The Impact of These Ideas on Me

Because I came to Columbia with almost no background in any social science, the frame of reference imposed by sociology at Columbia was the only frame I had. I reflected, both in the methodological skills I developed and in the substantive problems I saw as important, the then current Columbia definition of sociology. But my development as a sociologist reflected more specifically the confluence of three streams of activity: the activities of Lazarsfeld, the activities of Merton, and the activities of Lipset. From Merton came not only the vision of sociology as a calling but also a focus on sociological determinants of individual behavior, following in the pattern of Durkheim, on whose analysis of

social determinants of suicide (not The Division of Labor or The Elementary Forms of Religious Life, both of which had more a systemic, less individual, focus) Merton lavished attention in his lectures on theory.

Whether it was Merton's own orientation, the individualist direction of sociology at Columbia, or a positivist orientation I carried over from the physical sciences, midway through graduate school I concluded that Durkheim was engaged in one kind of endeavor and Weber in an altogether different one, and that I would choose Durkheim's path. I saw that path as the study of the force of social structure and social organization on the individuals embedded in that structure. Weber I thought of as engaged in the other side: the study of consequences of individuals' values, and the actions following from them, for social organization. Parsons I saw as trying to realize a Weberian program of theory construction and research, and although I felt it might have been done better by another, part of Parsons's problem seemed to lie in the Weberian starting point.

I have subsequently changed my orientation toward social theory to one more consistent with the Weberian program by taking purposive action as a starting point for social organization rather than social structure as a starting point for individual action. This problem, which can be described as the micro-to-macro problem, was not what I took away with me from Columbia. The Durkheimian orientation was far more compatible with the quantitative methods and the mathematical sociology I was learning there.

My orientation toward mathematical sociology came from Lazarsfeld via three channels, in addition to the Lazarsfeld-Nagel seminar and the evening university seminar on mathematics in the social sciences. First was the project he gave me of translating Rashevsky's mathematical biology for a sociological audience. Although this work (published in Mathematical Thinking in the Social Sciences ) had little, if any, significance for the discipline, it was important for my development. Rashevsky was a mathematical biophysicist who had turned to sociological problems and had developed models of various social processes. What was important for me about Rashevsky's work was what was important later in Nagel's orientation and in some of the mathematical models we studied in the Lazarsfeld-Nagel seminar: it was mathematics dedicated to mirroring social processes . It was not just a compilation of statistical indices nor a static representation of structure (as was Lazarsfeld's latent structure analysis, a development that excited La-

zarsfeld but left me cold). In this it was compatible with the applications of mathematics I knew from physics and physical chemistry and reinforced my belief that this mirroring of social processes was the most profitable application of mathematics in sociology. The orientation I found in Rashevsky, Nagel, and much of Herbert Simon's work (which influenced me while I was still in graduate school) is not one I have seen among users of mathematics in sociology whose background is in mathematics or statistics rather than the physical sciences or engineering. Nor did I find it in Lazarsfeld, whose preoccupation was with latent structure analysis and index construction.

The second channel was the Behavioral Models Project, for which I contracted to do a review and exposition of mathematics as applied to the study of small groups. In that work I learned about systems of nonlinear differential equations, beginning with Simon's modeling of Homans's propositions about small-group processes, and then discovering A. J. Lotka's work and Volterra's equations for predator-prey models—again a reinforcement of the process orientation I had found in Rashevsky. I also stuffed my head full of the work on random and nonrandom nets by Rashevsky's colleagues (Anatol Rapoport and others) at Chicago, which I applied to, and modified for, social networks. This work, however (which has recently been rediscovered by sociologists engaged in the study of social networks), attracted me far less than process modeling.

The third channel was Lazarsfeld's Panel Project, designed to develop methods for the analysis of panel data. The problem of panel analysis that Lazarsfeld posed was one I leaped on because it was hospitable to a model of a process; but his own solution to that problem involved index construction rather than construction of such a process model. It was the struggle to find a solution to this problem that led (with the aid of a suggestion from Richard Savage the year after I left Columbia) to the continuous-time stochastic process models that form the core (chapters 4–13) of my Introduction to Mathematical Sociology (1964) and to subsequent work in the same direction, as represented most recently in Longitudinal Data Analysis (1981). The compatibility of this direction of modeling with the substantive problems surrounding mass communications and mass behavior is not accidental, for those were the problems occupying Lazarsfeld as he formulated the problems of panel analysis.

