The premise of a physical symbol system is that the symbols have real world interpretations.
There is no plausible process for evolving a mechanism for a discrete irriducible system in evolution.
If you examine the steps of how a species might evolve a symbolic logic system, it kind of makes sense how the brain might actually work. A natural symbolic logic might have these components;
- The brain would need to evolve some sort of generic token system to hold the symbols.
- It would need some sort of process to map the external objects to the tokens
- It would need some system to collate the symbols and tokens, and to provide persistence.
- It would need some plastic system of rules to implement the reasoning.
Species dont carry around "potentially" beneficial characteristics, if there is biological cost to maintaining them in the genome.
If you look at each of the steps 1-4 there are continuous natural analogues to each of the requirements of symbolic logic, but none of them are discrete, reliable or positive.
I see symbolic logic as an "other hill" problem in fitness landscapes. There is a basic simplicity and efficiency in symbolic system, but there is no way to get from our fuzzy continuous meaty monkey world to the distant hill.
Fortunately monkey hill is sufficiently high, that we can see logic hill from over here. So we evolved a continuous approximation to that hard cold logic and use the meat logic to build tools that provide logic hill capabilites.
Notes: this is a brain dump of my AI revision, and could probably do with some more effort to make in fit for consumption.