MCCULLOCH, WARREN STURGIS

(b. Orange, New Jersey, 16 November 1898; d. Old Lyme, Connecticut, 24 September 1969),

neurophysiology, cybernetics, cognitive science.

McCulloch was a central figure in the early cybernetics movement. Cybernetics emerged from the interaction of anthropologists, mathematicians, neurophysiologists, physicists, psychiatrists, and psychologists with the financial backing of the Josiah Macy Jr. Foundation. Cybernetics analyzed biological, psychological, and social phenomena in terms of positive and negative feedback, communication, information, and computation. Cybernetic ideas have since diffused widely, but most notably into the fields of artificial intelligence, cognitive science, communication, and computer science.

McCulloch’s education ranged widely. At Carteret Academy in Orange, New Jersey, a private high school for boys, he devoted extra study to mathematics and Latin and also had a keen personal interest in poetry and secret codes. In the fall of 1917, he enrolled at Haverford College in Haverford, Pennsylvania, but with the outbreak of World War I, he transferred to Yale to be a part of the Yale Naval Training Unit. After a few months of active duty spanning the end of the war, McCulloch decided to remain at Yale. He graduated in 1921 with a major in philosophy. In 1923 he graduated from Columbia University with a master’s degree in psychology, then turned his sights to psychiatry at Columbia’s College of Physicians and Surgeons, graduating with an MD in 1927. McCulloch completed his internship and residency at Bellevue Hospital in New York City; he then returned to Columbia for a year to work in the Neurosurgical Laboratory in the Department of Neurology before returning to Bellevue for a year. He spent a year in graduate work in mathematical physics at New York University in 1931–1932 before serving a stint on the admission service at Rockland State Hospital in New York. What followed was a thirty-five-year career of scientific and philosophical investigation of the mind and brain.

The majority of McCulloch’s subsequent scientific career falls into three periods corresponding to the institutions where he drew his salary: the Laboratory of Neuro-physiology at Yale University’s Medical School; the Illinois Neuropsychiatric Institute at the University of Illinois Medical School; and the Research Laboratory for Electronics at the Massachusetts Institute of Technology (MIT).

At Yale . Johann Gregorius Dusser de Barenne was Sterling Professor of Physiology and head of Yale’s Laboratory of Neurophysiology from 1930 until his death in 1940. McCulloch joined Dusser de Barenne on a fellowship in 1934, but eventually worked himself up to the rank of assistant professor in 1940. During his years at Yale, McCulloch established himself as a neurophysiologist and authority on the sensory and motor cortex, working primarily with macaques and chimpanzees. An important component of his research concerned the functional organization of the brain. On the hypothesis that a particular region of the brain is dedicated to one specific function, say, motor control of the arm, the question of functional organization asks how that region is connected to other regions dedicated to other functions, such as motor control of the leg. Dusser de Barenne and McCulloch’s principal tool was strychnine neuronography. This technique involved applying a dilute colored solution of strychnine to the exposed surface of an animal’s cortex. The strychnine would induce distinctive, easily measured electrical activity in the brain region to which the poisoned neurons projected. This method was a great advance over degenerative techniques that involved performing one surgery to destroy a portion of the brain, then waiting several weeks to perform a second surgery to ascertain what pathways in the brain may have deteriorated as a result of the destruction. Clearly, much more data could be collected in one day with strychnine neuronography than could be collected in months with degenerative methods. Further, by measuring the effects of the strychnine with electronic apparatus, the method allowed for more precise measurements than mere gross sensory or motor responses to stimuli.

A second significant research project for Dusser de Barenne and McCulloch concerned the ways in which an electrical stimulus applied to one point of the cortex facilitates or extinguishes motor responses to a second test stimulus. For many decades, it had been known that a given stimulus could facilitate the motor effects of a second stimulus. Dusser de Barenne and McCulloch, however, discovered that there are conditions under which a given stimulus could extinguish the motor effects of a second test stimulus. What made this interesting to them was that the extinguishing effect could not be explained in terms of fatigue of the stimulated neurons. Large initial stimuli did not induce the extinction as well as did just supraliminal stimuli. Were the effect a matter of fatigue, they reasoned, the effect would be proportional to the

intensity of the initial stimulus. In other words, the extinguishing effect seemed to reveal something about the mechanisms of normal brain function. Five years’ worth of investigation on these facilitating and extinguishing phenomena led them to the conclusion that multiple factors were involved, one of them being reverberating electrical activity in closed loops of neurons. This was especially rewarding to them, since just a few years before this Rafael Lorente de Nó had used anatomical evidence to argue for the existence of closed loops of neurons.

