Stahl679 bmp

THE HOLE OF MODELS IN THEORETICAL BIOLOGY 203

(1959, 1962), and Overall and Williams (1961)—the latter work has the specific title “Models of Medical Diagnosis.” Computer simulations of human diagnosis may be based on nonspecific decision algorithms, Boolean logical methods, probability theory such as the Bayes theorem, and also nonspecific statistical methods. In regard to similarity criteria, the simplest approach is to disregard mechanism and only consider “performance scores.” Alternately, detailed relational and numerical invariants may be evaluated, but it is very difficult to obtain their true values for the human prototype.

B. Mathematical Learning Models

The flourishing field of “mathematical psychology” is largely based on an entirely different modeling concept, namely construction of numerical models of ąuantifiable behavior, such as learning of a simple task or word association. They are readily distinguished from algo-rithmic thought models simply by the characteristic mathematical tools applied for the simulation: Computer or automata programs for algo-rithmic models and numerical or probabilistic eąuations, possibly exe-cuted on a Computer, for the usual type of psychological model.

Numerous important works are available on mathematical psychological models, including a symposium volume by Luce (1960) on all aspects of the field, a boolc devoted to stochastic learning models by Bush and Mosteller (1955), and a volume dealing with simulation of complex conditioned-reflex learning by Steinbuch (1961). Much of the basie theory of psychological learning is based on numerical stochastic methods and specific reports on stochastic models are available by Bower and Theios (1964), Horvath (1963), Overall (1960), and others. Atkinson and Crothers (1964) have reported on the “comparison of paired-asso-ciate learning models having different acąuisition and retention axioms.” In this latter case, choice of the “axioms” would constitute an algo-rithmic design, but course of learning is followed by numerical criteria.

A rather different approach to psychological simulation is given in a report of Hinde (1960) entitled “Energy Models of Motivation.” The energy in this case is clearly not of the Newtonian “physical” type, but rather defined as the summation of impulses arriving in certain portions of the brain. Freudian hbidinal energy can perhaps be similarly under-stood. An artificial model dealing with psychological energy phenomena could be created with an analog or digital Computer and would be based on entirely different concepts than most learning or algorithmic thought models. Other models of “psychodynamic” interest include one dealing