We have devised a method for the isolation of viable neuronal growth cones from neonatal rat forebrain. The method involves differential and density gradient centrifugation and exploits the relatively low buoyant density (approximately 1.018 g/cm3) of growth cones. There are no known biochemical markers for growth cones and it was necessary therefore to monitor for their presence during the isolation using transmission electron microscopy. Several criteria were used to identify isolated growth cones including the presence of filopodia, an extensive system of branching, tubular smooth endoplasmic reticulum and a region rich in microfilaments subjacent to the plasma membrane. These morphological features are similar to those of growth cones identified unequivocally in intact developing brain and in tissue culture. Electron microscopical analysis showed that greater than 90% of membrane-bound, identifiable objects in one fraction were growth cones by these criteria. The major contaminant consisted of membrane sacs and vesicles of unidentified origin. There were only small amounts of isolated rough endoplasmic reticulum and mitochondria. Isolated growth cones were roughly spherical in shape with a diameter of 1.9 +/- 0.5 micron (mean +/- 1 SD). They usually contained mitochondria, large granular vesicles and small vesicles, and occasionally contained coated vesicles, lysosomes, lamellar bodies and multivesicular bodies, and only very rarely, intermediate filaments. Occasionally, growth cones had rudimentary synapses on them. The viability of isolated growth cones was investigated by observing their behaviour in short-term culture. After a few hours in culture on poly-D-lysine-coated coverslips, growth cones flattened down and extended filopodia-like processes. This behaviour was inhibited by cytochalasin B and reversibly by cold (4 degrees C). We conclude that physiologically active growth cones can be isolated rapidly and in large numbers by the method described here.