A convenient, preparative scale, nonlytic separation of mouse T lymphocytes into Lyt2.2+ and Lyt2.2- populations is reported. Immunoglobulin-negative (Ig-) spleen cells, Ig- lymph node cells, and peanut lectin-unagglutinated (PNA-) thymocytes were incubated under sterile conditions at 0 degree C with monoclonal mouse antibody to the Lyt2.2 T cell differentiation antigen. The antibody-treated cells were washed and placed in polystyrene tissue culture dishes that had been precoated with antibody to mouse Ig. Nonadherent populations were depleted to Lyt2.2+ cells and were essentially devoid of cytotoxic T lymphocyte precursors (CTLp), but contained helper activity for in vivo T-dependent IgM, IgG and IgA antibody formation. Adherent cell populations were enriched for Lyt2.2+ cells and for CTLp. The graft-vs.-host activity of the separated, adherent (Lyt2.2+) and nonadherent (Lyt2.2-) cells in the Simonsen spleen assay in neonatal (C57BL/6 x BALB/c)F1 mice was less than of unfractionated cells, but the activity of remixed Lyt2.2+ plus Lyt2.2- cells was higher than the sum of the contributions of these cells tested separately, and equal to that of the unfractionated cells. PNA- thymocytes were also separated into Ly2.2+ and Lyt2.2- populations by fluorescence-activated cell sorting. Nonlytic separation allows the recovery of the Lyt1+2+ population, which is lost in cytotoxic elimination experiments. Under the conditions described for the plate separation, the purity of the separated cells and recovery of activity approaches that of cells separated by sorting. Therefore, the plate separation offers a convenient alternative to fluorescence-activated cell sorting when large numbers (i.e. up to 5 x 10(7) positively selected cells) are needed, as in studies of in vivo cell-mediated immune reactions.