|dc.description.abstract||Hilary Putnam’s ‘Psychological Predicates’ (1967) represents the first formalization of the argument for the multiple realizability theory (MRT) of mental states. Few years later, Jerry Fodor (1974) conceived the generalised version of the MRT over times which supported the irreducibility of ‘special sciences’ as the necessary consequence of Putnam’s theory. Every science grounded on the study of the mind became autonomous from any fixed ‘bridge law’ that could bind it to physics.
Recently, some philosophers have challenged the likelihood of the argument (Zangwill 1992, Shapiro 2000) whilst others have stressed the failures of the predictions implied by the generalised theory (Bechtel and Mundale, 1999).
My contribution to the controversy consists in a connectionist perspective that helps in showing two weak points in the standard approach to the MRT.
Firstly, philosophers usually forget that Turing machines cannot compute (or simulate) every possible mathematical function and it is still to demonstrate if they are able to perform the mathematical functions computed by neural systems. Such a result may be only achieved using a mathematical approach to the processes implemented by parallel systems and there are good reasons to believe that the unique features showed by these systems cannot be simulated in any serial devices.
Secondly, the possibility to isolate single parts of a complex system identifying their functions, should be rejected for several reason, the most important being that such functions lead to partial descriptions of the system itself. These descriptions may be useful in some contexts (i.e. if we do not have access to complete descriptions), but they fail to give us exhaustive explanations and should be considered weak as the ground for an explanatory model.
In conclusion, I claim that a complete description of an information processing system relies on the formalization of the ‘mathematical functions’ it realizes (i.e. the mathematical description of the processes implemented by the system). This formalization can only be achieved studying the physical matter that generates the mathematical functions.||en