We have argued that several different metrics for the `valuation' of bundles of commodities are possible in principle, most of them logically incompatible with the idea that any scalar quantity is conserved in exchange. But the fact that individual commodities are separable, and separately tradable, imposes one particular metric, corresponding to what we called commodity value space--and this metric is consistent with a conservation law. This formal argument does not in itself prove that any identifiable `substance' is in fact conserved, nor does it establish the credentials of labour time as prime candidate for conservation. That is an empirical matter; and we have shown that the conservation of socially necessary labour time holds as a fairly close approximation in fact. Alternative candidates for conservation in exchange fare much less well empirically, and besides involve theoretical problems that can plausibly be circumvented in the case of labour. The Ricardian-Marxian `law of value' may be given a precise definition, on which, moreover, it turns out to be valid.