EINSTEIN's statistical theory of Brownian motion has two aspects, one dealing with distribution in time, the other with distribution in space. It is known1 that, while the spatial distribution is described, as in the chronology of a Geiger counter, by Poisson's enumerating law of independent rare events, the temporal distribution turns out to be determined by the classical theory of independent frequent errors. Precisely in view of the latter theory, and in view of the ergodic hypothesis, the spatial distribution requires a more specific model than does the temporal distribution; the former dealing with accidental fluctuations, the latter only with the underlying average state.