In 1976 S. Eilenberg and M.-P. Schützenberger posed the following diabolical question: if A is a finite algebraic structure, Σ is the set of all identities true in A, and there exists a finite subset F of Σ such that F and Σ have exactly the same finite models, must there also exist a finite subset F′ of Σ such that F′ and Σ have exactly the same finite and infinite models? (That is, must the identities of A be “finitely based”?) It is known that any counter-example to their question (if one exists) must fail to be finitely based in a particularly strange way. In this paper we show that the “inherently nonfinitely based” algebras constructed by Lawrence and Willard from group actions do not fail to be finitely based in this particularly strange way, and so do not provide a counter-example to the question of Eilenberg and Schützenberger. As a corollary, we give the first known examples of inherently nonfinitely based “automatic algebras” constructed from group actions.