It is submitted that, for the very large number of different traditional type formulae to determine price indices associated with a pair of periods, which are joined with the longstanding question of which one to choose, they should all be abandoned. A method is proposed whereby price levels associated periods are first all computed together, subject to a consistency of the data, and then price indices that are all true are determined from their ratios. An approximation method can apply in the case of inconsistency.