The Interneuron

By Mary A. B. Brazier

This title is part of UC Press's Voices Revived program, which commemorates University of California Press’s mission to seek out and cultivate the brightest minds and give them voice, reach, and impact. Drawing on a backlist dating to 1893, Voices Revived makes high-quality, peer-reviewed scholarship accessible once again using print-on-demand technology. This title was originally published in 1969.

• Language: English
• Publisher: University of California Press
• Release date: Nov 15, 2023
• ISBN: 9780520324268

Mary A. B. Brazier

Enter the Author Bio(s) here.

Book preview

UCLA FORUM IN MEDICAL SCIENCES

VICTOR E. HALL, Editor

MARTHA BASCOPÉ-EsPADA, Assistant Editor

EDITORIAL BOARD

Forrest H. Adams William P. Longmire

Mary A. B. Brazier H. W. Magoun

Carmine D. Clemente C. D. O’Malley

Louise M. Darling Sidney Roberts

Morton L Grossman Emil L. Smith

Reidar F. Sognnaes

UNIVERSITY OF

CALIFORNIA, LOS ANGELES

THE INTERNEURON

UCLA FORUM IN MEDICAL SCIENCES

NUMBER 11

THE INTERNEURON

Proceedings of a Conference held September, 1967

Sponsored by the Brain Research Institute, University of California, Los Angeles

EDITOR

MARY A. B. BRAZIER

UNIVERSITY OF CALIFORNIA PRESS

BERKELEY AND LOS ANGELES

1969

EDITORIAL NOTE

The present volume contains the proceedings of the first in a series of conferences on Neural Interaction, organized by Arnold B. Scheibel, Carmine D. Clemente, Mary A. B. Brazier and John D. French of the Brain Research Institute, UCLA School of Medicine.

Acknowledgement for the support of this conference is owed to the following: Abbott Laboratories; Hoffmann-LaRoche; Merck Sharp & Dohme; Sandoz Foundation; Schering Corporation; Smith, Kline & French; E. R. Squibb & Sons, Upjohn Company and Wallace Laboratories.

CITATION FORM

Brazier, M. A. B. (Ed.), The Interneuron. UCLA Forum Med. Sci. No. 11,

Univ, of California Press, Los Angeles, 1969

University of California Press

Berkeley and Los Angeles, California

University of California Press, Ltd.

London, England

© 1969 by The Regents of the University of California

Library of Congress Catalog Card Number: 69-16504

Printed in the United States of America

PARTICIPANTS IN THE CONFERENCE

ARNOLD B. SCHEIBEL, Chairman

Brain Research Institute,UCLA School of Medicine

Los Angeles, California

MARY A. B. BRAZIER, Editor

Brain Research Institute, University of California

Los Angeles, California

W. Ross ADEY

Brain Research Institute, UCLA School of Medicine

Los Angeles, California

PER ANDERSEN

Neurophysiological Institute, University of Oslo

Oslo, Norway

THEODOR W. BLACKSTAD

Normal-Anatomisk Institut, Aarhus Universitet

Aarhus, Denmark

THEODORE H. BULLOCK

Department of Neurosciences, University of California San Diego

La Jolla, California

CARMINE D. CLEMENTE

Department of Anatomy, UCLA School of Medicine

Los Angeles, California

JOHN D. FRENCH

Brain Research Institute, UCLA School of Medicine

Los Angeles, California

M. G. F. FUORTES

Laboratory of Neurophysiology

National Institute of Neurological Diseases and Blindness

Bethesda, Maryland

SUSUMU HAGIWARA

Scripps Institution of Oceanography

University of California San Diego

La Jolla, California

VICTOR E. HALL

UCLA Forum in Medical Sciences, UCLA School of Medicine

Los Angeles, California

G. ADRIAN HORRIDGE

Gatty Marine Laboratory, University of St. Andrews

Fife, Scotland

MASAO ITO

Department of Physiology

University of Tokyo Faculty of Medicine

Tokyo, Japan

ERIC R. KANDEL

Departments of Physiology and Psychiatry

New York University School of Medicine

New York, New York

DONALD KENNEDY

Department of Biological Sciences, Stanford University

Stanford, California

I. M. H. LARRAMENDI

Department of Anatomy

University of Illinois College of Medicine

Chicago, Illinois

CHAN-NAO LIU

Department of Anatomy and Institute of Neurological Sciences

University of Pennsylvania School of Medicine

Philadelphia, Pennsylvania

RODOLFO R. LLINÁs

Department of Neurobiology, Institute for Biomedical Research

American Medical Association—Education and Research Foundation

Chicago, Illinois

ANDERS LUNDBERG

Department of Physiology, University of Göteborg

Göteborg, Sweden

DAVID S. MAXWELL

Brain Research Institute, UCLA School of Medicine

Los Angeles, California

DONALD M. MAYNARD

Department of Zoology, University of Michigan

Ann Arbor, Michigan

DOMINICK P. PURPURA

Department of Anatomy, Albert Einstein College of Medicine

New York, New York

MADGE E. SCHEIBEL

Los Angeles, California

JOHN D. SCHLAG

Brain Research Institute, UCLA School of Medicine

Los Angeles, California

ROBERT F. SCHMIDT

II. Physiologisches Institut, Universität Heidelberg

Heidelberg, Germany

José P. SEGUNDO

Brain Research Institute, UCLA School of Medicine

Los Angeles, California

STEN SKOGLUND

Department of Anatomy, Karolinska Institutet

Stockholm, Sweden

WM. ALDEN SPENCER

Department of Physiology, New York University School of Medicine

New York, New York

COSTAS STEFANIS

Department of Neurology, Athens National University

Athens, Greece

L. TAUC

Laboratoire de Neurophysiologie Cellulaire

Centre d’Études de Physiologie Nerveuse

Centre National de la Recherche Scientifique

Paris, France

C. A. G. WIERSMA

Biology Division, California Institute of Technology

Pasadena, California

WILLIAM D. WILLIS, JR.

