Cell clones were isolated from a population of cultured tomato (Lyeopersicon esculentum Mill cv VFNT-cherry) cells and their tolerance to polyethylene glycol (PEG)-induced water stress was measured. Considerable variation for tolerance among the clones was found. Tolerance differences between clones appeared to be spontaneous and were different from tolerance differences between adapted and unadapted cells. Unlike adapted (selected by exposure to PEG) cells, cell clones retained their relative tolerance for many generations in the absence of selection pressure, and tolerance of both relatively tolerant and intolerant clones was very dependent on growth cycle stage and inoculum density. Analysis of subelones isolated from relatively tolerant and intolerant parent clones revealed that each parent clone gives rise to progeny with tolerances near the mean tolerance of both parents. However, progeny populations of both tolerant and intolerant parents are enriched with individuals with phenotypes nearer the mean response of their respective parent populations. When exposed to PEG, relatively tolerant and intolerant clones alike become adapted to the level of PEG to which they are exposed, and have the same phenotypic level of tolerance. Thus, selection by exposure to stress is unable to discriminate (on the basis of growth) between the innately tolerant and intolerant cell types within the population. This is indicated also by the fact that clones isolated from a population of cells adjusted to growth on 25% PEG do not show an enriched frequency of tolerant phenotypes when grown in the absence of PEG compared to the nonselected normal cell population which has never been adjusted to growth on PEG.The use of plant cell cultures to obtain variants much akin to the selection of variants in bacterial cultures has been proposed and discussed in the literature for some time. In fact, many plant cell variants have been selected and these efforts have been reviewed (6, 12). The possibility of using plant cell cultures to obtain variants which would have practical application to agriculture is receiving increased attention. Among such agriculturally useful variants are stress-tolerant phenotypes which might be selected directly from cultured cell populations exhibiting spontaneous or induced variation, or might be obtained by the introduction of specific genetic information through genetic recombination technology and subsequent plant regeneration.In spite ofthe relative success ofselection attempts with cultured higher plant cells, there is little information concerning the genetic basis for much of the observed variation. Although it was assumed earlier that variation among cultured plant cells was due largely