When dealing with the Pareto-Walrasian general equilibrium theory, Pascal Bridel repeatedly underlined that from one generation of economists to another, the methodology ruling this theory has been continuously changing: from Léon Walras's idealism to Vilfredo Pareto's positivism; from Paul A. Samuelson's operationism to Robert E. Lucas's instrumentalism; and so forth. In this regard, one major methodological turn was the move from cardinalist utility functions, resting on psychological foundations, to ordinalist ones. Samuelson's contribution, both methodological (introducing operationism in economics) and theoretical (the well-known revealed preference theory), is probably the most important, although recent works in economics increasingly reintroduce psychological elements in economic theory. In our essay we attempt at drawing a parallel between Samuelson's operationalization-of-consumer theory through revealed preference theory, and his works in production theory from the late 1940s on, in particular the non-substitution theorem (NST).

In 1949, the NST was first demonstrated simultaneously and independently by Samuelson and Georgescu-Roegen for a so-called (static) ‘generalized model of Leontief’. The theorem raised the question as to whether substitutions among factors of production could not occur even if possible. Samuelson and Georgescu-Roegen showed that, although possible, in certain conditions if the structure of final demand changes, the structure of production remains unchanged and no factor substitution occurs.