Attempts to increase the degree of blurring of nonfocused layers in tomography are inevitably associated with a worsening in the amount of artefactual detail introduced. A mathematical optimization technique has been employed to indicate how these conflicting characteristics may best be reconciled. The technique uses plausible definitions, based on the transfer function, of the effectiveness and the fidelity of the blurring, and generates a sequence of spread functions in which these two aspects are optimally combined.