The cognitive nature of predicates is a central issue in Evolutionary Cognitive Science. Virtually all representational accounts of meaning postulate the existence of mental structures that involve at least one variable slot, of the form PREDICATE(X) (HURFORD 2003). Evolutionary Cognitive Science must explain how such representations could emerge as an interface between language (with syntactic dependencies like verb/subject or adjective/noun hierarchies) and reasoning (with logical mechanisms like negation or deduction).
A commonly shared assumption is that animals do have predicates, since they seem able to perform categorisation. The problem is that even bacteria should be granted with the ability to form predicates, as they can categorise their environment as, say, acid vs. neutral pH. Our claim is that there is a qualitative difference between mere object or situation recognition and explicit predication (GHADAKPOUR & DESSALLES 2003).
We developed a model in which predication results from mental contrast (GHADAKPOUR 2003). In this model, any perceptual representation P can be contrasted with the nearest prototype A (or, more generally, the nearest attractor, which may be also an exemplar or a good shape), resulting in another, generally less instantiated, perceptual representation V. Mnemonically: P – A = V. The representation V can be understood as a variation of A, which may be gradual or not, as when you see a foot which is huge or has an additional toe. C-predicates result from the application of elementary topological operations (inclusion, separation) on this variation. If we note V(P) and V(A) the projection of P and A on V, predication takes the form of explicit categorisation when V(P) is included in the standard deviation of V(A), while explicit discrimination is achieved through topological separation between V(P) and V(A).
Contrary to most models of conceptual representation, our model is inherently context-dependent. A percept P determines the nearest attractor A, and both determine the contrasting variation V. Contextual effects like priming may change the choice of A and V, as the links from P to A and from P–A to V are supposed to be associative. Hence, negation is context-dependent: in specific contexts, a prawn is not a shrimp, though they are alike in almost all perceptual dimensions. Note that a holistic system like a classical neural network can’t distinguish its first prawn from the known class of shrimps. Our model has no problem producing a contrast between a prawn and the shrimp prototype and giving some relevant difference V. Conceptual modification is context-dependent too (small molecule vs. small galaxy), since the modifier is evoked by the contrast. Note that no supplementary machinery is needed to account for metaphors in modification (e.g. a small idea).
Our model offers a variety of other advantages. The most obvious is that it provides symbolic interfaces between syntax and reasoning, while avoiding the intractable difficulties of postulating increasing mental lexicons (GHADAKPOUR 2003). C-predication does not require any inventory of predefined elementary predicates. C-predicates are transitory representations, constructed at the interface between language, reasoning and perceptual space.
At this point, our model is consistent, but still incomplete. Many aspects concerning semantic determination and thematic roles are not yet fully accounted for. Though, the ability to contrast perceptual representations (either perceived or memorised) can be seen as a qualitative difference that separates our species from others.
JEAN-LOUIS DESSALLES & LALEH GHADAKPOUR, 2003, “Object Recognition is not Predication” (commentary on “The Neural Basis of Predicate-Argument Structure” by J. R. HURFORD), Behavioral and Brain Sciences, 26: 290-291.
LALEH GHADAKPOUR, 2003, Le système conceptuel, à l'interface entre le langage, le raisonnement et l'espace qualitatif : vers un modèle: de représentations éphémères. Thesis Dissertation: École Polytechnique.
STEPHEN J GOULD & NILES ELDREDGE, 1977, “Punctuated Equilibria: the Tempo and Mode of Evolution Reconsidered”, Paleobiology, 3: 115-151.
JAMES R HURFORD, 2003, “The Neural Basis of Predicate-Argument Structure”, Behavioral and Brain Sciences, 26: 261-283.