On The Causality Of Neural Communication

Neuroscience often describes neural interactions with the language of communication but rarely states what a process must satisfy to count as ‘communication’. Outside neuroscience, the dominant answer places the criterion on the sender, i.e., a signal forms a communication channel when it is produced with the intention that the receiver recognise that intention. However, this criterion cannot be readily applied to neurons since they cannot be attributed with communicative intentions. Thus, here, we argue that while neural signals are unintentional, they are still functional, meaning that they can assert a causal influence on the receiver. We therefore define communication as having taken place if, and only if, the receiver behaves differently than it would have done had the message not arrived. Our point is that establishing neural communication is fundamentally a problem of causal inference since it requires identifying the receiver’s counterfactual behaviour. We then set out two families of approaches to deal with the counterfactual: soft tests, which estimate it from computational models and other forms of heuristics, and hard tests, which intervene on the channel directly.