The Kolmogorov model for turbulence in the atmosphere is used to predict the power spectrum of phase fluctuations measured by a radio interferometer. Spectral density plots, where the phase fluctuation power is proportional to the area under the curve, are shown to be very useful for appreciating the relevant timescales, and there are straightforward scalings to different baseline length and airmass. The model is extended to include the thickness of the turbulent layer and the orientation of the baseline with respect to the wind direction.
The data examined so far are in broad agreement with the model. There
can be significant phase variations on timescales as long as one hour.
There may also be more than one component of turbulence contributing to
the phase fluctuations. There is no clear evidence for an outer scale
of turbulence on scales less than 
km, in contrast to optical
observations.
The author would like to thank Rachel Akeson, John Carlstrom, Peter Papadopoulos and David Woody for many useful comments, and acknowledges a Robert A. Millikan Fellowship from Caltech.