We have studied the influence of ATP, inorganic monophosphate, MgCl 2 and CaCl 2, alone or in combination, on the formation of synthetic myosin filaments by means of electron microscopy. Both crude and column-purified rat skeletal myosin were studied systematically, and in some instances parallel experiments were carried out using crude rabbit skeletal myosin. The behaviour of filaments formed by a standard polymerization procedure at pH 6.8, in the absence of ATP or inorganic phosphate, is not influenced by MgCl 2, CaCl 2 or ethylenediaminetetraacetic acid at concentrations up to 5 m m. Such filaments are homogeneously 30 to 50 nm wide and 5 to 15 μm. long, i.e. larger than physiological size. They do not systematically display tapered ends. Filaments are built up from 2 to 3-nm wide threadlike subunits, arranged roughly parallel to the long axes. Sometimes filamentous projections up to 60 nm long, not associated with accidentical filament bending, are seen. In heavily contrasted preparations these projections are replaced by irregular globular structures. This transformation of the projections is accompanied by a slight decrease in the diameter of the filament shaft. When polymerization is carried out in the presence of millimolar amounts of ATP or inorganic phosphate, but in the absence of divalent cations, regular filaments are not formed. Instead long branched structures, or small twig-like filaments are obtained, which are 10 to 15 nm wide and 0.2 to 0.4 μm long. The addition of ATP or inorganic phosphate to preformed regular filaments does not bring about this structural disorder. The presence of millimolar amounts of MgCl 2 or CaCl 2 in the polymerization medium efficiently counteracts the disorganizing effect of ATP and inorganic phosphate. The size of filaments formed under these circumstances critically depends upon the nature of the divalent cation. CaCl 2 induces the formation of filaments similar in every respect to those already described. In the presence of MgCl 2 distinctly thinner ones, with close to physiological diameters (15 to 17 nm), are obtained. In both cases the lengths are still 5 to 15 μm. The filaments with physiological diameters consistently display tapered ends. They are built up from the same threadlike subunits as wider ones, and most display filamentous projections up to 60 nm long, systematically pointing in the direction of the filament ends. Central zones of polarity reversal are easily identified. In heavily contrasted preparations these filamentous projections are again replaced by irregular globular structures and the apparent diameter of the shaft is slightly diminished. In no case did we notice any significant influence of the origin, or degree of purity, of the myosin on filament behaviour.