Commonly in neuroscientific research today, scientists build models that can perform cognitive capacities and compare their activity with neuronal activity, with the purpose of learning about brain computations. These models are constrained only by the task they must perform. Therefore, it is a worthwhile scientific finding that the workings of these models are similar to neuronal activity. This is a promising method to understanding cognition. However, I argue that it is likely to succeed in explaining how cognitive capacities are performed only when the capacities’ etiology is considered while choosing the modeled capacities. Otherwise, it may lead scientific practice astray.

Godfrey-Smith recently introduced the idea of representational ‘organization’. Representations from an organized family are tokened on different occasions and systematically interrelated (eg. analogue magnitude representations). Organization has been elided with structural representation, but the two are in fact distinct. An under-appreciated merit of representational organization is the way it facilitates computational processing. When representations from different organized families interact, they form a processing structure. These processing structures can be computationally useful. Many of the cases where organization has seemed significant, but which fall short of structural representation, are cases where representational organization underpins a computationally useful processing structure.

Behavioural systems present a relevance problem: there’s too much information about them to include all of it in our explanations, so we must decide what information should be included in and excluded from explanation. One popular solution is to (a) include only information at a single level and (b) exclude information at other levels. This excludes relevant information about interlevel processes. However, neuroscientists have found an implicit way to include relevant interlevel processes in their explanations of behavioural circuits. I argue that we can rationally reconstruct an interlevel theory of relevance for behavioural systems from their explanations of behavioural circuits.

Terminological inconsistency in neuroscience obscures ontological relations between neural circuits and cognition. Meanwhile, the dominant view among philosophers is that human cognition is neurally realized. It remains an open question whether the extensive philosophical literature on (multiple) realization sheds light on the ontological unclarity in neuroscience. Here I identify the kinds of experiments in neuroscience that are relevant to determining whether and how cognition is neurally realized. I then argue on empirical grounds that realization is not a relation between individual circuits and cognitive phenomena.