Why try to prove that bees can ‘master’ concepts?
This recent PNAS paper, by Avargues-Weber et al., recounts a series of Y-maze experiments where bees are trained to distinguish two classes of visual pattern. The training patterns are always composed of two different parts which can be left and right of each other or above and below. Bees learn to favour the appropriate spatial relationship, as well as learning to prefer patterns that are composed of different elements. The authors justify their description of this as bees being able to learn 2 “abstract concepts”, primarily because there are no simple perceptual rules which can explain to choice.
There are 2 points I would like to raise. Firstly, I disagree with the descriptions of “Left-Right”, “Above-Below” and “Same-Different” as abstract concepts. It is very easy to imagine how they can be grounded within simple sensori-motor interactions with the world. Spatial properties of a scene are captured by first or second order visual parameters and can be used directly for navigation or recognition without the need for them to be represented as abstract concepts. Similarly a “Same-Different” judgement within a single scene may be captured through an ecologically relevant visual symmetry mechanism.
Thinking about possible mechanisms, leads me to my second point: In my opinion, the value of insect behaviour research is that we can gain deep access to a system, where we understand the task that the insect is trying to perform. This may allow us to consider how apparently sophisticated cognitive behaviours emerge from sensori-motor primitives. These sensori-motor primitives are likely to be phylogenetically widespread, and therefore the building blocks of more cognitive processes across taxa. This represents a principled bottom-up approach to comparative cognition. I worry that an alternative approach, of trying to argue for the sophistication of a behaviour, without considering the mechanisms, will lead to a less deep understanding of animal cognition.