Analyzing cognition

Here's a nice quote from the article "Why are animals cognitive?" by Richard Byrne & Lucy Bates in Current Biology (Jun 2006, 16(12), R445-R448):

Indeed, whether an animal’s behaviour is cognitive, and thus by implication ‘clever’, or associatively learnt is not an empirical question at all. These are simply two different ways of studying the same behaviour, and in the complex natural environments of most species only the cognitive approach leads to testable predictions.

The nicest thing about this is a rather clean exposition of why the "cognitive" can (should!) be considered as a separate level of representation. Specially for understanding animal behavior. And of course for understanding human behavior. Why the attribution of mental states as a level of explanation should be a mystery to anyone is a mystery to me...