Department of Anatomy

The University of Texas Southwestern Medical School

Dallas, Texas

FOREWORD

Within the frame of contemporary neurophysiology, a conference devoted to the interneuron seems a natural choice. For with the advent of the intracellular microelectrode and recognition of the hyperpolarizing potential as an invariable concomitant, if not causative agent, of postsynaptic inhibition, the presynaptic progenitor of this effect has invariably been identified as a local circuit cell—an interneuron—in the immediate neural environment. A number of different cell families have been implicated in the development of this effect. Internuncials in the intermediate spinal gray, so-called Renshaw cells in the ventral motoneuron pool, basket cells in hippocampus and cerebellum, and stellate cells in neocortex are among those cited. All of these elements are believed to share certain qualities which distinguish them from other neurons. The most important of these are the presumably local (short-axoned) nature of their axonal trajectory, their characteristically intercalated position between long-axoned projection neurons, and their functional capacity for reversing the sign of excitation from facilitation to inhibition.

Selection of elements such as these to serve as chapter and verse for this symposium immediately meets with difficulties. In the first place, their number forms but a fraction of that grand ensemble which rightfully deserves the name of Interneuron. Hie definition of Bullock & Horridge (1) suggests panoramic diversity:

Interneuron. An internuncial neuron; one which is neither sensory nor (purely) effector-innervating, but connects neurons with neurons…

At best, this definition allows us to discard from consideration first order sensory cells, motor neurons, and possibly a few central elements whose peripheral processes run a centrifugal course to control gain levels at the receptor stage. All others, and this must constitute the great majority of elements of the vertebrate nervous system, fall within our category.

For over three quarters of a century the short-axoned (Golgi II) internuncials have seemed to constitute a cell type in search of a function. But, despite the undoubted advantages implicit in their identification as the obligatory, intercalated interneurons mediating inhibition, facts have begun to handle the theory unkindly. For one thing, the Renshaw cell, long considered the paradigm of this type, can no longer be accepted as a short-axoned cell. If indeed there is a special category of spinal neurons responsible for the burst-type firing and IPSPs generated concomitantly in local ipsilateral motoneuron populations, structural data demand that they be long-axoned projection cells, similar in all respects to other proprioneurons. The same reservation applies to interneurons more dorsally located and implicated in the generation of primary afferent depolarization. For another, the conceptual restraints imposed by Dale’s thesis no longer seem so binding in the light of recent data pointing to mediation of both excitatory and inhibitory functions by the same synaptic substance. With the development of the receptor-mosaic notion of postsynaptic membrane organization, so far shown only in invertebrate material, it is true, the presence of the intercalated sign-reversing cell seems less urgent. In fact, careful study of spinal cord sections with degeneration and Golgi methods now suggests that some loops of known inhibitory function may well be direct rather than interneuron- mediated.

With the advantages of exclusivity already under attack, how wide must we enlarge our set so that no worthy candidate is omitted? Following the definition of Bullock and Horridge, all elements beyond the level of the first order sensory cell, and prior to the final common pathway motoneuron must be considered. This includes most spinal neurons, the brain stem axial core, rostral sensory projection systems, the great descending motor paths, pyramidal and extrapyramidal, and all cortical tissue including cerebellum and cerebral hemispheres. Clearly we have invoked a situation demanding consideration, not of a set, but virtually the set of all possible sets.

In similar fashion, the category of functions commonly attributed to interneurons will no longer suffice. In the course of phylogenetic development, only the protozoa, mesozoa, and porifera seem without identifiable nervous systems. The coelenterate nerve nets are made up exclusively of sensory and motor components, fashioned in random nets, some of which are through- conducting, and others non-through-conducting. But, even in this primitive matrix, it has been found that activity in the non-through may connect directly or progressively to the through depending on frequency and (or) repetition rate of the stimulus. Perhaps in this plastic relationship between the two nets where output relates to, but is not precisely the same as, input, we can recognize an Anlage of interneuron function. The next phy- letic step, to platyhelminthes presents at once the prospect of a complex ganglionated chain and, with it, of cells whose axonal course is run entirely within ganglia and connectives. Suddenly, many of the familiar characteristics of vertebrate reticular neurons are present including bifurcating axons running long distances rostral and caudal along the chain, and dendrite-like structures immersed in afferent neuropile. And, in successive invertebrate phyla built to this scheme, a bold functional plan emerges. Interneurons are more than the bridge between input and output. They are repositories of the output pattern for which the afferent signal is but the trigger. The sensory input serves only a flip-flop function and the master plan, already coded into the interneuron and its connectivity pattern through the effectors, is ac-

tivated in invariant form once the sensory input switches to yes. Here is a class of interneuron activity directly on the route, not just to sign change or relay function but to the most sophisticated strategy-forging levels of cortical operation in the brain.

With progressive enrichment of the set, both in number and in function, it becomes increasingly clear that a symposium on interneurons should properly encompass almost the entire phyletic series, and the vertebrate nervous system from conus to cortex. The range of communications which follows gives substance to this belief.

M. E. and A. B. Scheibel

REFERENCE

1. BULLOCK, T. H., and HORRIDGE, G. A., Structure and Function in the Nervous Systems of Invertebrates. Freeman, San Francisco, 1965: p. 1602.

This slide was prepared personally by Santiago Ramón y Cajal and was presented to

the Rrain Research Institute by Professor Clemente Estable of Montevideo, Uruguay.

The sections were impregnated with silver according to the techniques of Golgi and,

as can be seen by the label in Cajal’s own handwriting, are from a "Cat of 15 days,

acoustic cortex. The bue is short for bueno", which Cajal used to label his best

preparations.

Contemporary photomicrograph made from slide on facing page, showing a cell

in the second section in the left column. It shows a short-axoned neuron of the Golgi

type at about X 740 magnification. These cells are among those in the cerebral cortex

generally considered to be interneurons.

