•  The length of myosin filaments was measured in three types of smooth muscle using serial electron microscopy. •  The frequency distribution of myosin filament length for all three types of smooth muscle followed an exponential decay pattern. •  The same frequency distribution pattern was observed in activated tracheal smooth muscle, although the average length was shorter compared with the filaments in relaxed smooth muscle. •  Analysis suggests that the distribution pattern reflects a dynamic equilibrium between competing processes of linear polymerization and de‐polymerization of myosin dimers. Abstract  Myosin molecules from smooth muscle and non‐muscle cells are known to self‐assemble into side‐polar filaments in vitro. However, the in situ mechanism of filament assembly is not clear and the question of whether there is a unique length for myosin filaments in smooth muscle is still under debate. In this study we measured the lengths of 16,587 myosin filaments in three types of smooth muscle cells using serial electron microscopy (EM). Sheep airway and pulmonary arterial smooth muscle as well as rabbit carotid arterial smooth muscle were fixed for EM and serial ultra‐thin (50–60 nm) sections were obtained. Myosin filaments were traced in consecutive sections to determine their lengths. The results indicate that there is not a single length for the myosin filaments; instead there is a wide variation in lengths. The plots of observation frequency versus myosin filament length follow an exponential decay pattern. Analysis suggests that in situ assembly of myosin filaments in smooth muscle is governed by random processes of linear polymerization and de‐polymerization, and that the dynamic equilibrium of these processes determines the observed length distribution.