Learning and Long-term Physiological Regulation

A complex animal is composed of many different semiautonomous homeostatic mechanisms, and compromises must frequently be negotiated to achieve and maintain a stable overall physiological state optimal for survival. As evidenced by general skeletal behavior (McFarland, 1971), the brain has the integrative capacity required to resolve competing demands of these mechanisms. However, the conception of the brain as a process control computer—receiving data from an array of critically placed transducers, comparing the measurements to established set points, and dispatching instructions under a fixed program to a network of switches, valves, and pumps—is, although an appealing analogy, not tenable in the light of available data. Critical physiological variables are regulated to within narrow limits for periods of weeks or even years. Feedback-stabilized electromechanical regulators, such as thermostats, can maintain steady-state conditions indefinitely, but to do so they employ physical sensors or transducers which remain calibrated indefinitely. In contrast all interoceptors adapt (Chernigovsky, 1960; Mount-castle, 1980; Widdicomb, 1974), and analogous biological control schemes using interoceptors as sensors could not maintain regulation for extended periods of time.