In the theory of context-sensitive associative memory (CLAM) described in Eliashberg (1969Eliashberg ( ,1979Eliashberg ( ,1981Eliashberg ( ,1989) a broad range of psychological phenomena of short-term memory (STM) and temporal context (mental set) can be naturally understood as implications of the states of "residual excitation" in neural elements. Such hypothetical states of analog dynamic memory were referred to as E-states. The usefulness of the phenomenological concept of E-states is suggested by a variety of psychological, neurobiological and information processing considerations. The mathematical models of the dynamics of such states can be derived from many different assumptions about microscopic cellular mechanisms (see, e.g, Kuffler at al, 1984). One of the interesting possibilities (Eliashberg, 1989) is to connect the dynamics of the macroscopic E-states with the statistical dynamics of the conformations of the protein molecules in neural membranes. The corresponding formalism can be viewed as a natural system-theoretical extension of the classical Hodgkin and Huxley (1952) theory.
A single protein molecule is treated as a probabilistic finite-state machine. The probabilities of transitions between different conformations (states) of such a molecule (machine) am affected by different external inputs (membrane potential, concentrations of neurotransmitters, etc.). The average numbers of molecules in dlfferent conformations affect ionic currents, the rates of catalytic synthesis of second messengers, etc. These average numbers are identified with the E-states. Though the dynamics of a single molecule is discrete, the E-states change in a continuous fashion. This continuous dynamics can be highly nonlinear, because it is governed by a potentislly quite sophisticated logic within each molecule. The paper discusses some nontrivial possibilities of the outlined formalism