Sidney Siegel--The Professional Years
|Sid and Alberta married in the summer of 1954, and with Jay moved to Pennsylvania, where Sid had been named to the faculty in the Department of Psychology at Pennsylvania State University. As always, his driving energy and enthusiasm were quickly mobilized in a variety of directions: he taught courses in statistics and social psychology, he directed several students in research on authoritarianism, he participated in research evaluating the effectiveness of certain experimental courses in modifying students' attitudes, and he continued to work on the problem of the measurement of utility. He modified the approach, which had been developed in the collaboration with Davidson and Suppes, based on the earlier work of Ramsey, applying it to discrete objects. With such objects, utility may be measured in a higher-ordered metric scale -- a scale between an ordinal scale and an interval scale in strength. Sid wrote a paper on the method for ordered metric scaling and then worked with a graduate student, Paul Hurst, on the measurement of utility for cigarettes among prisoners at the Lewisberg Federal Prison, where cigarettes constitute an informal medium of exchange. Hurst collected the data, and Sid never let him forget that a prisoner had been killed there not long before in a fight over a carton of cigarettes.
In his second semester at Penn State, Sid taught a graduate course in statistics. He presented a comprehensive introduction to the theory of measurement, suggested the implications of this theory for statistical practice, and then taught a variety of nonparametric tests as being the techniques most suitable to the analysis of data of the form typically employed in the behavioral sciences. In this course, he was frustrated by the difficulty of finding suitable source materials on nonparametric statistics and spent a good deal of time collecting and reproducing for student use the various tables associated with the tests. There was so much interest in these that he considered publishing a collection of tables for nonparametric tests, thinking that a student or research worker could use these in connection with several summary sources then available. He had found the existing sources to be incomplete and not always clear, however, so instead he decided to write a handbook about nonparametric statistics, addressing it to behavioral scientists whose skills were in research and not in mathematical statistics. He devoted the summer of 1955 to this project.in addition to teaching a full schedule of courses.
In writing Nonparametric Statistics for the Behavioral Sciences, Sid had several goals. One was to bring useful material together in one book. Another was to indicate the relation between the level of measurement achieved in research and the statistical tests suitable for analyzing the data. His position on measurement was a strong one, and it was boldly stated in the book and in a paper he wrote shortly afterwards. Later on, he came to modify this position somewhat, and the stance he eventually took on the measurement-and-statistics issue was presented in two later papers. A third goal of the book was to present statistics as an adjunct to research, capitalizing on the reader's interest in research to help him learn something about statistics. This purpose underlay the use of examples from real research in the text in place of the usual discussion of coin tossing. Still another goal was to present the material in a form convenient for later reference, even though this might require repetition of the same point at several places in the book. Finally, his aim was to write at a level comprehensible to the mathematically naive graduate student or research worker. Alberta argued very strongly for the last mentioned goals, and she is partly responsible for some of the "taking the reader by the hand" which some of Sid's more sophisticated colleagues find tiresome.
Although the publishers regarded the book as a probable money loser when they issued it in 1956, Nonparametric Statistics has been among the most widely adopted statistics books in the United States every year since its publication, and it has appeared in several foreign languages. Throughout the remainder of his life, Sid maintained a large correspondence with readers of the book, advising them on statistical problems and replying to questions about contents. As his research interests developed and came to focus on decision-making and choice behavior, he became rueful at being identified as "Nonparametric Siegel," as he often was when introduced to colleagues at meetings, but of course his dominant reaction was pride in the wide use his book enjoyed.
In the two years after the statistics book went to press, from 1955 through 1957, Sid directed his attention to several research problems, in addition to continuing to carry a heavy load of classroom teaching and supervision of graduate student research. First, struck by the meaningfulness of the assertion that "science is measurement," he worked on demonstrating the centrality of careful and ingenious measurement to scientific work. In addition to the study with Hurst on an application of ordered metric scaling, he participated in two other small studies to demonstrate that a fresh measurement technique could enable progress in some stale and overworked areas of social psychology. For a study of social distance as measured by his technique; he collaborated with a graduate student from the South, Irma Shepherd, who had access to Southern women whose attitudes could be assessed. In a study of political preference as revealed by the technique of ordered metric measurement, the collaboration was with a colleague, George M. Guthrie, and a graduate student, Selwyn W. Becker.
