SummaryWhen an aerial is used to survey the distribution of radio brightness over the sky, the observed distribution is smoother than the true distribution; the broader the beam of the aerial, the greater the smoothing. It is shown that the aerial does not register those spatial Fourier components of the true distribution having frequencies beyond a cut-off determined by the aerial aperture. Components of lower frequency are registered but their relative strengths are altered.Two important consequences follow. (i) There are invisible distributions which produce no response when scanned by the aerial. Consequently there is not a unique solution to the problem of correcting for aerial smoothing. The established method of correcting by successive smoothing, leads to the principal 8olution, in which Fourier components accepted by the aerial have been restored to their full values, but the components rejected by the aerial are still not represented. (ii) In conducting a survey it is sufficient to observe at discrete intervals. The measuring points must be closer together than half the period of the Fouri,er component at cut-off. For an aperture of width w, this peculiar interval is equal to fA/w (radians).