Radar is a system of measurement rather than communication, yet it is quite possible to apply information theory to it, in order to see whetherc the very small received signals inherently contain as much information as those of an ideal communication system working at the same signal-to-noise ratio (SNR). It turns out that they do, very nearly, but this is not r eally what makes radar a suitable topic for this symposium. The main interest is in the type of coding it represents. Shannon has pointed out that when the natural number of dimensions of a message is artificially increased by mapping non-topologically into a signal space of higher dimensions, a marked threshold effect is produced. Radar exhibits such a threshold particularly well and it is to this that I wish to direct attention. It is seen that there are in radar two quite different conditions of reception. There is ambiguous reception in which the information rate is high and the intelligibility low, and there is unambiguous reception in which the information rate is low but intelligibility is high. It would therefore appear that merely to evaluate quantities of information, and compare them with the ideal limits is not adequate guide to the behavior of a communication system.