[g22 



II WDBIIllK OP I'lIYSIllI.OGY 



NKtROPIIYSIOI.OGY III 



to give appropriate weight n> each entry, the prob- 

 ability nl the particular observable occurring in the 



particular disturbance is programmed — headache is 

 the rule with a brain tumor, the exception with a 

 dislocated disc. Then some value ratings are attached 

 to the decisions: a positive error in diagnosing appen- 

 dicitis and operating, rather than choosin" 'indiges- 

 tion' and giving a soothing potion, is more serious 

 than is a similar positive error in diagnosing choic- 

 er stitis and prescribing lied rest; whereas a negative 

 error, in diaa;nosinsj appendicitis or cholecystitis as 

 'indigestion' is more serious in the reverse order. 

 I ,ater, tin- initial settings of the machine are readjusted 

 in terms of the feedback from its own experience — 

 corrected for clinical judgment, autopsy findings and 

 the like so Uuit its performance steadily improves. 

 The initial information is supplied by human judg- 

 ment from book--, records and panels of experts, and 

 no sensible person would prefer the computer's 

 diagnosis to that of his doctor. But, in time, the 

 collective wisdom of man could be integrated in a 

 single instrument and enriched by its experience at a 

 i. Mi impossible to a single brain. Then the physician 

 will -.eek aid from it. 



Another computer, properly programmed and 

 given the axioms of the Trincipia Mathematical has 

 solved over three fourths of the first 50 theorems 

 (219, 220); a third can learn to discriminate patterns 

 without prior instructions (252); a fourth behaves 

 directly as a neuron assembly and serves to test 

 parameters (-249). One is being developed (250) to 

 simulate .1 society (or a brain, for that matter 1 with 

 originally naive members and loose connectivities 



that develop special roles and organizations while 



achieving effective performance. Indeed, learning by 

 a system may involve little Learning by its compo- 

 nents whether a brain (203) or .1 structural group 

 (Bavelas, personal communication) or, for that 

 matter, .1 gene pool. With computers learning to 

 learn, the epigenetic invention of life, great advances 

 an- imiuiiii mi See also von Neumann (287), Simon 

 & Xewell (264) and Green (127). A valuable further 

 reference on processing neuroelectric data is Technical 

 Report ;",i, Massachusetts Institute oi fechnology 

 Research Laboratory of Electronics, 1959, by Rosen- 



blith and others.] 



My hope in this ess.iv is to help direct the attention 

 ■ •I neurophysiologists, especially oi those starting their 

 careers in formal physiology departments, to the 

 mounting challenge and opportunities in this sector 

 oi behavioral science. Behavior is rooted in the 

 metabolism ol neurons and flowers in the meat 



imaginative creations of mankind. Understanding 



it will demand the pooled skills of many presently 

 disparate disciplines. 



Architecture of Knou ledgi 



LEVELS. A later section of (his chapter will consider 

 the nervous system at different levels of organization 

 from that of uross anatomy to that of molecules, so a 

 brief over-all consideration of general level properties 

 may be helpful. Any svstem with sufficient identity 

 iu permit studv has some sort of a boundary which 

 separates it from its environment. This environment 

 may itself be a larger svstem in which the system 

 gi\ en attention is an element, or the environment may 

 be entirely external to any organization of which the 

 system is an integral part. Internally, also, the 

 system is composed of discriminahle elements, not 

 necessarily all alike, which mav themselves be 

 subsystems containing further subordinate units. 



It has proved helpful in considering the broad 

 scientific scene to plot the levels on an ordinate axis 

 and the attributes of swems on the abscissa. The 

 levels that concern biology run from the molecule up 

 through the organelle, cell, tissue and organ, organ- 

 ism and small hereditary or ecological group, to a 

 larger society or ecosystem. The relevant attributes 

 an- 'being' or architecture, 'behaving 5 or function and 

 'becoming 1 or development. The table (fig. 1) thus 

 generated (115) has proved useful in many ways. For 

 one thing, knowledge has started with man's sensory 

 experience, roughly at the individual level and in 

 the descriptive structural column; it has then spread, 

 .is .1 series ni expanding semicircles, horizontally to 

 function and development and vertically to constit- 

 uent or collective units. 1 his is exemplified in the 

 history of our knowledge o) the nervous svstem which 

 has moved from structure to function to development, 

 and from i^niss anatomy to gray centers to neurons, 



111 membranes and Xissl granules and molecules. 

 Modern concern with neural organization has moved 

 from the structural to the dynamic to the learned, as 



appeals also in ihis chapter. 



1 1 xi iiiixvi 1 mis. The structural entities, in the 

 first Column, regularly obtain earliest attention, the 



functional ones, in the second column, appear with 

 advancing sophistication. Neurophysiology clearly 

 falls in the second column ami extends from cell to 

 individual, trailing into neui 01 liemistrv at the 

 molecular level and heading into individual and 

 group psychology at higher ones Social roles, indi- 





