When BHK21 fibroblasts adhering to glass were incubated in trypsin they became spherical within a few minutes. They did not, however, respond simultaneously; the most trypsin-resistant cells in an unsynchronized population were of greater length, had more processes projecting from the cell body, and were more spread out than the most trypsin sensitive cells. In no instance did trypsin detach cells from their substrate, even when the incubation period was prolonged to 5 h in trypsin concentrations almost sufficient to cause cell lysis. Scanning electron-microscope observations showed that initially flat cells rounded up in trypsin to reveal persistent adhesion sites joined to the cell body by retraction fibres. Such cell could be dislodged only by agitation; when this occurred parts of the cell remained attached to the substrate in the form of small spheres and fibres; these were remnants of the retraction fibres and adhesion sites. We propose that the adhesion sites are not susceptible to proteolytic degradation, presumably because of steric hindrance, and that this causes detachment to be dependent upon mechanical dislocation. We suggest that some of the descriptions of substrate-associated 'cell exudates' in the literature may refer to these cell remnants on the substrate which consist of membrane-bound fragments of cytoplasm rather than secreted products.