The second line of work was related to this. The notion of an ordered metric scale of utility, in which objects are ordered according to the subject's preference for them and, in addition, the distances between the objects are ordered to reflect his preference, suggested a fresh way of looking at level of aspiration. Sid suggested that a person's level of aspiration is a point on his utility scale for a set of achievement states and that ordered metric measurement of the utility of those states is sufficient for identifying the level of aspiration, since it is that point which has the largest distance between it and the next adjacent state below. Having formulated this idea, he enlisted the collaboration of Selwyn W. Becker in testing it. In both of the tests they conducted, they studied students' aspirations for grades on an examination in a college course. The fact that the level of aspiration could be determined behaviorally rather than verbally in a realistic situation having a significant outcome for the subject seemed especially important to Sid in evaluating the validity of the approach. The data from the studies, of course, were themselves validating data for the concept, as were the latter data from the bargaining studies in which the notion of level of aspiration was invoked.
The third focus of research during these years proved to be the one with the greatest long-term fruitfulness. In his search for an event having subjective probability .50, Sid had experimented briefly in 1953 with the two-light situation which attracted so much interest from learning theorists in the 19508. He was intrigued by some of the phenomena he observed with pilot study subjects. More important, he became interested in following the emerging research literature on the situation, and he was struck by the apparent incongruity between the statistical association theories which were dominating work with this situation and the decision-making hypothesis of maximization of expected utility. Working with a graduate student, Donald A. Goldstein, he devised an experiment to test the notion that in fact subjects behave as if they were maximizing expected utility in the situation. This was demonstrated by the fact that altering the payoffs built into the situation alters the stable-state behaviors of subjects.
Early in his third year at Penn State, the first major professional recognition for his research came to Sid: he was invited to be a Fellow at the Center for Advanced Study in the Behavioral Sciences. He accepted for the following year, 1957-1958. The Center had opened on the Stanford campus just as we had left Stanford, and of course the unique opportunities it offers for independent work were well known to us. We moved back to California in June of 1957, embarking on a fifteen-month period which was to be significant in Sid's work and which was also to be among the happiest times of his life. Sid spent the summer as a visiting member of the Stanford faculty, then moved up the hill to the Center .
The Center provides its Fellows with an opportunity for uninterrupted work on tasks of their own choosing. Its administrative staff, Ralph W. Tyler, Preston Cutler, and Jane Kielsmeier, have succeeded in creating a context in which noncompetitive and mutually helpful relations develop among the Fellows. Its isolation, beautiful setting, and informal atmosphere make the Center a place where lasting friendships are quickly formed. Sid's tenure there was enriched by companionship with a group of scholars whose work he admired and whose friendship he esteemed. Although he was interested by the work of Fellows in several disciplines, he was especially attracted by the work of the economists at the Center that year -- Kenneth J. Arrow, Milton Friedman, Melvin W. Reder, Robert M. Solow, George J. Stigler -- and by the mathematically advanced status of their field in contrast to the status of other behavioral sciences. He became more convinced that the notion of maximization of expected utility has central explanatory value in the analysis of human decision-making and that experimental work with utility is important because that concept had been notably resistant to non-tautological empirical rendering.
At the Center, Sid formalized the notions which had underlain the initial experimentation with the two-light situation, stating a theory which he was to test, revise, and test again in his laboratory during the ensuing years. He also arrived at a more formal statement of the model for ordered metric scaling. Finally, the work of which he was proudest was his collaboration with John W. Tukey in developing a nonparametric statistical test for differences in spread, thereby filling a gap in the battery of available nonparametric tests by providing a nonparametric analogue to the F-test for differences in variability. The tables they developed for use with the test are also useful in other ranking tests, serving as more convenient and more comprehensive alternatives to other tables available for such tests.
Freed from the usual pressures, at the Center Sid was able to think in terms of long-term projects and commitments. In retrospect, he felt that his pattern had been to "hatch a bright idea," write a theoretical paper about it, and then collaborate with students in two or three quick studies providing tests of the idea. He decided that it would now be more appropriate for him to settle on a program of research, developing in depth his interest in a single area. He determined to embark on a series of studies of choice behavior, following the lines sketched in his theoretical paper, and he received generous support from the National Science Foundation for that plan.
