In striving for ever greater angular resolution in riometer (relative ionospheric opacity meter) observations it is tempting to try to use a nonfilled array to reduce the cost of antenna elements. We analyze as an example the Mills Cross configuration, compare its performance as a riometer with that of a filled square array of the same angular resolution and inquire about the penalty of using a sparsely filled array. We find that the susceptibility to interference from sources in the sidelobes is greater than in the filled array, that strong point sources in the sidelobes can cause both positive and negative excursions in the output and that the noise performance is inferior and requires longer integration times for a given rms error than the filled array.