The large differences in He and Ar diffusivities in silicate minerals could result in fractionation of the He/Ar ratio during melting of the mantle, producing He/Ar ratios in the primary mantle melts that are higher than those of the bulk mantle. Modeling noble gas diffusion out of the bulk mantle into fast diffusion pathways (such as fractures or melt channels) suggests that significant (order of magnitude) He/Ar fractionation will occur if the fast diffusion channels are spaced several meters apart and the noble gas residence in these diffusion channels is of the order days to weeks. In addition, the 15% difference in 3He and 4He diffusivities could also produce isotopic fractionation between the melt and its solid source. Modeling the behavior of He and Ar during melting shows that small increases (few %) in 3He/ 4He should be correlated with larger variations (factor of 5) in 4He/ 40Ar. However, in order to test this hypothesis the effects of subsequent He-Ar fractionation that occur during degassing have to be corrected. I describe a scheme that can separate He/Ar variations in the primary melt from overprinted fractionation during magmatic degassing. Using the degassing-corrected data, there is a correlation between the primary melt's 4He/ 40Ar and 3He/ 4He in mid-ocean ridge basalts (MORBs). The slope of the correlation is consistent with the models of preferential diffusion of 3He relative to 4He and of 4He relative to 40Ar from the solid mantle into the melt. Diffusive fractionation of noble gases during melting of the mantle can also account for low 4He/ 40Ar ratios commonly found in residual mantle xenoliths: preferential diffusion of He relative to Ar will produce some regions of the mantle with low 4He/ 40Ar, the complement of the high 4He/ 40Ar ratios in basalts. Diffusive fractionation cannot, however, account for differences between the He and Ne isotopic compositions of MORBs compared with ocean island basalts (OIBs); not only are the extremely high 3He/ 4He ratios of OIBs (up to 50 Ra) difficult to produce at reasonable mantle time and lengthscales, but also the Ne isotopic compositions of MORBs and OIBs do not lie on a single mass fractionation line, therefore cannot result from diffusive fractionation of a single mantle Ne source. If preferential diffusion of He from the solid mantle into primary melts is a significant process during generation of MORBs, then it is difficult to constrain the He concentration of the mantle: He concentrations in basalts and the He flux to the ocean essentially result from extraction of He from a larger (and unknown) volume of mantle than that that produced the basalts themselves. The He concentration of the mantle cannot be constrained until more accurate estimates of the diffusion contribution are available.