Numerical geological time-scales rest firmly on the analysis of radiometric data. Since the discovery of oceanic magnetic stripes and their signification, alternative time-scales have been constructed whose numerical predictions are based on the hypothesis of a constant rate of sea-floor spreading. The most recent such scale uses only five tie-points, plus present day, to subdivide a span of over 80 million years. The present paper demonstrates that the rate of ocean-floor spreading has varied markedly, and without obvious pattern both in time and in space. Thus, magnetostratigraphic scales based on large interpolations are fundamentally flawed. An alternative scale, based on a consideration of all radiometric data and using magnetostratigraphic data only for small interpolations is commended.