When Sid returned to the Penn State campus in the fall of 1958, the Department of Psychology had acquired a new chairman, Arthur H. Brayfield. Sid and Art quickly discovered that they shared many goals for psychology and had many of the same values--and the same peeves--concerning academic life. With the university provost, L. E. Dennis, Art gave strong administrative support to Sid's program of research and teaching. Sid was named Professor of Psychology at Penn State in 1959, five years after he had joined the faculty, and at the beginning of 1961 he was named Research Professor of Psychology, an appointment which gave him autonomy in determining the course of his work. With this administrative support, Sid organized a laboratory for the study of choice behavior.
The laboratory was important to Sid because of his beliefs concerning graduate study. He saw graduate work as an apprenticeship in which the student's most valuable learning occurs through working at the side of an experienced investigator in his laboratory. Further, he believed that the contents and methods of a field are not the only topics to be mastered by graduate students; equally important are the attitudes, enthusiasms, and loyalties they develop through graduate study. He felt that a student's commitment to psychology and to academic values would come through a process of personal identification with his professor. The laboratory provided a physical setting in which a student could become identified with his professor and his work.
Sid did not encourage distinctions between "teaching assistants,” “research assistants," and "graduate fellows." Every graduate student who worked with him, whatever the title of his appointment, was drawn into the total academic enterprise. Sid tried to arrange for every student who worked with him to have a desk in the laboratory, to teach a weekly section of his elementary statistics course, to participate in all the discussions of on-going research, to observe data collection in every study under way, to attend one or two national meetings of professional associations each year, and in other ways to serve as an apprentice participating in all facets of academic life. Occasionally he would ask a student to help him in replying to a letter seeking advice on research, to assist him in proofreading a paper for publication, or to offer comments on a paper he was refereeing for a journal. As always, the purpose was to involve the student in the total enterprise. In addition, of course, Sid developed personal friendships with his students, who were among the most frequent and welcome guests in our home. The students who worked most closely with Sid at Penn State from 1958 through 1961 included Julia McMichael Andrews, Robert Evanson, Joseph Gaberman, Donald Harnett, James McKendry, Joseph Robertson, Conrad Weiser, and Patricia Yost. In addition, Robert Radlow was a postdoctoral fellow in the laboratory, pursuing studies of decision making and learning.
In common with other men as they become established in their fields, Sid found that an increasing amount of his time was given to service to his profession. He served frequently as a referee for research-grant proposals to the National Science Foundation or the U.S. Public Health Service. He was successively a Consulting Editor and then an Associate Editor for Sociometry during the editorship of his friend John A. Clausen. He gave editorial consultation to the Journal of the American Statistical Association, the Psychological Review, occasionally to other journals in his field, and to commercial publishers. At Penn State, he served as President of the campus-wide Social Science Research Center.
Shortly after his return to Penn State from the Center, alight with enthusiasm over the mathematical sophistication of economics, Sid became acquainted with Lawrence E. Fouraker, one of the ablest economists at the university, and they began to discuss the possibility of devising experimental tests of economic hypotheses. They chose some classical propositions with a long and honorable history in economic theory, propositions which had received no precise test because the only available data had been naturalistic observations of the behavior of businessmen. Economists, like psychologists, had become suspicious of naturalistic observations not only because economic data are fragmentary and incomplete, but also because they are confounded in ways only too familiar to any social scientist who has relied on Nature as an experimenter. Fouraker was sympathetic to the notion of collecting data in a social psychological laboratory under controlled conditions, an alternative to naturalistic observation which had not yet gained vogue among economists. Sid suggested some approaches that might be employed, and soon he was swept up in a series of experiments in economic bargaining. They centered on bargaining between bilateral monopolists -- between the single seller of the commodity and the single buyer.
Although the bargaining studies diverted some of his attention from the choice-behavior experiments to which he had intended to devote his undivided concern, Sid found them too challenging to forego. They represented a form of interdisciplinary collaboration for which there was little precedent -- -cooperation between an economist and a laboratory psychologist. Previous incursions of psychology into economics had represented the most applied branches of psychology, and the bridge had been to some of the least theoretical fields of economics. The bargaining studies, on the other hand, represented a natural alliance between experimental social psychology and one of the core fields of economic theory. Another attraction of the bargaining studies was the fact that the dependent variable was expressed in terms of money; this meant that the measurement problems in the research were minimal. Further, it was natural to employ monetary payoffs to the bargainers and by this time Sid had an ingrained suspicion of any study whose validity relied on the subject's motivation to cooperate, with no attention to the rewards to him contingent on his performance relevant to the variables under study. Finally, bilateral monopoly bargaining is a prototypical form of social conflict, analogous in structure to the conflict between the United States and the U .S.S.R. in, say, disarmament negotiations, so the social importance of an understanding of bargaining behavior was clear.
