Lymphokine-activated killer (LAK) cells can lyse a number of tumor target cells regardless of whether the tumors are natural killer (NK) sensitive or resistant. LAK can also lyse autologous lymphoblasts that have been modified with 2,4,6-trinitrobenzene sulfonic acid (TNBS). In this study, we examined the surface markers of murine LAK precursors. It was found that depletion of Thy 1- or Lyt 2-bearing precursor cells abolished the ability of spleen cells to generate LAK against TNBS-self, but had no effect on the generation of LAK against tumor cells. Depletion of asialo-GM1 (AGM1)-bearing precursors abolished the generation of LAK against all target cells tested. Normal spleen cells were fractionated on a Percoll density gradient and two fractions were examined: fraction (Fxn) 3, which is enriched for NK activity but depleted of the ability to generate cytotoxic T lymphocytes (CTL), and Fxn 5, which had no NK activity but was enriched for the ability to generate CTL. Both fractions were capable of generating LAK, although Fxn 5 required a relatively larger amount of interleukin 2 (IL 2). Upon examination of the surface markers of LAK precursors in these fractions it was found that the precursors in Fxn 3 giving rise to LAK against tumors were Thy-1-, Lyt-2-, AGM1+, whereas the precursors in Fxn 5 were Thy-1+, Lyt-2+, AGM1+. The precursors generating LAK against TNBS-self were Thy-1+, Lyt-2+, AGM1+ in both fractions. The time kinetics of LAK generation in both fractions were different, with Fxn 3 showing much earlier kinetics. These data delineate at least two different LAK precursors defined by their buoyant density, by their surface markers, and by their susceptible target cells. These data also may resolve the confusion in the literature regarding the phenotype of LAK precursors.