We have studied submicrosecond and microsecond rotational motions within the contractile protein myosin by observing the time-resolved anisotropy of both absorption and emission from the long-lived triplet state of eosin-5-iodoacetamide covalently bound to a specific site on the myosin head. These results, reporting anisotropy data up to 50 microseconds after excitation, extend by two orders of magnitude the time range of data on time-resolved site-specific probe motion in myosin. Optical and enzymatic analyses of the labeled myosin and its chymotryptic digests show that more than 95% of the probe is specifically attached to sulfhydryl-1 (SH1) on the myosin head. In a solution of labeled subfragment-1 (S-1) at 4 °C, absorption anisotropy at 0.1 μs after a laser pulse is about 0.27. This anisotropy decays exponentially with a rotational correlation time of 210 ns, in good agreement with the theoretical prediction for end-over-end tumbling of S-1, and with times determined previously by fluorescence and electron paramagnetic resonance. In aqueous glycerol solutions, this correlation time is proportional to viscosity/temperature in the microsecond time range. Furthermore, binding to actin greatly restricts probe motion. Thus the bound eosin is a reliable probe of myosin-head rotational motion in the submicrosecond and microsecond time ranges. Our submicrosecond data for myosin monomers (correlation time 400 ns) also agree with previous results using other techniques, but we also detect a previously unresolvable slower decay component (correlation time 2.6 μs), indicating that the faster motions are restricted in amplitude. This restriction is not consistent with the commonly accepted free-swivel model of S-1 attachment in myosin. In synthetic thick filaments of myosin, both fast (700 ns) and slow (5 μs) components of anisotropy decay are observed. In contrast to the data for monomers, the anisotropy of filaments has a substantial residual component (26% of the initial anisotropy) that does not decay to zero even at times as long as 50 μs, implying significant restriction in overall rotational amplitude. This result is consistent with motion restricted to a cone half-angle of about 50 °. The combined results are consistent with a model in which myosin has two principal sites of segmental flexibility, one giving rise to submicrosecond motions (possibly corresponding to the junction between S-1 and S-2) and the other giving rise to microsecond motions (possibly corresponding to the junction between S-2 and light meromyosin). Both motions are slower and more restricted in filaments than in monomers. This study complements previous fluorescence and electron paramagnetic resonance studies by extending the time range and providing time-resolution, respectively, thus providing a more complete description of myosin rotational dynamics.