CONTENTS 1

CONTENTS 1

THE INTERPRETATION OF BEHAVIOR IN TERMS OF INTERNEURONS

THE CONTROL OF OUTPUT BY CENTRAL NEURONS*

EXCITATORY AND INHIBITORY PROCESSES

THE ORGANIZATION OF SUBPOPULATIONS IN THE ABDOMINAL GANGLION OF APLYSIA

REGULATIVE MECHANISMS FOR THE DISCHARGES OF SPECIFIC INTERNEURONS§§§§

REFLEX MATURATION

A STRUCTURAL ANALYSIS OF SPINAL INTERNEURONS AND RENSHAW CELLS*******

SPINAL COBD AFFERENTS: FUNCTIONAL ORGANIZATION AND INHIBITORY CONTROL

CONVERGENCE OF EXCITATORY AND INHIBITORY ACTION ON INTERNEURONES IN THE SPINAL CORD§§§§§§§§§§

THE LOCALIZATION OF FUNCTIONAL GROUPS OF INTERNEURONS

ELECTRON MICROSCOPIC STUDIES OF CERERELLAR INTERNEURONS

NEURONS OF CEREBELLAR NUCLEI

FUNCTIONAL ASPECTS OF INTERNEURONAL EVOLUTION IN THE CEREBELLAR CORTEX

THE NERVE CELL AS AN ANALYZER OF SPIKE TRAINS‡‡‡‡‡‡‡‡‡‡‡‡‡‡

STUDIES ON THE HIPPOCAMPUS: METHODS OF ANALYSIS

PARTICIPATION OF INHIBITORY AND EXCITATORY INTERNEURONES IN THE CONTROL OF HIPPO- CAMPAL CORTICAL OUTPUT†††††††††††††††

INTERNEURONAL MECHANISMS IN THALAMICALLY INDUCED SYNCHRONIZING AND DESYNCHRONIZING ACTIVITIES

INTERNEURONAL MECHANISMS IN THE CORTEX††††††††††††††††

SUMMATION

NAME INDEX

SUBJECT INDEX

THE INTERPRETATION OF BEHAVIOR IN TERMS

OF INTERNEURONS

G. ADRIAN HORRIDGE

St. Andrews University

Fife, Scotland

The aim of neurobiology is the explanation of behavior, and microelectrode techniques now provide the possibility of explaining it at a new level in terms of units which we believe correspond to the actual units of the nervous system.

I propose to deal with some implications of this new emphasis. In the physical sciences, vast expense is justified by the expectation of securing an explanation in terms of sub-units of the next size down; chemical properties in terms of molecules, molecules in terms of atoms. From experience we know that this type of explanation does not provide complete answers, or even any answers to many of the outstanding questions, but it makes old questions redundant, provides a particular style of mechanistic understanding, leads to new useful experiments, and sometimes reveals methods whereby we can exert control. Other kinds of explanation are less likely to be fruitful in these valuable ways. Further, to be most effective, the mechanistic explanation must be in terms of units which correspond to real structures, although, like chromosomes or atoms, these units may turn out later to be both divisible and mutable.

BACKGROUND

As units in the nervous system we have ready to hand the anatomical neurons of Ramón y Cajal and the physiological all-or-none properties which were worked out historically for the heart, the jellyfish bell, single axons, vertebrate fast muscle fibers and then parts of many other excitable cells. It was a great temptation to endow the anatomical neuron with the physiological all-or-none activity, and call the hypothetical product the functional unit of behavior. Circuits of these units could act as models with some of the properties of nervous systems. Examples are found in the study of interneurons in optokinetic responses by Lorente de Nó (1), in papers by Pitts & McCulloch (3), and by a few even to this day.

Studies with intracellular microelectrodes on a few advantageous preparations have now brought a profusion of variables so that many physiological properties of units are explained by changes at membrane level. The variables are potentials of all shapes and sizes; the mechanisms are changes in permeability to various ions; the effects are inhibitions and excitations, immediate, delayed or rhythmic.

So far as events at membrane level control those at the neuron or unit level, all are inhibitory or excitatory. All are summed in time and in space upon the branches or soma of each neuron by virtue of the electrotonic spread as governed by the time and space constant of the piece of neuron considered. To build up behavior from this is like writing mathematics with limited symbols such as plus, minus, zero and a few digits. We know that success depends upon close attention to the spatial pattern of the symbols or units. What is written at one place in such a system has little significance without the rest of the program. In the nervous system the same dependence on the whole pattern holds. If we are to have explanations of critical choices and operations in small parts of behavioral patterns they must ultimately be expressed in terms of local excitations and local inhibitions in the anatomical context of the active nerve fibers, and in order to exclude alternative explanations the pattern must be known in detail in space and time. This adds an immense anatomical problem, but is the only realistic analysis, in contrast to blanket explanations with black boxes, and it must be stressed over and over again that the principal causal factor seems to be the spatial geometry of the actual synapses and fibers. The problem of analysis is easier in two circumstances which are well worth examination—in very simple examples where all alternatives might easily be explored, for example, the ear of a moth with two receptor cells, and in very specialized examples where the larger part of the system serves a single known function, as in the auditory system of bats. But for all situations in nervous systems a great deal of microanatomical description is required before any exclusive explanation of behavior can emerge.

THE PROBLEM

The task involves description of two kinds, (a) anatomically observed connectivity patterns and (b) causal relationships as revealed by the activity of single units. Although we may expect that these refer to the same substrate of investigation, experience warns us to distinguish carefully the actual phenomena seen with different techniques from the inferred entities and their supposed relationships.

A unit is a physiological entity defined as the apparent output of a supposed anatomical neuron or axon, and is often recognized by the all-or-none properties of an active propagated response (called a spike); however, we are now moving into a period when units are not necessarily recognizable by spikes. For example, first and second order cells on the visual pathway of insects and vertebrates apparently have only graded non-propagated potentials. Numerous small neurons in all groups of animals may turn out not to have spikes but to work by electronic transmission over small distances. Therefore they will have to be analyzed as units. Furthermore, a neuron, defined anatomically as a cell, may have several regions each of which can act as a separate unit. To complicate the issue still more, where several neurons interact in a compact cluster and operate without spikes, neither the anatomical neuron nor the physiological unit is appropriate to understanding the nature of the system of interactions. For example, electron microscopy reveals that in the neuropile of the vertebrate retina or cortex, or in any invertebrate neuropile, the proximity and serial occurrence of the synapses provides a situation where simultaneous complex electrotonic interaction between many inputs may prevail. The ommatidium of Limulus and the optic cartridge of arthropods are other examples of tightly interacting groups of units. The responses of the output neurons as units are observable but are not necessarily predictable from synaptic potentials or other membrane phenomena of any one cell in the group. Despite the complexity of the interactions which bear upon the spike-initiating process, however, the information carried by the spike is the justification for the examination of the output at the unit level.

