Neween, Albert & Bartels, Andreas - Animal Minds and the Possession of Concepts

10/26/2007

Philosophical Psychology Vol 20 No 3 June 2007

This quick moving paper takes a combination of empirical, intuitive and reasoned approach to giving an epistemic model of concept possession. That is to say: it tries to lay out conditions where we can ascribe the possession of concepts in non-human animals. The first step is to lay out what it means to have a concept. Here, authors borrow from Peacocke in using a psychological-functional mental-capacities approach to concepts. (pg 284) Here also some preliminary ground work is done for what a concept is, how to individuate it, and so on.

The beginning attacks prior writers (Chater & Heyes) that argued that there are three theories of what a concept is: definitional, exemplar, and prototype. The problem is that only prototype concepts can be 'formulated' nonlinguistically, so thus nonlinguistic animals can only have this type of concept. But often it is impossible to tell the difference between conditioned stimulus-response (S-R) systems and prototype concept use. Using principles of parsimony, there is no reason to think animals have concepts. Authors take this to be a challenge of determining the differences between S-R and concept use. (pg 285-6)

Authors consider various approaches, e.g. Dretske, Allen & Hauser/Allen. Dretske gives the idea that you need a flexible response to the same stimuli at a minimum (to distinguish between S-R and concept use). Allen suggests that concepts require integration of more than one source of information in order to modify behavior-- stimulus independent, or as Pylyshyn suggests, 'transcendence of particular stimuli' (pg 287). Using this we arrive at the value of 'stable representations' that are independent of particular stimuli. This constraint is only necessary (not sufficient), since it could merely give us naming abilities for objects (see collie discussion pg 288).

Authors consider the Davidson argument about belief possession, and say flat out that it is wrong because it presupposes second-order belief usage in order to have first-order beliefs. That is, you need to have a concept of a concept before you can have a concept. Authors deny this could be realistic (pg 289), but agree with Davidson that there is a normative element to concept possession. Authors instead take this more modest metacognitive criterion from Allen:
ii. An Organism is capable of detecting some of its own discrimination errors between veridical applications of concept X and veridical applications of concept X (pg 289)

The next step is to have the ability to extract classes from perceptual data-- to separate properties from objects. Authors examine the case study for Alex the parrot, that was able to successfully 85% separate objects into piles according to the categories color, shape, material. Thus showing not just that Alex could learn red, blue, etc., but also higher-level categories. (pg 292-4) Authors distinguish between three different categories of cognitive capacities: nonconceptual representations (S-R), conceptual representations and propositional representations. Here is an interesting discussion about the 'cognitive' and what it means for something to be cognitive. [Clearly, 'cognition' isn't the same as 'thinking' anymore if an ant can do it to get home!] (pg 294) Here we are homing in on conceptual representations. In order to have a first-order perceptual concept (e.g. RED):
C1: The system has a stable representation of the property across a range of objects
C2: The system must be able to represent not only property X but also other properties of the same object
C3: The system has a distinguishing capacity that picks out particular properties of an object that is independent of the stimulus of the property itself (e.g. the system can, on its own, pick out the relevant property without its making itself salient somehow)
C4: The system has a minimal semantic net where the first-order concept fits into a proper dimension (e.g. RED fits into COLOR rather than STOPS or DANGER-WARNINGS).
There is extended discussion on these necessary conditions. (pg 296-7). Authors give their suggestions for what propositional representations are-- the combination of two concepts that are strongly stimulus-independent. Authors then look at the empirical evidence to see how non-human animals fit into the criteria. (pg 298-300) Authors reply to a few objections and summarize their claims. (pg 303-305)