During the winter of 1958-1959, Sid and Larry conducted a series of five experiments in bilateral monopoly. They reported them in a monograph they wrote in the summer of 1959. This short book, Bargaining and Group Decision Making: Experiments in Bilateral Monopoly, was awarded the $1,000 monograph prize in the social sciences for 1959 by the American Academy of Arts and Sciences. [Included in the series of studies was one in which the concept of level of aspiration was invoked, following the formulation Sid had presented some years earlier, and aspiration levels were induced experimentally. The report of that study has been adapted from the book for inclusion in the present collection.]
Larry Fouraker was on leave at Harvard during the 1959-1960 year. In his absence, Sid conducted some pilot studies in bargaining behavior in collaboration with Martin Shubik, but devoted his principal attention to the choice-behavior studies. Upon Fouraker's return for 1960-1961, a second and more extensive series of bargaining studies was launched, Attention was given to bargaining in economic situations which are more complex and thus perhaps more interesting than the situation of equal strength bilateral monopoly. First, negotiations between pairs of bargainers in a situation of bilateral monopoly under price leadership were studied. Then, in a larger group of studies, negotiations in oligopoly were studied. In all, about 400 different subjects were observed in bargaining under 15 different experimental conditions. In addition, to test the generality of their findings with student subjects, Sid and Donald Harnett replicated two of the bargaining conditions with experienced businessmen in a study conducted in New York City in June, 1961, under the auspices of the Behavioral Research Service of the General Electric Company. Sid and Larry devoted the spring and summer of 1961 to writing prepublication technical reports of their bargaining studies, intending to prepare the report for publication during the 1961-1962 year.
The program of research in choice behavior spanned the same period of time as the bargaining studies. A succession of experiments was performed to test the theoretical model, which underwent repeated modifications and extensions as the research progressed. Fundamental to the model, which concerns choice behavior in a repeated-choice situation, is the notion that a person's strategy in making choices -- the proportionate distribution of his choices over the alternatives open to him -- is one which maximizes expected utility. That is, with experience in a choice situation, the person comes to distribute his choices in a way which maximizes his expectation of satisfaction from the situation.
After the first two years of work, Sid presented a progress report on the research program in a paper for the New York Academy of Sciences, in May, 1960. At about the same time, the first paper from the work went to press; conducted in collaboration with Julia McMichael Andrews, it concerns the choice behavior of young children in a game which in its essential features is analogous to the decision situation employed with mature subjects. Sid did not take time to write reports on the other studies in the series, for he had been invited to return to the Center for the 1961-1962 year and he intended to devote that year to preparing a comprehensive and integrated report of the total project.
In September of 1961, Sid and Alberta went to the Center for Advanced Study in the Behavioral Sciences, where we were both to spend a year in residence. Julia Andrews joined us there to work with Sid in the preparation of the material on choice behavior; she had been a graduate student in his laboratory during the previous three years and had collaborated in both theoretical and empirical work on the project. Sid readily became involved with the current group of Fellows at the Center, participating in a seminar on measurement, discussing his own research, and exploring new ideas with Albert Madansky, a mathematical statistician with whom he soon identified shared interests.
Sid's death came suddenly on November 29, 1961. Its cause was coronary thrombosis. Although there was a familial history of cardiovascular disorder, Sid's own health had been excellent, and his death was totally unexpected. Perhaps it is not too sentimental to remark that it occurred at a place -- the Center-which he loved and which had played a very special role in his life.
Mrs. Andrews and Alberta decided to attempt to complete the report of the research on choice behavior, dealing with the manuscript materials and technical reports already at hand. The National Science Foundation assisted us materially in this project, and we were fortunate in having the services of Albert Madansky as a technical consultant. Thus Sid's book Choice, Strategy, and Utility will appear posthumously. Larry Fouraker generously assumed the responsibility for completing the jointly authored book, Bargaining Behavior, which also appeared after Sid's death.