From the 1920s to the present, seismic imaging (“migration”) has helped the oil and gas industry locate hydrocarbon traps inside the Earth. Migration has evolved and improved over the years, and it is now used routinely for structural imaging, seismic velocity estimation, and amplitude analysis, among other applications. It is applied to narrow- and wide-azimuth towed-streamer marine data, to marine data with multicomponent sensors placed on the sea floor, to land data from desert and mountainous areas, to data acquired in transition zones, to sparsely and densely acquired data, and even to blended data. Migration is applied in many different kinds of settings, and it comes in many different shapes and sizes. Was migration in the 1920s anything like migration today? Perhaps surprisingly, the answer is yes. Even the most advanced current methods are based on principles that drove early innovators to exploit the nature of recorded seismic wavefields in mapping subsurface reflectors – sound waves bouncing off reflectors and sending localized packets of wiggle energy back toward recording devices. Of course, the details changed as, first, specialized machines replaced pencil and paper and, later, the digital revolution replaced analog devices and allowed the modeling of wavefields inside the computer. Some methods fell by the wayside and others emerged as computers became more powerful and allowed us to migrate more and more wiggles on more and more traces. Is migration finished, or even mature? Mature, yes, but by no means finished. We now often apply a complete two-way wave equation when migrating seismic data, but the wave equation we use is almost always an acoustic one applied to waves propagating in an elastic Earth. This itself is a big approximation, and it affects our ability to estimate velocity and analyze amplitudes. Sometimes migration does not use a complete acoustic equation – the equation might not admit two-way propagation, or it might use an approximate ray-based solution. In fact, depending on the seismic acquisition, migrated images using an “incomplete” wave equation are often preferable to more theoretically correct images. So there is still a lot of work to do, both in improving the fidelity of our high-end methods and in refining our lower-end methods to accommodate less-than-ideal acquisition.