A Chinese phonological enigma

Recent cross-language research has yielded strong statistical evidence in support of the idea, advocated by André Martinet and widely accepted by linguists, that languages avoid adopting sound-changes which would create many homophones. Yet we know that the history of Chinese phonology has been marked by repeated phoneme mergers and losses which led to a very high incidence of homophony, forcing the monomorphemic vocabulary of the classical language to be replaced by a largely bimorphemic modern vocabulary. This paper examines various ways in which this apparent contradiction might be resolved. None seems fully satisfactory, yet some resolution must exist.