of labour necessary to the production of a thing is diminished, then the
value of that thing diminishes also. Hence no increase in the
productiveness of labour can increase the total sum of values. This,
however, is a fundamental mistake, for what depends upon the amount of
labour is merely the _relative_ value of things--the exchange relation in
which they stand to other things. This is so self-evident that Ricardo
himself cannot avoid expressly stating that he is speaking of merely the
'relative' value of things; nevertheless, this relative value--which,
strictly speaking, is nothing but a value relation, the relation of
values--is treated by him as if it were absolute value.

And yet Ricardo's error is a not less important step in the evolution of
doctrine than those of his previously mentioned predecessors. It signifies
the revival of the original problem of political economy, which had been
lost sight of since Adam Smith; and Ricardo's follower, Marx, is in a
certain sense right when, with bitter scorn, he denounces as 'vulgar
economists' those who, persistently clinging to Smith's optimism, see in
the _productiveness_ of labour the measure of the increase of _actual_
wealth. For all that was brought against Ricardo by his opponents was known
by him as well as or better than by them; only he knew what had escaped
their notice, or what they saw no obligation to take note of in their
theory--namely, that the actual facts directly contradicted the doctrine.
It by no means escaped Ricardo that his attempted reconciliation of the
theory with the great problem of economics was absurd; and Marx has most
clearly shown the absurdity of it. The latter speaks of the alleged
dependence of value, not upon the productiveness of labour, but upon the
effort put forth by the labourer, as the 'fetishism' of industry; this
relation, being unnatural, contrary to the nature of things, ought
therefore--and this, again, is Marx's contribution to the progress of the
science--to be referred back to an unnatural ultimate cause residing, not