Hidden Variables Simulating Quantum Contextuality Increasingly Violate the Holevo Bound

In this paper we approach some questions about quantum contextuality with tools from formal logic. In particular, we consider an experiment associated with the PeresMermin square. The language of all possible sequences of outcomes of the experiment is classified in the Chomsky hierarchy and seen to be a regular language. Next, we make the rather evident observation that a finite set of hidden finite valued variables can never account for indeterminism in an ideally isolated repeatable experiment. We see that, when the language of possible outcomes of the experiment is regular, as is the case with the Peres-Mermin square, the amount of binary-valued hidden variables needed to de-randomize the model for all sequences of experiments up to length n grows as bad as it could be: linearly in n. We introduce a very abstract model of machine that simulates nature in a particular sense. A lower-bound on the number of memory states of such machines is proved if they were to simulate the experiment that corresponds to the Peres-Mermin square. Moreover, the proof of this lower bound is seen to scale to a certain generalization of the PeresMermin square. For this scaled experiment it is seen that the Holevo bound is violated and that the degree of violation increases uniformly.