Single cell suspensions prepared from embryonic chick or rat spinal cords were separated into morphologically and functionally distinct subpopulation based on their buoyant densities The lightest fraction (F-1) was highly enriched for cells containing the enzyme choline acetyltransferase (CAT), a marker for developing motoneurons. The morphology biochemistry, and in vitro development of this and other spinal cord cell fractions isolated by the outlined procedure were investigated. Spinal cords, dissected from 6-day chick or 12-day rat embryos, were dissociated with trypsin and applied to iso-osmotic metrizamide density gradients. After brief centrifugation, biochemical analysis revealed that cholinergic cells migrated to lower densities than other spinal cord cells. The use of discontinuous density gradients allowed rapid and simple isolation of three fractions of viable cells (designated F-1 to F-3, lowest to highest density). Characterization of chicken and rat embryo cell fractions gave similar results. The cells in Fraction 1 were large with prominent nuclei and nucleoli, while those in F-2 and F-3 were distinctly smaller. Fraction 1 was highly enriched for cholinergic cells. The CAT specific activity (CAT/cell) was increased 400% in Fraction 1 compared to unfractionated cells, while CAT specific activity in F-2 and F-3 was reduced to 25% and less than 4% that of unfractionated cells, respectively. The recovery of cholinergic cells using this procedure was much better than with other published procedures; greater than half the spinal cord CAT activity was routinely recovered in the enriched fraction. The cholinergic-enriched cells (F-1) were unique in their in vitro growth characteristics. All fractions had neuronal cells, while non-neuronal cells were distributed primarily in F-3, fewer in F-2, and were essentially absent from F-1. Neurons in F-2 and F-3 remained viable under a variety of conditions, most of which were not supportive of F-1 cell survival. The cholinergic-enriched F-1 cells survived and developed only in the presence of muscle cells or in muscle-conditioned medium on highly adhesive substrata. Large, multipolar neurons predominated under these conditions. The method described provides a means of characterizing the factors involved in the development of distinct populations of cells from the embryonic spinal cord.