Native and synthetic vertebrate smooth muscle thin filaments have been examined by electron microscopy in order to determine the arrangement of the regulatory protein caldesmon. In synthetic filaments of actin-caldesmon, long slender molecules were sometimes seen running along the thin filament, suggesting that caldesmon can associate with actin along its length, while at other times lateral projections were observed. In native filaments, containing actin, caldesmon and tropomyosin, we found no evidence for lateral projections extending from the filaments, suggesting that caldesmon does not act as a crosslinking protein in vivo. In contrast, elongated molecules were clearly seen following the long pitch actin helices. We suggest that these may represent an association of caldesmon and tropomyosin. Antibodies developed against an N-terminal fragment of caldesmon caused thin filaments to aggregate laterally into arrays displaying approximately 35-38 nm repeats; thin filament aggregates with this periodicity were obtained previously (Lehman et al., 1989) using antibodies to the C-terminal segment of caldesmon. These results suggest that both ends of caldesmon are closely associated with the shaft of the thin filament, supporting a model in which the elongated caldesmon molecule runs along the filament, possibly interacting with tropomyosin, following the long pitch actin helices.