In June 1940, Dusser de Barenne died suddenly of heart disease. Rather than continue its support of neuro-physiology, the faculty of Yale’s Medical School decided to reallocate Dusser de Barenne’s endowed chair to another discipline, leading to the incorporation of the Laboratory of Neurophysiology into John Fulton’s Laboratory of Physiology. With considerable disappointment and frustration, McCulloch felt compelled to search for another academic appointment. With the help of friends, he was appointed director of research at the newly created Illinois Neuropsychiatric Institute starting 1 September 1941.

At the University of Illinois . While McCulloch’s stature was growing during his time at Yale, the move to Chicago enabled him to flourish as his own intellectual, working on research projects of his own conception. It also dramatically expanded McCulloch’s range of scientific connections. At the Neuropsychiatric Institute, he continued to work on purely physiological and neurophysiological projects, such as tests of glucose tolerance, carbohydrate metabolism, and the anatomical mapping of neuronal pathways in the brain. His best-known contributions from his time in Chicago, however, were the product of interdisciplinary collaborative projects beyond the confines of the institute. He also indulged his literary side through his involvement with the Chicago Literary Club and in later years was proud of the fact that a collection of his poetry, The Natural Fit (1959), was the first poetry ever published by the society.

Among his earliest new friends in Chicago was a medical school student, Jerome Lettvin, who introduced McCulloch in turn to his younger friend, Walter Pitts. For many years, McCulloch had thought that it should be possible to describe the activity of networks of neurons using some form of Boolean logic of sentences. This was to be a formalization of some of the early-twentieth-century qualitative theories of the role of networks of neurons underlying psychological processes. From his work at Yale, McCulloch had also come to believe that closed loops of neurons played an important role in brain activity. Pitts was just the partner McCulloch needed in order to formalize his logical conception of nervous nets. Though still a very young man of eighteen, Pitts had for some years been studying logic and mathematics on his own. Moreover, he had allied himself with Nicolas Rashevsky’s group in mathematical biophysics at the University of Chicago and for about a year had been working on a project to use differential equations to describe the electrical activity in closed loops of neurons. Realizing their confluence of interests, McCulloch and Pitts set about forging what came to be their most significant paper, “A Logical Calculus of the Ideas Immanent in Nervous Activity.” This work, published in 1943 in Rashevsky’s journal, Bulletin of Mathematical Biophysics, gave logico-mathematical form to the idea that networks of neurons compute logical functions.

“A Logical Calculus”: Two Ideas . There are two principal ideas in “A Logical Calculus.” The first, and most influential, is that it is possible to establish formal equivalences between formulas in an extension of Boolean logic and networks of on-off neurons that contain no closed loops. As a simple example, consider the network of three neurons shown in Figure 1. The formula McCulloch and Pitts would use to describe this network would be N₃(t) ≡ N₁(t –1) v N₂(t –1), which asserts that the neuron N₃ will fire at time t if, and only if, either neuron N₁ fires at time t –1 (an earlier instant of time) or neuron N₂ fires at time t –1. McCulloch and Pitts assumed that it required two excitatory inputs to cause a neuron to fire; they inserted two such inputs from both neuron N₁ and neuron N₂. As a second example, there is the network shown in Figure 2. McCulloch and Pitts would describe this network using

the formula N₃(t) ≡ N₁(t –1) . N₂(t –1), which asserts that N₃ will fire at time t if, and only if, both N₁ fires at time t –1 and N₂ fires at time t –1. A third example is the network shown in Figure 3, with positive excitatory synapses shown as black dots and negative inhibitory synapses show as open circles. McCulloch and Pitts would describe this network using the formula N₃(t) ≡ N₁(t –1) . ~N₂(t –1), which asserts that N₃ will fire at time t if, and only if, both N₁ fires at t –1 and N₂ does not fire at t –1. The technical challenge of this idea lay in showing how any network not containing closed loops could be described using a formula in the system McCulloch and Pitts had developed and in showing how any formula in the system could be used to generate an equivalent network.