‘Connectivity’, like ‘synapse’, is a term based on either functional or anatomical observations, or both, in every individual instance. As favored by some, for a mathematical relationship, or by investigators of vertebrate CNS for inferred physiological or synaptic interaction, connectivity is a functional term. As favored by others it means what is connected to what, in actuality, by anatomical synapses which can be mapped. Apart from serial sections under the electron microscope we have as yet almost no sure tools with which to study anatomical connectivity. The relatively gross anatomy of which tracts run over or under, which layers contain what fibers, the density, size and shape of neuron branches, and so on, is of little consequence so far as we know. If the feature of interest is what is connected to what, the pattern of axon terminal branching or dendritic tree matters little, although particular growth patterns may represent easy ways of establishing the required anatomical connectivity. I wish it were otherwise, but almost the whole of neuroanatomy of all classes of animals needs to be reworked in terms of anatomical connectivity. The anatomy of neurons revealed by the methylene blue or Golgi techniques even today takes us little further than Cajal or Retzius was able to see. At the detailed level of synapses the anatomical complexity is so great that a complete description is out of the question. Only carefully chosen areas of critically relevant neuropile can be described, and one must hope that they are representative of similar areas under the electrode in physiological experiments. At a gross level this is adequate but the correspondence inevitably breaks down at some level. The question is whether the variability between samples of the same critical region from different animals allows an exact analysis of the physiological results.

Electrophysiology employed alone leads to the inference of physiological pathways and these can be equated exactly with individually identified neurons only in a few relatively simple examples which are mostly giant neurons. In the vertebrates the problem of equating any single physiological pathway with any single anatomical connectivity pattern at unit level is that the same neurons (except Mauthner fibers) cannot be found twice. Who has marked a vertebrate neuron at the conclusion of recording its normal activity, found the part it plays in behavior and where it lies in the pathway, then identified its connections anatomically? The point reached in vertebrate studies is the analysis of the numerically most abundant types of physiological unit and a partial equating with the most abundant types of anatomical neurons. This can be most readily achieved where there are central projections of the body senses or muscles. We prefer to build models and draw arrows between boxes in two dimensional diagrams of a typical digested subsystem which ignores the individuality of the units. The boxes are sometimes drawn as nerve cells.

A central nervous system is always marked by a choice of action that is dependent upon its own structure in interaction with the whole history of sensory impressions, and by a behavioral output that is highly adapted to the animal’s environment. This adaptive choice of alternative activity patterns distinguishes a CNS from a cardiac, sympathetic, or other peripheral ganglion. Central nervous systems of all animals consist of three components, terminals of axons of sensory neurons, dendrites of motoneurons, and thirdly those neurons which lie between sensory and motor—the interneurons. The proportion of the latter rises with increased behavioral complexity. Interneurons are not just interpolated small inhibitory black boxes on physiological pathways. Beyond this point many definitions do not hold for situations such as nerve nets, outlying ganglia or ciliated cells.

The study of interneurons has in practical terms become the interpretation of repeated sampling of units by microelectrodes and the description of all experimentally controlled factors which appear to modify their changes in spike frequency. On the sensory side this has led to the important concept of the receptive field. For each type of stimulus available, the receptive field is the plot of contours of equal stimulus intensity which give selected standard strengths of response. For chemical stimuli, each experimentally applied stimulating substance has to be treated as if it lies in a different dimension. The receptive field is the natural way of expressing what a neuron responds to, and the present effort is largely directed to the full description and classification of the receptive fields. For the numerous units which are spontaneously active, or which have non-specific inputs, the problems of classification are very great. We also have to face the possibility that every neuron in every animal is in some way unique. The classification into categories is a human means of ordering the data, but a particular classification scheme may be unrelated to the animal’s organization and will therefore subsequently impede progress until superseded.

The uniqueness of neurons adds to the fascination of the problem. On account of the geometrical arrangement of sensory neurons on the body, no two sensory units have identical receptive fields. In the retina, in the skin, and along the cochlea, receptors are distributed in particular spatial patterns so that no two can be identical. In vertebrates, even in the chemical senses, it is an experimental finding that every receptor and interneuron is in some small way unique. In vertebrates the ascending sensory interneurons form classes with beautiful progressive changes in receptive field as we pass from unit to unit at one level within a single nucleus, but it is a matter of experience that no two interneurons are exactly duplicated. Considered anatomically, every neuron in the central nervous system has its dendritic field in a different place, and must have its own pattern of anatomical connectivity: every effector neuron runs to its own differently situated muscle or gland cells.

Although every analysis has classified units into classes, advance has been most rapid in those systems where each unit is provided with a unique label because it occupies its own position, as in a projection from the retina, the cochlea or from the body surface. Apart from its place in the projection, most vertebrate interneurons are not described in an individually recognizable way by electrophysiological studies. The reason for shunning the issue of uniqueness is that in vertebrates there are as yet no techniques for returning to work over and over again on the same neuron, identified even by receptive field together with projection. This lack of techniques becomes crucial in the experimental study of the growth of neurons that are in process of forming new connections. For example, the distinction between fast and slow vertebrate motoneurons made possible the analysis of how each group connects with and modifies the properties of its own type of muscle fiber. Comparable analysis on central neurons can be made only to the degree of detail which the recognition of individual differences permits. The interneurons recognize each other, as proved by their formation of synapses in particular ways. We have not succeeded in distinguishing the basis of their specificity even after they have accentuated their differences as cells by their establishment of unique anatomical and functional connections.

