After a brief review of recent work on melt inclusions in diamond and in normal magmatic environments, the nature of the fluid-inclusion evidence for various stages of magmatic immiscibility is summarized. During magmatic differentiation by crystal fractionation, from originally low-volatile content silicate melts (yielding normal plutonic rocks), through hydrous silicate melts (yielding pegmatites), to late-stage water-rich fluids (yielding quartz veins and ore deposits), the composition of the residual liquid changes drastically. In some geologic environments, this change in composition of the residual liquid may be continuous, with no phase change in the fluid part of the system. Under most natural conditions, however, it is far more likely to have one or more stages in the evolution during which globules of a separate, new immiscible fluid phase exsolve. The term immiscibility is used here in its more general sense to refer to the existence, at equilibrium, of two or more non-crystalline polycomponent solutions (fluids), differing in properties and generally in composition. I believe that the great bulk of the magmatic inclusions that are being studied today have come about through the intervention of one or more such stages of immiscible fluid separation, i.e., magmatic differentiation by fluid immiscibility. Examples include silicate melt/dense CO 2 fluid, silicate melt/sulfide melt, silicate melt/silicate melt, silicate melt/low density vapor (i.e., vesiculation), and, as presented in more detail in this paper, various stages of silicate melt/hydrous saline melt/aqueous fluid/CO 2 fluid immiscibility. Generally, only parts of these stages are recorded by the trapping of fluid inclusions, and the interpretation of the inclusion record is commonly ambiguous and difficult, particularly that stage between hydrous silicate melts and the hydrous saline melts (and aqueous solutions) responsible for many ore deposits. The ambiguity results mainly from problems of inclusion origin and nonrepresentative trapping within any given individual inclusion. Thus the probability of trapping of only one, vs. various amounts of two, immiscible fluids, or of trapping solid phases along with the fluid, present severe difficulties. In addition, there is the almost universal superposition of various stages of trapping of fluids within any given sample. These problems can only be approached by very careful petrography and analysis of optimum material, combined with field data and experimental P-V-T-X studies of pertinent systems. Completely unequivocal statements of the course of evolution of the fluids in such natural systems may never be possible.