[en] The implications of Dirac's Large Numbers Hypothesis for the past history of the Earth's rotation are examined. Three variants of the hypothesis (early Dirac, Additive creation and Multiplicative creation) make differing predictions for the variation over geological time of the numbers of days in the month and year. The record of these variations preserved by biological growth rhythms in fossil bivalve molluscs is used to eliminate the tidal component and compare the predictions of the Large Numbers Hypothesis with observation. Early Dirac and Additive creation are rejected. However, Multiplicative creation is in as good agreement with the fossil data as the conventional physical and astronomical theory that assumes dG/dt/G = 0. (author)