Commentary on Phillipe G. Schyns, Robert L. Goldstone, and Jean-Pierre Thibaut (1998) The development of features in object concepts. BBS 21:1–54.
Functional identification of constraints on feature creation
|Phillipe G. Schyns a1, Robert L. Goldstone a2 and Jean-Pierre Thibaut a3|
a1 Department of Psychology, University of Glasgow, Glasgow, G12 8QB, United Kingdom
a2 Department of Psychology, Indiana University, Bloomington, IN 47405
a3 Department of Psychology, Université de Liège, Batiment B32, Sart-Tilman, 4000 Liège, Belgium
Dawson's provocative comment makes three connected points: (1) to be falsifiable, theories that assume flexible features must constrain their feature creation and mechanisms, (2) the explanatory power of such functional theories is rooted in the properties of their underlying physical mechanisms, and (3) to derive the relevant constraints of feature creation from these mechanisms, it is critical to avoid the scope slip. We will argue here that even though we agree with (1) and (2), (3) confuses two different levels of analysis of computational systems: the functional identification and the physical implementation of relevant constraints.