Semantic systematicity and context in connectionist networks

Fodor and Pylyshyn argued that connectionist models could not be used to ex hibit and explain a phenomenon that they termed systematicity, and which th ey explained by possession of composition syntax and semantics for mental r epresentations and structure sensitivity of mental processes. This inabilit y of connectionist models, they argued, was particularly serious since it m eant that these models could not be used as alternative models to classical symbolic models to explain cognition. In this paper, a connectionist model is used to identify some properties which collectively show that connectio nist networks supply means for accomplishing a stronger version of systemat icity than Fodor and Pylyshyn opted for. It is argued that 'context-depende nt systematicity' is achievable within a connectionist framework. The argum ents put forward rest on a particular formulation of content and context of connectionist representation, firmly and technically based on connectionis t primitives in a learning environment. The perspective is motivated by the fundamental differences between the connectionist and classical architectu res, in terms of prerequisites, lower-level functionality and inherent cons traints. The claim is supported by a set of experiments using a connectioni st architecture that demonstrates both an ability of enforcing, what Fodor and Pylyshyn term systematic and nonsystematic processing using a single me chanism, and how novel items can be handled without prior classification. T he claim relies on extended learning feedback which enforces representation al context dependence.