Nucleation fields for the reversal of magnetization in thin iron‐nickel films were experimentally measured. The reversal was observed by the Kerr magneto‐optical effect. Different regions or types of imperfections were observed at which lower nucleation fields are required than for the more perfect regions. Small regions of study in the ``interior'' of the film were localized by using an inhomogeneous field a millimeter or less in extension. Nuclei formed reproducibly about a center, with a statistical probability of formation which depended on the pulse length and field strength. A mode value of nucleation can be associated with a given center. The average of the mode values for different centers increased somewhat with film thickness, and showed a marked dependence upon composition, with a minimum value for films containing 80–85% nickel.Repeated pulsing with fields smaller than those required for nucleation increased the size of nuclei, while maintaining the same lenticular shape, to limiting dimensions which increased with film thickness. At higher fields, wall motion was observed in detail for two types of motion: (1) the tip of the domain was displaced along the easy axis with an axial or ``endwise'' extension; (2) the ``180° wall'' was displaced along the hard axis with a normal or ``sidewise'' extension. Endwise extension takes place at lower fields than sidewise extension. Evidence was obtained to indicate that an energy barrier exists for sidewise motion, e.g., repeated 1.5‐&mgr;sec pulses caused no motion, whereas a single 20‐&mgr;sec pulse did. With very narrow drive wires or with thick films, additional nucleation took place instead of ``sidewise'' motion of the wall; otherwise, domains could be enlarged at will. Some films were produced which required larger fields to ``erase'' or contract the domain lengthwise than for tip extension.The instabilities and general morphology of the domains as they depend on composition, film thickness, distance of extension of the inhomogeneous field along the easy axis, and the magnitude of the drive field were studied. Certain configurations led to a morphology of many small nuclei, which eventually were dense enough to join into a single domain; other configurations showed domain growth by wall motion and a large domain formed from very few nuclei. Very small nuclei were generally unstable, disappearing in a short time after formation. The smallest stable nucleus observed was about 0.25 mm×0.04 mm. The transition in behavior between 1.0‐ and 0.5‐&mgr;sec pulses stresses the importance of multiple‐nucleation processes in the nonuniform rotation region. The finescale observations on fields for nucleation and wall motion make possible comparisons with the processes taking place when the complete specimen is switched in a B‐H loop tester.