The possible routes of transvascular migration of leukemic cells in the liver were studied in guinea pigs with an L2C lymphoblastic cell-line inoculation leukemia. Invasion of the hepatic parenchyma theoretically can occur in three ways: Through the intact sinusoidal endothelium, utilizing either preexistent gaps (normal in the liver), or newly created pores, whether interendothelial or intraendothelial. We could not convincingly demonstrate this, but could not wholly exclude this either. After destruction or retraction of the endothelium, either on account of the remarkable sinusoidal engorgement and distension by masses of leukemic cells, or by direct assault on the endothelium by the leukemic cells. We can clearly demonstrate the former, and hold it to be the major cause of hepatic infiltration. Evidence for a direct endotheliolytic effect was not uncovered in our studies. Secondary infiltration from the portal triads. Heavy leukemic infiltration of the triads, whether from the portal or hepatic veins, or from the lymphatics, is indeed and early an consistent feature--but the infiltration of the hepatic lobule shows no peripheral, or any other zonal preference. In both portal and hepatic veins, leukemic cells transverse the endothelium through a cytoplasmic "pore", adjacent to cell junctions, without obvious damage to the endothelium.