Convection of silicate melts in magma chambers is considered as a possible mechanism for producing significant or, at least, detectable chemical fractionation by the Soret effect. Thermal boundary layer analyses show that Soret fractionation would be, at best, an extremely weak process in evolving magmatic systems. For very large amplitude thermally driven convection at horizontal chamber margins, both the magnitude of the temperature gradient and the time scale for residence of magma in the unstable thermal boundary layer are entirely inappropriate for chemical fractionation of the magma at levels comparable to those obtained in laboratory experiments (5-60%). For a typical convecting body a relative concentration enhancement of 0.04% is obtained as an estimate of the upper limit of Soret fractionation. Even if exceedingly large temperature gradients (of the order of 104 ¿C/m) in a transient thermal regime are attained by compositionally driven convection, the magma residence time in the thermal boundary layer is so brief (100 s) that much less chemical fractionation results (0.002%). For convection near vertical margins a kinematic model of a countercurrent flow regime provides estimates of chemical separation produced by the thermogravitational fractionation mechanism. Incorporating a range of physical and chemical parameters that characterize magma chamber convection, steady state values of concentration enhancement are even smaller than for fractionation near horizontal boundaries.