Conceptual Spaces: The Geometry of Thought

by Peter Gärdenfors

Within cognitive science, two approaches currently dominate the problem of modeling representations. The symbolic approach views cognition as computation involving symbolic manipulation. Connectionism, a special case of associationism, models associations using artifical neuron networks. Peter Gardenfors offers his theory of conceptual representations as a bridge between the symbolic and connectionist approaches.

Genres: Philosophy, Linguistics, Nonfiction, Science, Psychology, Mathematics, Neuroscience

Unknown Binding

First published March 20, 2000

About the author

Professor of cognitive science at the University of Lund, Sweden.

Reviews

[Nat · Rating 4 out of 5]

Gärdenfors sketches an intermediate level of representational content, between the "symbolic" (syntactic and propositional) and associationist levels, and argues that this intermediate level of conceptual content helps resolve a variety of deep problems for cognitive science, such as how thinkers acquire novel cognitive content, how to understand metaphor, and how induction works.

My interest is mainly in philosophy of language, and there are many interesting suggestions for how to think differently about the meaning of expressions. Gärdenfors advocates cognitive semantics, which eschews reference and focuses on conceptual structures as the meanings of expressions. On the conceptual space picture of cognitive semantics, properties (which normally would be understood as functions from possible worlds to extensions) are understood SPATIALLY, organized according to a variety of topological principles. So, for example, color properies are "convex" regions of the 3-D color solid. Objects are located at particular points in color space, so a judgment that x is green is true just in case x is located in the "green" region of the color solid.

Even mainstream referential semantics turns to "scales" to handle the meaning of gradable adjectives, which is a step in this geometrical direction. Gärdenfors's conceptual spaces seem best suited to handle spatial content (such as that had by prepositions), color, and other domains with an obvious structure. It seems less well-suited to handling even relatively simple medium-sized dry goods like "apple" (what dimensions make up "fruit space"? Gärdenfors takes a stab at listing some of them.).The book raises more questions than it answers, but that's a good thing.

[Jonathan Cashman · Rating 4 out of 5]

It is very clear and concise.
I really like the treatment of metaphor.