The second major idea in “A Logical Calculus” was the suggestion that closed loops of neurons could be used to explain memory, purposive behavior, homeostasis, and various psychiatric disorders. So, McCulloch and Pitts’s work provided formal descriptions of what was then the intuitively appealing hypothesis that short-term memory consists of electrical activity in closed loops of excitatory neurons. This kind of short-term memory could complement a kind of long-term memory constituted by modifications to the connections among neurons. The links to purposive behavior and homeostasis came about through McCulloch’s extramural interactions. While working on “A Logical Calculus,” McCulloch learned of Arturo Rosenblueth's, Norbert Wiener's, and Julian Bigelow’s ideas of positive and negative feedback loops and their relation to purposive behavior. According to their analysis, purposive behavior is simply behavior guided by negative feedback. Aiming for a goal is a matter of having some mechanism that detects deviation from that goal and uses negative feedback to correct for the deviation. Closed loops of neurons might provide such a mechanism.

“A Logical Calculus”: The Upshots . “A Logical Calculus” thus helped establish common ground between McCulloch and Pitts, as well as others who were interested in the emerging theory of computation, computer engineering, information, and communication and their application to a broad spectrum of biological, psychological, and sociological phenomena. The practical upshot of this common interest was the initiation of a series of interdisciplinary conferences on “circular causality” to be sponsored by the Josiah Macy Jr. Foundation, with McCulloch as conference chairman. Each conference, initially held at six-month intervals, then later at one-year intervals, featured some five or perhaps ten papers separated by ample time for extensive discussion. In fact, the discussion was widely heralded as the high point of the conferences. In time, the ten conferences of the series attracted an impressive array of mid-twentieth-century luminaries, including the anthropologists Gregory Bateson and Margaret Mead, the engineer Claude Shannon, the mathematicians John von Neumann and Norbert Wiener, the philosopher F. S. C. Northrop, the physiologists Ralph Gerard and Rafael Lorente de Nó, the psychiatrist Lawrence Kubie, and the psychologists Heinrich Klüver and Donald Marquis.

Aside from its influence on the cybernetics movement, “A Logical Calculus” for many years inspired additional research in Nicolas Rashevsky’s mathematical biophysics. Attempts to clarify, simplify, and extend the results in “A Logical Calculus” also led more mathematically inclined investigators to develop what came to be known as automata theory. In later years, the paper served as an important reference point for other mathematical treatments of neural networks, such as Frank Rosenblatt’s work on perceptrons.

The other noteworthy collaboration between McCulloch and Pitts at about this time was “How We Know Universals: The Perception of Auditory and Visual Forms” (1947). This paper showed how one might account for Gestalt invariants in the visual perception of sizes and shapes and in the auditory perception of musical chords in terms of structures that could be mapped onto specific brain structures.

At MIT . During the course of the Macy conferences, McCulloch finished his eleventh year at the Neuropsychiatric Institute. By then, he had had his fill of the responsibilities of directing a research operation involving some forty faculty, students, and staff. He and his wife, Rook, also wished to return to the East Coast to be closer to their summer home, a farm in Old Lyme, Connecticut. Topping it off, there was the prospect of working in a neuro-physiology group with Lettvin and Pitts, who had already moved to Cambridge, with the further hope of attracting Patrick Wall to their group. Adding a neurophysiology group to MIT’s Research Laboratory for Electronics was intended to more fully implement Wiener’s vision of a center for cybernetic ideas. MIT would, thus, become the institutional home to the fruits of the Macy conferences. All of which made a move to MIT in October 1952 seem eminently attractive.

Unfortunately for all involved, the move did not bear the fruit that was expected. Just as the negotiations for McCulloch’s move were reaching their conclusion, a personal rift arose between Wiener and McCulloch, Pitts, and Lettvin. Much to McCulloch's, Pitts's, and Lettvin’s disappointment, Wiener refused to have anything to do with them. This was the most dramatic development of the MIT years, but diverse interests also led the members of the group in different directions.