If we had the facility to extend a probe into a distant city and record individual conversations (but not purpose, motive or thoughts) of the inhabitants at random, it would be possible, in time, to gain a good deal of insight into the activity of that city, and to classify even the most individualistic inhabitants. As an aid to understanding mechanisms, the classification would be the essential first step only, and many types of classifications would assist little in the identification of causes and effects in the city’s affairs. A street plan would be some help; a map showing the paths of movement and possible communication points of every individual might seem desirable for analysis, but too complex to be useful if secured. The problems of learning anything about behavior from interneuron studies are in some ways similar.

This analogy immediately suggests that the most interesting units are likely to be numerically very rare but to have widespread effects. They will be the ones most likely to be omitted from preliminary classifications of numerically numerous types whose actions cannot be understood without them. Vast amounts of work upon irrelevant facets can easily be entirely wasted and yet tiny details can turn out to be of great significance. The problem lies in the choice of features which will be found to be relevant.

Low INFORMATION CONTENT OF BEHAVIOR

Apart from feats of human memory, and the specialized actions of primates and a few other mammals such as elephants, an act of behavior in an experimental test of a cat, rat, octopus or crab is relatively simple. The total number of choices described in all psychological literature could be punched on comparatively few yards of tape. Although justly regarded as marvelous, the behavior of most animals is strictly adapted to the environment in which they live, and special-purpose machines to perform these tasks would not require many components. I feel competent to say this because I am a zoologist and I strive to make the following point. The complexity of the nervous system seems out of all proportion to the job it has to do. A small worm with little interesting life operates with about 100,000 neurons; a sea-anemone has many more. The complexity of impulse patterns and possible cross-correlations in neurons, even in lower animals, is astronomical, like the detail of leaves in woodland undergrowth. There is literally no end to the potential analysis of it. Most of the detail and correlations will therefore not be relevant to the required explanation. Yet for medical reasons alone we must strive to understand the basis of action by the nervous system. The important topic for armchair discussion is how to proceed without waste of effort upon details which later are proved irrelevant.

SUBCATEGORIES OF BEHAVIOR

For a long time now a number of terms have been employed in explanations of that highest category, behavior. They are words like arousal, pattern abstraction, sign stimulus, innate behavior pattern, selection of alternatives, attentiveness, reflex, efferent copy, reinforcement, learning. Experimental analysis of behavior is full of such terms. All can be defined satisfactorily; many are open to experimental analysis, but all carry the expectation of being underpinned by a lower category of entities. They all need to be explained at neuronal or unit level. The point is that these terms referring to the behavioral level are very obviously abstractions of the human mind; they are not necessarily components of real mechanisms in the animal. We should therefore not be disappointed that when we start to explain this kind of concept at a lower level we find that each of our magnificent comprehensive terms covers a multitude of possible mechanisms. Gunpowder and ice have in common an ability to change under mechanical pressure, but this is hardly explainable by a mutability which they hold in common. Detailed analysis of each, in terms of component units, leads to a different world of interactions, where the mutability feature is not lost but is given less consideration.

Arousal is a term which conjures up the vertebrate (or perhaps only mammalian) reticular system, and we might ask whether arousal is necessarily nonspecific. In neuronal terms, if there are particular units which, when excited, arouse others, are they not specific? If they are nonspecific with reference to the stimulus, does that mean a wide receptive field or inability or failure to explore the receptive field? What is the total receptive field of that arousal neuron? What are its total connections and when is it active? Well, of course usually we do not know. It is about as far as we can go to record an electrical or neuronal correlate of arousal at all. There is therefore a temptation to hang on to the term for as long as possible before substituting half a dozen types of spatial summation at terminals of arousal axons with known pathways.

Pattern abstraction, as a property of higher order sensory interneurons, is still an exciting new discovery. Some interneurons respond to warble notes falling between certain sound frequencies, others to movement in a particular direction irrespective of contrast. These stimuli are features of the environment of interest to the animal. The responses of interneurons are explicable as the consequence of particular patterns of summation and inhibition of earlier order neurons which impinge upon them. With hindsight it is easy to say that pattern selection is inevitable wherever dendrites have excitatory and inhibitory terminals impinging on them from several sources. To stress that every unit selects a pattern is no longer novel; however there is a considerable amount of spade work still to do, even after the foundations are drawn out. Only recently has it been demonstrated that the directional discrimination of a sound is a type of pattern selection that is made possible by the summation and inhibition of excitation from the two ears upon a common interneuron. Different higher order interneurons select sounds with different time delays between the two ears. There is a population of these neurons such that sounds which originate in different directions are signaled along separate pathways. From this arises a line-labeled representation of the outside world—not necessarily a topographical map.

Many other fascinating examples of the same kind wait to be worked out, but why should they be explored if they really are of the same kind? The answer is not because they are there, or because Ph.D. students require topics, but because new principles of explanation in terms of component units emerge only from analysis of those very components. This can only be done by hours of recording of responses to thousands of test stimuli, as the receptive fields are mapped out in all dimensions. During the course of this work, for instance, some of the well-known sign-stimuli of behavioral studies may prove to be examples of pattern abstraction by single interneurons. The in tellectual challenge, however, comes from those units (which are airead)’ known to be common in all animals) where the receptive fields do not correspond to the obvious structure of the environment, e.g., multimodal units.

