Gas-particle partitioning is examined using a partitioning constant K(p) = (F/TSP)/A, where F (ng m-3) and A (ng m-3) are the particulate-associated and gas-phase concentrations, respectively, and TSP is the total suspended particulate matter level (mu-g m-3). At a given temperature and for a given sample of particulate matter, compound-dependent values of K(p) tend to be correlated with the sub-cooled liquid vapor pressure (p(L)0, torr) according to log K(p) = m(r)log p(L)0 + b(r). Theory predicts that b(r) values should be somewhat similar, and that m(r) values should be near -1. This is supported by field and laboratory work. However, there is still noticeable variability in reported m(r) and b(r) values, even when obtained by the same researchers sampling in the same location. Three possible thermodynamic sources of variability include variability in the compound-to-compound differences in the thermodynamics of adsorption, event-to-event variability in the specific surface area of the aerosol and event-to-event variability in the ambient temperature. Non-thermodynamic sources of variability include sorption of gaseous analytes to the filters used in differentiating between F and A, the presence of non-exchangeable components in the measured F values, within-event adsorption/desorption kinetics, within-event changes in contaminant levels, and within-event changes in temperature. Each of these sources of variability operate in their own way to cause variability in m(r) and b(r). In general, one can expect there to be a correlation in the observed m(r) and b(r) of the form b(r) = m(s)m(r) + b(s). For the study of Yamasaki et al. (1982, Envir. Sci. Technol. 16, 189-194), one obtains m(s) = 5.77 and b(s) = -2.18, with r2 = 0.91. In the presence of such a correlation, one can expect that all log( F/TSP)/A vs log p(L)0 plots will tend to intersect at the same (x, y) point given by (-m(s), b(s)). Existing field and laboratory data show this tendency.