The break with Wiener was especially difficult for Pitts. Although always a troubled individual, the following years saw Pitts simply refuse to complete the work he needed to do for his PhD. Despite heroic efforts by Lettvin, McCulloch, and others, Pitts progressively lost interest in research and turned to long periods of self-destructive behavior. Lettvin and Wall perhaps remained most true to the original conception of an experimental group in neurophysiology. They took the lead in experimental work on the spinal cord and the use of strychnine during the early 1950s. In the late 1950s, Lettvin and Humberto Maturana produced some of the classic results in vision science. Using sophisticated microelectrodes, they found that cells in the frog’s retina respond selectively to moving black dots. In other words, the frog’s eye responded to things that looked like the frog’s prey, namely, moving flies. Their subsequent publication, “What the Frog’s Eye Tells the Frog’s Brain” (1959), with McCulloch and Pitts as coauthors, became one of the most frequently cited papers in vision science.

McCulloch, for his part, contributed to the experimental work the group was doing on the spinal cord, the use of strychnine, frog vision, and neuron physiology, but his heart had moved away from experimental work in the direction of more purely logical and mathematical investigations. One of his later projects at MIT concerned how the brain could compute reliably given the unreliable, probabilistic activity of individual neuronal action potentials. He became interested in probabilistic, multivalued logics in contrast to the deterministic, two-valued logic of McCulloch-Pitts networks. Another project, pursued most extensively with William Kilmer, was the development of a theory of the reticular formation of the brain as a command and control center.

While at MIT, McCulloch indulged many more of his personal proclivities, some to good effect and some to not so good. A naturally gregarious extrovert with a talent for public speaking, during the 1950s and 1960s, McCulloch was frequently on the road giving lectures. He spoke on computers, cybernetics, and bionics at universities and conferences in the United States and in Europe. He also served as an advisor to the National Aeronautics and Space Administration and was the first elected president of the American Society for Cybernetics. Over the course of his travels, McCulloch met many bright and aspiring scientists whose careers he did his best to promote. Perhaps in part because of his experiences with the young Pitts and his fondness for young people in general, McCulloch also frequently gave lectures at high schools in Connecticut, Massachusetts, New Jersey, and New York. Less good for McCulloch was his love of scotch and filterless cigarettes. They apparently cut his life short while he was reading a research paper at his farm in Old Lyme.

BIBLIOGRAPHY

McCulloch’s professional papers are available for scholarly use at the American Philosophical Society Library in Philadelphia.

WORKS BY MCCULLOCH

With W. Pitts. “A Logical Calculus of the Ideas Immanent in Nervous Activity.” Bulletin of Mathematical Biophysics 5 (1943): 115–133.

With W. Pitts. “How We Know Universals: The Perception of Auditory and Visual Forms.” Bulletin of Mathematical Biophysics 9 (1947): 127–147.

The Natural Fit. Chicago: Chicago Literary Club, 1959. A pamphlet of poetry, reprinted in The Embodiments of Mind.

With J. Y. Lettvin, H. R. Maturana, and W. H. Pitts. “What the Frog’s Eye Tells the Frog’s Brain.” Proceedings of the IRE 47 (1959): 1940–1951.

The Collected Works of Warren S. McCulloch. Edited by Rook McCulloch. Salinas, CA: Intersystems Publications, 1989. An extensive, but still incomplete, collection of McCulloch’s publications and lectures. The primary omissions are published abstracts from conference proceedings.

The Embodiments of Mind. Cambridge, MA: MIT Press, 1989. A sampler of McCulloch’s publications and lectures from 1943 to 1964. This collection obviously does not at all represent the work McCulloch carried out at Yale.

OTHER SOURCES

Abraham, T. “(Physio)logical Circuits: The Intellectual Origins of the McCulloch-Pitts Neural Networks.” Journal of the History of the Behavioral Sciences 38 (2002): 3–25.

_____. “Integrating Mind and Brain: Warren S. McCulloch, Cerebral Localization, and Experimental Epistemology.” Endeavour 27 (2003): 32–36.

Arbib, M. “Warren McCulloch’s Search for the Logic of the Nervous System.” Perspectives in Biology and Medicine 43 (2000): 193–216.

Kenneth Aizawa