piperidine and the heat capacities calculated from the observed vibrational frequencies. In the second study, heat capacities for a benchmark group of 27 organic molecules were calculated using vibrational frequencies from AM1, HF/3-21G(d), and B3LYP/6-31G(d) theory level, and the results were compared to three recently updated additivity schemes. In the third study, semiempirical corrections to the heat capacities of n-alkanes were investigated. The strength of additivity schemes is that they are easy to use and understand, and often give reasonable results, but predict the same thermodynamic properties for isomers containing the same functional groups, and are of limited accuracy when the near-neighbor interactions are strong or if the input data set is not appropriate. Ab initio vibrational frequencies can rapidly provide accurate heat capacities using the harmonic oscillator model, especially if one conformer dominates. The results can be comparable to additivity, and better for rigid molecules containing heteroatoms. Furthermore, ab initio results do not require a calibration data set. (C) 2004 American Institute of Chemical Engineers.