Information in the Brain: From Metaphor to Truth

ABSTRACT Despite explicit warnings from Shannon to tread carefully when applying Information Theory to fields for which it was not designed, contemporary neuroscientists adopting the framework of Information Theory have fallen right into the traps Shannon and others cautioned against. What makes the neuroscientist more than anyone prone to fall prey to confusion is that neuroscience looks back on a tradition—which long predates Shannon—in which information and related semantic communication notions (message, codes) are invoked as central explanatory posits. In the first part of this paper, I shall give a selective historical overview of this tradition, which runs via Aristotle over Descartes to the groundbreaking work of the early nerve physiologists. At the same time, I will show how the history of explaining brain activity in semantic communication terms is characterised by a process of de‐metaphorisation. Next, I will show how semantic notions, including semantic information, are still routinely relied upon within contemporary information theoretic approaches to the brain, despite the fact that Shannon deemed meaning irrelevant for his theory. There are, then, two very different notions of information at work within one and the same scientific discipline. I will demonstrate how these two notions get as a rule tangled up here. Next, I will show how invoking semantic‐intentional information at the level of the brain raises two specific problems: first, notions that aim to explain by invoking semantic content or meaning, like the pre‐theoretical notion of information, in principle resist integration within causal‐mechanistic explanations; second, such notions are part of a mode of explanation which logically presupposes a cognitive agent. As I will also show, however, we indeed find that the neuroscientist has embraced the habit of treating neurons as cognitive agents, attributing to them capacities we would normally only attribute to persons.