The reflex as a unit of behavior is easily recognizable as an appropriately adaptive movement in response to a normal environmental stimulus. In terms of interneurons it implies pattern selection on the input side and further selection of patterns of interneurons by motoneurons whereby the appropriate set of motoneurons are excited. Traditionally we are led to believe that in vertebrates the running commentary provided by the receptors of joints, skin and muscles, provides a continuous control of the course of the movement. In arthropods however, where the critical experiments are possible, the general rule is that reflex movements are the result of centrally determined patterns of motor impulses that are substantially unchanged by the removal of all proprioceptors, periphery and musculature. The great change of emphasis in the past five years made by workers on arthropods justifies illustration by a few examples, because it is no longer possible to begin with the assumption that coordinated responses (at least in invertebrates) are reflexes that are dependent upon continuous sensory monitoring of the environment including their own consequences. Ventilation of the trachea, oscillation of the wings in flight, song of insects and swimming movements, are each the product of a central program of the central ganglia. Moreover, where a rhythm is modified by sense organs these do not necessarily have an effect at the instant when they are excited. As shown by Wilson and his colleagues (5), wind blowing against hairs on the head of a flying locust causes a flow of impulses in second order neurons down the neck nerves to the thoracic ganglia. While these fibers are active, at more or less any frequency or pattern, regular bursts of motor impulses oscillate the wings through the direct wing muscles. Even when the wing muscles have been removed these bursts of impulses resemble those in normal flight. In addition, proprioceptors can influence the efferent pattern. Near the hinge of the wing is a large sensory cell which increases the frequency of the wing beat and shifts the position of certain motor impulses to two of the muscles. By this shift in the position of an impulse in each cycle, lift is added at the time when it is most effective. The receptor fires only once at the top of each wing cycle and its impulse has no immediate effect, acting only when summed over many cycles. Immediately acting reflexes also contribute; for example, contact of the insect’s feet with the ground stops the flight rhythm at once and orientation reflexes control turning. There is, therefore, a centra] rhythm with peripheral control of some aspects. As in all animals, the response as a whole is a result of summation of the mechanical effects of different muscles at the gross anatomical level.

A more complex but equally clear central rhythm controls horizontal optokinetic eye movement of the crab. The eyes of the crab are on the ends of movable stalks which can follow the movement of a striped drum rotating around the animal. One eye moves towards the midline, the other away from it. The eyes follow slowly for 10° to 15°, then flick back to start again. Nine muscles, each with fast and slow motor axons, and phasic and tonic muscle fibers participate in this regularly repeated movement, which is called the optokinetic response. The eye moves steadily across the orbit because the motor impulses to one of the muscles slowly change in frequency. For each motor frequency the eye is pulled over to an equilibrium position which is determined by the balance between the muscle tensions and the elastic recoil.

The forward movement and periodic fast return phase are due to centrally determined programs which depend only upon the history of the total apparent movement seen by either eye. A large movement of all contrasts in the visual field causes a large movement of the eye, during which the sequence of slow forward and fast return phases may run through several cycles. However, for a given stimulus relative to the eye the response is the same whether or not the eye is allowed to move. Numerous tests on the immediate control of eyestalk position revealed no evidence of utilization of proprioceptors. The only information which the animal has about the direction in which the eye points at any moment appears to have been converted into the motor impulse pattern to the nine eye muscles at that moment. A vertical or a horizontal movement of contrasts in the visual field, or a tilt of the crab in any plane adds a little excitation of the appropriate motoneurons and, if it is free to move, the eye moves in the appropriate direction. If it is not free, the change in the motor impulse pattern takes place nevertheless. The central program is predetermined to bring the eye to the right place for every relevant stimulus situation. Suppose the input is an instantaneous instruction to move 20° to the right, but the eye can move only 12°: it will flick back and then complete the remaining 8⁰. This is achieved by an entirely central mechanism with no proprioceptive feedback, and the motor output, including the flick-back, is unaltered by clamping the eye.

Recently I have been looking at two aspects of this system. If the eye’s traverse is caused by a pattern of central origin it is difficult to see how the positions of the ends of the traverse agree with the shape of the socket. The motor impulse frequency of the pertinent muscle is not constant from crab to crab. The new relevant observation is that when the edge of the eye socket is repeatedly stimulated while the optokinetic response is going on, the fast flick-back soon begins its onset at a position which is earlier in the cycle. This has the consequence that the eye does not move so far in its slow phase across the socket. The change in the onset of flick-back lasts for many minutes whereas the shock treatment lasts for a few seconds. The direction of this long-term change in the central program is such that an eye which touched the edge of the socket would eventually be limited to a more centrally situated track. The brain utilizes no proprioceptive information for the immediate control of onset of flick-back. The long-term adaptive change of the central pattern, however, could provide the necessary adjustment to a new shape at each moult of the cuticle. It is a type of learning, without association but with reinforcement that is appropriate for trimming the central pattern to suit the animal’s anatomy. The adaptive plasticity of the response is therefore no more than one more item which is written into a centrally determined program.

The other new feature of interest appeared during a study of the protective eye withdrawal, in which the eye is pulled back into its socket. As the eye extends to its former position it sweeps across contrasting objects in its visual field. The question is whether the crab distinguishes between real movement and the apparent movement which its own eyecup extension creates. This is an old problem which von Holst tried to solve by supposing that every efferent signal is accompanied by an efferent copy which is a signal that neutralizes the expected effects of the motor output. In the crab a right eye in process of being voluntarily extended after a protective withdrawal certainly continues to see. This is proved by the observation that if the left eye is blinded both eyes continue to respond to small background oscillations of a surrounding striped drum while the right eye is in process of extending. A voluntary extension of the right eye fails to drive the blind left eye by the amount of apparent movement which the extension generates. When the right eyecup is slowly pushed in a forced extension, however, the blind left eye makes a movement in the opposite direction, which is certainly caused by the apparent movement that is generated. There is therefore a difference between the voluntary and the forced cases. But when the relative movement caused by a voluntary extension of the right eye is forcibly prevented, either by holding the eye or blinding it, the left eye makes no movement. Therefore no actual compensatory efference copy is passed to the other side.

The inescapable conclusion from these and many related experiments was that the crab somehow has learned, during the course of many spontaneous eye retractions, to distinguish and eliminate the central effect of its own eye movement, and makes use neither of efferent copy nor of proprioceptors.

