Abstract <p>Due to a high sensitivity to structural changes, the viscosity of a melt is often measured as an indirect method for studying the structural state of a liquid alloy. The temperature dependence of the viscosity of a metallic melt often exhibits anomalous behavior in the form of jumps, breaks, changes in the sign of the temperature coefficient or its derivative, a hysteresis of heating/cooling polytherms, and so on. These anomalies are usually related to temperature-induced changes in the structure of a melt, but there is no consensus on their nature. This is due to the contradictory data obtained by different authors. In this work, we discuss the nature and possible mechanisms of the anomalies observed by us on studying the temperature dependence of the viscosity of a metallic melt using the oscillating crucible method. In some cases, viscosity polytherm anomalies can be of a methodological nature, which is reflected in the dependence of the obtained results on experimental conditions. Methodological anomalies include the specific features of polytherms that are caused by the influence of film effects, wetting phenomena, and sedimentation of a crystalline phase in the mushy zone on measurement results. It is important to detect and exclude these anomalies, since they promote the formation of erroneous viewpoints on the structure of metallic melts and the nature of the phenomena detected in them. In addition, the specific features of the temperature dependence of the viscosity of liquid aluminum and the melts based on it with small additions of nickel, cobalt, and iron have been revealed as deviations of polytherms from the Arrhenius equation, and they are interpreted as liquid–liquid transitions. Aluminum-based melts alloyed with transition metal exhibit specific features of the temperature and time dependences of viscosity, which are caused by long relaxation processes in a melt after the crystal–liquid phase transition.</p>