1. Isolated rat liver mitochondria do not synthesize labelled phosphatidylcholine from CDP-[(14)C]choline or any phospholipid other than phosphatidic acid from [(32)P]phosphate. The minimal labelling of phosphatidylcholine and other phosphoglycerides can be attributed to microsomal contamination. However, when mitochondria and microsomes are incubated together with [(32)P]phosphate, the phosphatidylcholine, phosphatidylinositol and phosphatidylethanolamine of the reisolated mitochondria become labelled, suggesting a transfer of phospholipids between the two fractions. 2. When liver microsomes or mitochondria containing labelled phosphatidylcholine are independently incubated with the opposite un-labelled fraction, there is a substantial and rapid exchange of the phospholipid between the two membranes. Exchange of phosphatidylinositol also occurs rapidly, whereas phosphatidylethanolamine and phosphatidic acid exchange only slowly. There is no corresponding transfer of marker enzymes. The transfer of phosphatidylcholine does not occur at 0 degrees , and there is no requirement for added substrate, ATP or Mg(2+), but the omission of a heat-labile supernatant fraction markedly decreases the exchange. 3. After intravenous injection of [(32)P]phosphate, short-period labelling experiments of the individual phospholipids of rat liver microsomes and mitochondria in vivo give no evidence for a similar exchange process. However, the incubation of isolated microsomes and mitochondria with [(32)P]phosphate also fails on reisolation of the fractions to demonstrate a precursor-product relationship between the individual phospholipids of the two membranes. 4. The intraperitoneal injection of [(32)P]phosphate results in a far greater proportion of the dose entering the liver than does intravenous administration. After intraperitoneal administration of [(32)P]phosphate the specific radioactivities of the individual phospholipids are in the order microsomes > outer mitochondrial membrane > inner mitochondrial membrane. 5. The incorporation of (32)P into cardiolipin is very slow both in vivo and in vitro. After labelling in vivo the radioactivity in the cardiolipin persists compared with that of the other phospholipids, whose specific radioactivities in the microsomes and mitochondrial fragments decay at a similar rate to that of the acid-soluble phosphate pool. 6. The possibility of phospholipid exchange processes occurring in the liver cell in vivo is discussed, and it is suggested that only a small but highly labelled part of the endoplasmic-reticulum lipoprotein pool is involved in the transfer.