Although the fixed programs are largely inherited, some of their detail and, in particular, compensation for their own movements, may well be later determined by flexible phrases in a central program. To a complex animal with many sense organs in a varied environment, the effects of its own actions must frequently be unpredictable. Many lower animals have startle reflexes which are set off by sensitive receptors. Examples are the jump of the cockroach as triggered by an air current on the anal cerci, and the contraction twitch of the earthworm at the slightest touch. But these giant fiber responses are stilled when the animal itself moves. The automatic cancelling mechanism is not understood even in these simple examples, although it will probably turn out to be wholly central. One can predict that the motor output is accompanied by an inhibition of the giant fiber reflex even when, by section of motor nerves, the slow movement is not allowed to take place. In more complicated examples of control by a centrally determined program there must be many types of underscribed plasticity. I am in fact suggesting that the simple type of postural learning, with non-specific reinforcement, that I first described in the insect ventral cord, or which I have just mentioned in the control of the position of onset of the fast flick-back of the crab eye, could be a common mechanism of adjustment of otherwise centrally determined sequences. The central program contains phrases which are adjustable under the influence of inputs that are only remotely related in pattern to the output. I believe that once the technique of repeated recording of detailed central programs from groups of identified neurons is widely utilized it will lead to a new understanding of the part played by centrally fixed patterns of efferent impulses as components of reflexes and of conditioned reflexes, even in vertebrates. The central consistency in the face of variation in sensory input also helps us to ignore one great slice of nervous activity, the sensory excitation which the animal itself ignores. Plasticity controlled by nonspecific inputs evolves easily into learning mechanisms controlled by reinforcement neurons.

The electrophysiological analysis of arthropod movement teaches that each neuron can be individually named and each is a unique pattern perceiving neuron. The analysis of reflexes by recording from named motoneurons reveals a fixed central mechanism. Thinking of this in terms of interneurons suggests that reflexes and central rhythms grade into each other and each have many mechanisms at the neuronal level.

Learning, analyzed at the interneuron level, presents a range of new obstacles which arise from the difficulty of recording from the required units exactly when the learning occurs. Learning is a troublesome term from another method of observation and there is no reason to expect a single mechanism at a neuronal level. Learning is difficult to achieve in experimental situations where electrophysiology is possible and vice versa. If physiological changes are discovered at the time that learning occurs there is no obvious way of knowing whether they are primary causes or not; because units are so numerous the observed changes are almost certainly not primary causes. Moreover, if long-term physiological changes in a unit are suspected as significant, the constancy of the synaptic inputs to that unit and the absence of interfering small cells, must be demonstrated. In electrophysiological tests during a learning experiment the stimulus situation must be presented repeatedly while a number of units are tested. Therefore by the time the units of interest have been selected, they have been modified. At the critical spot in the nervous system, presumably upon part of a neuron, a crucial change in learning must occur at some time during the series of presentations; the problem is to anticipate by probing at that spot before applying the right stimulus. In a nervous system where every neuron can be individually identified, this is not a fundamental uncertainty principle after the Heisenberg model, because patient elucidation will eventually discover the unit of interest and make possible experimentation upon it alone before it is modified by repeated test stimuli. However, if units are not individually identifiable, the problem appears to be insuperable by a sampling process. Worse than this, there is no means of knowing what kind of primary change to expect. Changes of activity will certainly be found; perhaps the only way of testing whether they are significant is to introduce them artificially.

FEATURES OF INTERNEURONS IN INVERTEBRATES

The numerous interneurons which have been investigated in arthropods, particularly by Wiersma (4) in the crayfish, have the following general characteristics that can probably be extended to many invertebrate phyla.

Interneurons, identified by function, are constant in position in the ganglion or in location in a bundle. Many, which are inferred to be of lower order in the hierarchy, have receptive fields which are readily defined and constant. This is a strong practical justification for adhering to their classifi- ciation by receptive field. The receptive fields at higher levels are more complex and are explainable by taking sums and differences of lower order units. Some interneurons have inputs of one modality; others of two or more modalities. In the crayfish, about 10 per cent of interneurons can be identified, and counts of fiber numbers in the electron microscope show that 90 per cent are below the present size limit of recording with microelectrodes. Receptive fields with a geometrical regularity such as inhibitory surrounds or with exclusion-type interactions are rare in invertebrates, though common in mammalian sensory systems. The arthropod interneurons seem to function purely by selection of particular patterns of sensory input but, so far, it has rarely been possible to make sense of their responses in terms of the normal environment or of normal behavior. Every stimulus excites numerous interneurons but, strangely enough most stimuli have no other consequence, and, in particular, no overt behavioral effect. We must therefore infer that a behavioral output of interneuron activity is caused only by certain combinations of stimuli when these act for relatively long periods of time upon ganglia which are in a receptive state. This is unscientific until the receptive state is separately definable.

More difficult to interpret in terms of known interneurons is the general finding that motor activity in arthropods is largely governed by centrally determined sequences that are relatively independent of immediate sensory pattern. We are led to the conclusion that the interneurons are signaling when their part of the sensory combinations are appropriate for a central sequence to be emitted, allowing that some kind of receptive state is also essential centrally. The motoneuron pool seems to chum out programs like a jukebox but to be comparably deaf to feedback. In such a system we find novel types of plasticity of the detail of central sequences, and we observe diurnal rhythm in their execution but, in contrast, pattern selection by interneurons shows neither plasticity nor diurnal rhythm.

The large cells of molluscs have revealed other features of interneuron organization, as more fully outlined by Kandel in this symposium. A single cell with two axons can cause excitation at one terminal but inhibition at the other, apparently by the release of the same transmitter at each. This shows that the nature of the response depends on the postsynaptic cell. Different identified single neurons have quite different sensitivities to drugs and to transmitter substances. They sometimes have different ionic contents, different sensitivity to light and their own characteristic rhythm of activity. Identifiable pairs of cells, with one member on each side of the animal, are physiologically symmetrical where tested. In the case of one pair of cells, an obvious asymmetry proved to be only of the gross anatomical position of the cell body while the physiological pathways to these cells and their membrane properties were similar. As for anatomically symmetrical neurons in many groups of animals, it seems reasonable to conclude that the symmetry of connections in the two sides of the central nervous system arises from a similarity in the DNA-governed differentiation of the symmetrical cells. It is significant that in many animals symmetrical neurons often demonstrate their recognition of the metabolic similarity of their partner and form low resistance pathways from one to the other where they meet, whereas they are less likely to form such pathways with other neurons.

