Implications of Hadley’s Definition of Systematicity

In the ongoing debate between connectionists and proponents of the LOT Hadley [2] has contributed a new definition of systematicity, a term brought into the debate by Fodor [3]. Two important new aspects of this definition are that it provides a precise test of a systems level of systematicity, and that it puts strong emphasis on learning. Both of these aspects have their implications for the debate. This paper will try to make these implications clear.

An implication of the definition being more precise is that the discussion about whether or not a connectionist network behaves systematically can be greatly reduced (...). As a consequence, the definition raises a challenge for connectionists to strengthen their case on empirical grounds. We will discuss a project that aims to meet this challenge [4], and examine the arguments Hadley [1] uses to reason out that it still does not live up to his definition.

An implication of the emphasis Hadleys definition puts on learning is that the claim that the LOT does account for systematicity becomes less clear, as the trouble symbolic systems have in learning is notorious. Hadley cites [6] as an example of a symbolic system that exhibits systematicity. We will take a critical view on this example, to see whether the systematicity exhibited falls within the definition of the concept by Hadley.

Editors and Affiliations

1. Dutch Foundation for Neural Networks (SNN), Geert Grooteplein Noord 21, 6525 EZ, Nijmegen, The Netherlands

   Bert Kappen  & Stan Gielen  & 

Reprints and permissions

© 1995 Springer-Verlag London Limited

Policies and ethics