How to improve on Quinian bootstrapping - a response to nativist objections

How to improve on Quinian bootstrapping – a response to nativist objections Zoltan Jakab (zoltan.jakab@barczi.elte.hu) Institute for Psychology of Special Needs, Lorand Eotvos University, Ecseri ut 3 1097 Budapest, Hungary often sound intuitively plausible. 2 Hence Fodor’s early view on concept acquisition, according to which primitive concepts cannot be learned, and so they must be innate. In Fodor (1990, 1998) this idea is supplemented by argument that most of our concepts are primitive, giving rise to radical concept nativism. More recently, Fodor (2008) has found this conclusion much too weak, and formulated a stronger version according to which the obstacle to concept learning is not that most concepts are primitive, but rather that compositionality cannot increase the expressive power of cognitive systems. Very roughly, expressive power is the range of concepts and hypotheses (theories, conceptions) that a given representational system could formulate, or express, given its primitive symbols (concepts) and rules of combination. As we currently understand cognition, any case of learning seems to be underlain by some mental process that exploits compositionality: forming complex mental representations out of simpler ones governed by rules, plus adjusting certain parameters of the primitives. 3 If this is how cognition operates, then all that learning can achieve is the manifestation of what’s born with us: we actually come to express what we are innately capable of expressing. In formal logic, building new complexes out of primitives (symbols and rules of combination) does not count as increase in expressive power – only adding certain new primitives 4 does. The same restriction seems to apply in the realm of mental representation, if indeed compositionality is the only game in town for cognitive theorizing. In sum, since no cognitive mechanism that we can currently think of transcends compositionality, and compositionality cannot increase expressive power beyond one’s innate endowment, no cognitive mechanism can increase the expressive power of mental representational Abstract Quinian bootstrapping is Susan Carey's solution to Fodor’s paradox of concept learning. Carey claims that contrary to Fodor’s view, not all learning amounts to hypothesis testing, and that there are ways in which even primitive concepts can be learned. Recently Georges Rey has argued that Carey’s attempt to refute radical concept nativism is unsuccessful. First it cannot explain how the expressive power of mental representational systems could increase due to learning. Second, both Fodorian circularity charges and Goodmanian problems of indeterminacy apply to Carey’s examples of Quinian bootstrapping. I argue that Carey’s examples of bootstrapping can be amended to escape Fodorian and Goodmanian objections. I suggest some ways to improve on our models of concept learning to this end. I also argue that skill learning is the way for mental representational systems to increase their own expressive power, that is, to enrich their conceptual repertoire beyond what compositionality alone affords. Keywords: Quinian bootstrapping; learning; expressive power nativism; concept Introduction: Fodor’s paradox Jerry Fodor’s famous argument for concept innateness has taken different forms. According to its early version (Fodor, 1975, 1981), all learning is hypothesis testing, and to formulate a hypothesis one needs to possess all the concepts that the hypothesis involves. Therefore, in the course of learning one can seek for evidence supporting or undermining a hypothesis, but formulating hypotheses will never result in the acquisition of new concepts. In order to work, this argument needs an immediate qualification. Some concepts are structured, whereas others are not; they are primitive. At any rate, this distinction holds on Fodor’s own representational theory of mind which endorses compositionality. Complex concepts arise as combinations of primitives (or simpler complexes). For example, WHITE RAVEN 1 is a complex concept which has two constituents: WHITE and RAVEN; the two constituents are related by a conjunction. Now if one wishes to test the hypothesis that White ravens are quite rare, and it is conceded that forming WHITE RAVEN out of WHITE and RAVEN counts as learning a new concept, then evidently hypothesis formation makes room for concept learning. For more complex cases the idea that compositionality affords concept learning does For example, learning the idea of MOVING THE KING TWO SQUARES TOWARDS A ROOK ON THE PLAYER'S FIRST RANK, FOLLOWED BY MOVING THE ROOK ONTO THE SQUARE OVER WHICH THE KING CROSSED (i.e., CASTLING in chess) may strike one as a case of bona fide concept learning. Many examples support this generalization from the formation of perceptual prototypes to the construction of schemas, scripts, mental models, and propositional representations. As Carey (2009) says, it is a truism that all learning involves building new representations from antecedently available ones. I.e., adding new operators, predicates, etc. with content that no combination of the earlier set of primitives could represent. One example is adding the modal operators ’possibly’ and ’necessarily’ to classical propositional logic. Concepts here are denoted by the name of their referent typed in upper case letters.