DIFFERENCES BETWEEN INVERTEBRATE AND VERTEBRATE INTERNEURONS

Invertebrate interneurons are in many cases individually identifiable, constant in number and receptive field, few of each type, accurately duplicated in different individuals of the same species, and with constant anatomical branching pattern. In contrast, vertebrate interneurons are not identifiable individually, and therefore appear to be less easily defined physiologically: they are more numerous, with many overlapping classes with reference to one stimulus. The receptive fields of invertebrate interneurons are constant and independent of stimuli applied elsewhere (with one or two exceptions); on the other hand, many types of vertebrate interneurons have receptive fields which are modified by excitation in parallel channels or by nonspecific stimuli of all kinds.

As we consider interneurons, in coelenterates, worms, molluscs and arthropods and in the classes of vertebrates, it is evident that evolution has taken the form of a progressively greater diversity of types, and that this is often achieved by greater detail of line-labeling. Occasionally there is an apparent simplification of pathways, as in the primate fovea, but in reality the wiring diagram is more specified, by an increased individual exactness of neuron connections in the higher form. There is greater diversity of interneuron classes and their receptive fields are more clearly definable, in more dimensions, in the higher groups of animals. Vertebrate sensory pathways often have a neural gain control system near the receptor terminals, and the higher order interneurons have regularly arranged inhibitory surrounds with a topotopical representation of the periphery in a spatial array of interneurons . Receptive fields in vertebrates are controlled at many levels by inhibitory feedback circuits. These features of the organization are not typical of invertebrates, where pathways of interaction are mostly linear sequences from receptor to effector. In invertebrates even the control of movement by proprioceptors, where it occurs, is sometimes not achieved by immediately acting feedback arcs to particular subsystems. Some of the above distinct features of vertebrates derive from the amacrine cells of vertebrate sensory pathways. All central nervous systems have nonspecific systems which ramify everywhere, but in vertebrates the specific systems have dendritic fields which are ordered in space, as in most of the nuclei of the mammalian central nervous system. In the invertebrates highly ordered dendritic fields are found only in optic lobes.

The meaning of this regular anatomical order in vertebrates is not known but it does not necessarily have any relation to synaptic interaction, in which only the anatomical connectivity pattern is relevant. My own theory is that in a group of interneuron pathways which represent some kind of sensory projection, the necessary distinction of pathways is more easily reached in growth if the terminals and dendrites are spatially separated by a growth pattern which produces a regular array. Certainly cells which are adjacent are those most likely to interact at the next higher level. If the endings representing a projection are all confused together, then each output fiber has a more difficult task of connecting functionally with only its own particular input fiber if the projection is to be maintained. On this theory the regularly repeated arborizations in a vertebrate nucleus are mainly devices to assist in establishing correct anatomical connections with a minimum of individual specifications from the parent cell bodies. Perhaps vertebrates are driven to utilize more obviously regular growth patterns because they have many more neurons in each class than do most invertebrates.

PATTERN ABSTRACTION

The outstanding general feature of interneurons that has emerged in the last decade is that all are units of pattern recognition. On account of this property of responding best to a particular class of inputs, each neuron is telling us whether the stimulus is relevant for it, and for all interneurons the stimulus is an excited cluster of neurons that are lower in the hierarchy. Because interneurons are so numerous, the overall relevance of a stimulus to the animal can be determined only by direct testing of all boundary conditions in all dimensions of the stimulus and by recording interneurons all down the physiological pathway. A stimulus configuration which has components of interest but is not quite acceptable may not excite at the final interneuron stage and, as far as concerns this particular response, might as well not be there at all in the early stages. Conversely all stimulus configurations which are just adequate, however varied, are treated by a responding interneuron as a single successful class. Small discrepancies in the stimulus are ignored so long as the unit’s threshold is exceeded. An animal with a range of many partially overlapping interneuron fields (called range fractionation) has therefore a much greater capacity to reject at a later stage what was mistakenly responded to by the earlier stage neurons. In consequence there is a high premium on range fractionation by large numbers of pathways in parallel at each level. Most excitation patterns still cause no response at the motor levels.

The above paragraph summarizes what neurons say. What they do not say is equally relevant. Interneurons as individuals do not make decisions as a result of some computation, either on a basis of the probable needs of the animal or on the statistical properties of the stimulus. They are not comparators; they do not pause to consider before a decision. Usually the size of their response is not an accurate measure of the message which they convey by being active. Despite numerous efforts to establish the importance of temporal pattern, the spacing of impulses does not seem to be particularly important in higher order units, although clearly paramount in auditory directional sensitivity and in rate of movement detection mechanisms. Interneurons respond to the here and now as it impinges upon them: even if modified by past experience they still respond only when their own threshold is exceeded.

A pattern classifying system of any kind which discriminates between all combinations of n inputs requires at least 2n indicating units to signal its result. Therefore the classification must be adapted to the patterns which are likely to be encountered. This is why animals with a complex behavior pattern are highly adapted to their environment. Lack of universal flexibility caused by paucity of pathways is one of the evolutionary factors which makes adaptation essential, and a restricted repertoire certainly reduces the number of necessary outputs. The auditory and visual pathways of a bird, for example, are extremely complex but the abstractions they make which are of interest to the bird appear to be relatively restricted. In fact the number of outputs indicated by behavior is always fewer than the number of inputs indicated by the sensory interneurons. This is a measure of the extent to which the environment contains predictable situations. Most animals, having freedom of movement, have selected the customary niche to which they are adapted and there they stay.

PROBLEMS OF CLASSIFYING INTERNEURONS

Technical difficulties are too great and numerous to be overlooked. Arthropods and perhaps some other invertebrate groups have unipolar central neurons with cell bodies which are electrically far from the synaptic regions. Therefore recordings must be taken from axons that are too small to be individually localizable in neuropile. Only relatively giant fibers have been identified anatomically as well as functionally; all examples of individually identifiable cells show a complete fixity of anatomy and function; the same neuron always does the same thing when picked up