Radar is a uniquely powerful source of information about near-Earth asteroid (NEA) physical properties and orbits. Measurements of the distribution of echo power in time delay (range) and Doppler frequency (radial velocity) constitute two-dimensional images that can provide spatial resolution finer than a decameter. The best radar images reveal geologic details, including craters and blocks. Radar wavelengths (13 cm at Arecibo, 3.5 cm at Goldstone) are sensitive to the bulk density (a joint function of mineralogy and porosity) and the degree of decimeter-scale structural complexity of the uppermost meter or so of the surface. Radar can determine the masses of binary NEAs via Kepler's third law and of solitary NEAs via measurement of the Yarkovsky acceleration. With adequate orientational coverage, a sequence of images can be used to construct a three-dimensional model, to define the rotation state, to determine the distribution of radar surface properties, and to constrain the internal density distribution. As of mid 2006, radar has detected echoes from 193 NEAs, of which 107 are designated Potentially Hazardous Asteroids. Radar has revealed both stony and metallic objects, principal-axis and non-principal-axis rotators, smooth and extremely rough surfaces, objects that appear to be monolithic fragments and objects likely to be nearly strengthless gravitational aggregates, spheroids and highly elongated shapes, contact-binary shapes, and binary systems. Radar can add centuries to the interval over which close Earth approaches can accurately be predicted, significantly refining collision probability estimates compared to those based on optical astrometry alone. If a small body is on course for a collision with Earth in this century, delay-Doppler radar echoes could almost immediately let us recognize this by distinguishing between an impact trajectory and a near miss, and would dramatically reduce the difficulty and cost of any effort to prevent the collision.