The metabolism of the small pool of triglyceride in human high-density lipoproteins (HDL) has been studied. In 28 postabsorptive and unmedicated subjects the previously reported reciprocal relationships between the concentration of very low-density lipoprotein (VLDL) lipids and HDL cholesterol was confirmed. It was also found that the HDL triglyceride correlated positively and significantly with the concentration of VLDL triglyceride, and there was a significant negative correlation between the concentrations of triglyceride and cholesterol in HDL. Five normal and 3 hypertriglyceridemic subjects were given 5,000 units of heparin intravenously. Not only was there a pronounced reduction in VLDL triglyceride, but there was a 60 to 80 per cent reduction in HDL triglyceride; the low-density lipoprotein (LDL) triglyceride was unchanged. To test a possible precursor-product relationship between the triglyceride pools in VLDL and HDL, 5 subjects were given an intravenous pulse injection of (14-C)-palmitic acid. The specific activity-time curves of triglyceride in VLDL and HDL did not conform to a typical precursor-product relationship; rather the HDL curve reached its peak well before it crossed the VLDL curve, quite compatible with an origin of HDL triglyceride which was independent of VLDL. However, when the lipoproteins of density smaller than 1.006 (VLDL and chylomicrons) were labeled in the intestine after an oral tracer dose of (3-H)-palmitic acid in 2 subjects, the relationship of the triglyceride labeling in HDL to that in d smaller than 1.006 lipoproteins was very similar to the relationship observed after the intravenous dose. Although an independent intestinal origin of HDL triglyceride could not be excluded, this result did suggest that the labeling of triglyceride in d smaller than 1.006 lipoproteins and in HDL was related and was consistent with a transfer from the larger lipoproteins to the HDL. In three subjects given a constant intravenous infusion of (3-H)-palmitic acid, the specific activity of triglyceride in VLDL became constant after 3 to 4 hours; it also became constant in the HDL after about 4 hours but at a level of only 40 to 70 per cent of the VLDL level. It appeared that 30 to 60 per cent of the HDL triglyceride was either derived from an unlabeled source or else turned over too slowly to become labeled to any extent during the infusions. Whatever the explanation, the same phenomenon would be likely to influence the level of the peak specific activity of HDL triglyceride after the pulse intravenous injections of label. It is possible that the specific activity of the "labeled" fraction of HDL triglyceride may have reached its peak at the point at which it crossed the VLDL curve, which would be consistent with a precursor-product relationship between VLDL and one fraction of the HDL triglyceride. It has been concluded that HDL triglyceride is heterogeneous and includes at least one pool, possibly transferred from VLDL, which is relatively rapidly turning over.