Swamp creatures

From A conversation about the brain
Jump to: navigation, search

AG: Emma has asked me twice in the seminars on Shea's book[1] about swamp creatures. I thought it would be helpful to write down my simple (no doubt very naïve) view on the subject.

  • (i) Brains have only two ways of storing information, with synaptic weights and firing rates
  • (ii) Most information that is stored in the brain (past memories, etc) is stored as a pattern of synaptic weights
  • (iii) Information in the brain that determines the observer's perception must rely on current firing rates (in addition to synaptic weights).
  • (iv) 'Representation' is a complicated concept - best left to philosophers to define - but it seems to relate to both the information stored in synaptic weights and the information carried in firing rates.

I think most neuroscientists would agree with the above and even consider (i) to (iii) to be self-evident.

  • (v) A swamp man/macaque has an identical brain (synaptic weights, firing rates) to a real person/macaque with a history. Given that the brain has no other way of storing information, it is very hard to see how these two brains could represent things any differently from one another. What, at this physical level of explanation, is the alternative?
  • (vi) The flaw in the logic seems to be that someone might think they could reasonably imagine lightning (or similar) generating all the synaptic weights that make up the brain of someone who has had a long evolutionary and personal history. But this is not reasonable - the only thing in the real world that could generate such an extraordinarily complex pattern of synaptic weights is a very long evolutionary history and personal experience in the real world.
  • (vii) In summary, a neuroscientist's perspective is that there is nothing else to the brain (and representation in the brain) other than synaptic weights and firing rates. So, the swamp creature and the real person/macaque are identical at every level of description from physical to mental, including history, teleology, perception and thought. There is no way to tell them apart (by a neuroscientist's definition).

This is, of course, different from Shea's position. For example, p70, he says (about two swamp macaques, one of whom would be getting something wrong and one getting something right if they were real macaques with a history): "At the moment of creation neither monkey exemplifies the cluster of properties that underpins the explanatory purchase of representational content. Correspondingly, there is no substantial sense in which either one of them is getting it right or getting it wrong." I assume most neuroscientists would disagree with Shea here. Certainly, I do.

The rest of the pages in this wiki explore a simple relationship between synaptic weights and firing rates that might help explain representation in the brain (and is similar in many ways to representations generated by modern deep reinforcement learning) but, unlike the comments above about swamp creatures, this particular hypothesis is not a widely held view among neuroscientists.

See comments from Emma and Nick on Discussion tab.

Back to notes on Shea (2018).


  1. Shea, N. (2018). Representation in cognitive science. Oxford University Press.