I was reading this paper by Neil Taatgen on primitive information processing elements (PRIMs), as a type of machine-language for ACT-R. In the paper he claims:
The absence of variables means that the binding problem is no longer an issue
From what I understand, the binding problem is about how variables are represented neurologically. Namely how they can be combined and retrieved, which has typically been a problem for most Connectionist architectures.
Why does the lack of variables in PRIMs (they seem to be just operations) mean that the binding problem is no longer an issue?