In view of recently reported departures from bond additivity of chemical energies of formation, a purely empirical study has been made of the significance of chemical bond energies. If one assumes that the energy of formation is made up of bond energies and interaction energies between the bonds as well, it is shown that both of these two types of energy are indeterminate so far as thermochemical data are concerned. Under certain simple conditions on the interaction energies apparent bond additivity would obtain and one could then define apparent bond energies. In case these additivity conditions are not valid, it is further shown how the energy of formation can be expressed in terms of the apparent bond energies with a correction term involving the numerical departures from the additivity conditions. It is also shown that under a transformation of reference state to excited states of the atoms: (1) the additivity characteristics are invariant, and (2) the bond resonance energies as defined by Pauling are invariant, and therefore the reference of bond energies to excited states has no significance so far as these resonance energies are concerned. A discussion of atomic and bond additivity is given for physical quantities in general. It is pointed out that the energy connected with the nuclear motions, that is the heat content plus the zero‐point energies of the vibrations, may be of the order of one‐tenth the total energy of formation. Since this latter energy would not be expected on the whole to exhibit bond regularities, it is therefore proposed that ``electronic'' energies be introduced for the study of structural regularities in the energies of formation. A study is then made of available data in terms of the proposed scheme. Because of the lack of accurate data it is at present not possible to make a very thorough test. But a certain amount of information is obtained from the excellent data of Rossini on some of the CHO compounds, and so far as reliable data are available reasonable results are obtained, which indicate that in some cases the departures from additivity may be considerably greater than ordinarily supposed. It is further pointed out that Pauling's bond resonance energies could not accurately be considered as pure bond quantities if the interactions were appreciable, and that this may be the reason in part, that his electro‐negativity map cannot on the whole be used successfully to predict the electric moments of chemical bonds.