Though Lipset was less important to the social system of Columbia sociology than Lazarsfeld or Merton, he was no less important than they to my development as a sociologist. Through a series of fortunate

accidents I came in 1952 to be the third member (with Lipset and Trow) of the ITU research team. Lipset had started to plan the research as early as 1949, and it began as an organized research project in 1951. What was of central importance about the ITU study was the fact that it focused on a substantive macrosocial problem in political sociology (the social bases of political democracy) that had a rich history in sociological theory. Yet the research used quantitative data based on the new sample survey techniques to study the problem.

Macrosocial problems and sample survey techniques usually do not mix well in social research. How was Lipset able to bring about a successful mix? The answer, I believe, lies in the dominance of the problem over the data, a dominance buttressed both by the rich store of knowledge that Lipset had about the printers' union and by the fund of social and political theory bearing on the problem that he had at his fingertips. And as Union Democracy documents, the survey of New York printers was only one of a number of data sources, some of which (such as literature about printing unions and informant interviews with printers) Lipset had amassed before the survey project and some of which (such as observation of union meetings and chapel meetings, records of voting by locals in international elections, written material about union political issues gleaned from union publications) were obtained during the project itself. Although quantitative analyses of the survey data can be found throughout the book and indeed are central to the study, it was the framework of ideas from social theory that generated the analyses. Those ideas were, of course, modified by the results of the data analysis. For example, Lipset began with an idea of the importance of the occupational community for political participation in the union; but only in the analysis of the survey data did we discover the importance of certain formal clubs within the union for providing an organized base of opposition to the incumbents. Thus the initial ideas were richly developed, elaborated, and modified by the data. But the main point is that the framework of ideas was set by the macrosocial theory; the data analyses were forced to cope with those ideas, often partially testing a hypothesized macrosocial relation by testing the one or more implied micro-level processes necessary to generate it.

This data analysis, moving between theory such as that of the mass-society theorists and data that only indirectly bore on the theory, was far more important to the development of my data analysis skills than were any courses in research methods. It also led to orientations different from those I would have developed in such courses. One of these was a

low level of interest in statistical inference (an orientation characteristic of Columbia sociology as a whole). The complexity of the linkage between, on the one hand, survey analysis involving relations between attributes of individuals and, on the other, the structure of macro-level relations they were designed to test made a prima facie case for ignoring standard tests and using other criteria to aid in drawing inferences about the social bases of political democracy.

Quite apart from these theoretical and methodological aspects, I found the ITU study appealing because it was consistent with, and reinforced my interest in, political pluralism, the social sources of political diversity, and the structural bases for opposition to an incumbent authority. It was partly these ideas that led to the proposal for a study of high schools that I tried to get funded when I was at the Bureau in early 1955 and finally pursued at the University of Chicago in 1957 (published in 1961 as The Adolescent Society ). Treating the social system of adolescents in a high school as a partially closed social system, I wanted to study the effects of monolithic and pluralistic status systems on the behavior of adolescents, and to study the sources of these variations in status systems. (The proposal is included in an appendix in The Adolescent Society .) The theoretical aims of the study were never realized; The Adolescent Society was written largely for a lay audience and education professionals, with the theoretical work deferred to a later publication that never materialized. The contrast of The Adolescent Society, in which the theoretical aims did not dominate the data analysis, with Union Democracy, in which they did, is instructive, showing the ease with which the constraints of survey analysis can come to distort the original goals of the research.

Besides Merton, Lazarsfeld, and Lipset, others were also important to the development of my ideas. One was Ernest Nagel, both in his courses (logic, and the philosophy of science) that I audited and in the seminar with Lazarsfeld. Herbert Hyman, remarkably ingenious in using survey data to test social psychological ideas, was another. Others included statisticians, particularly Howard Raiffa and Theodore Anderson. Still another was a visitor, Harold Pfautz, who late in my Columbia career introduced me to some of the sociological classics I would otherwise have missed, including Adam Smith's Theory of Moral Sentiments, an extraordinary set of ideas about the self and society. (I had already been entranced by G. H. Mead's ideas, but I found Smith even more compelling.)

Quite apart from faculty, the set of fellow students and research associates at the Bureau was of considerable importance to each of us. It

was, for example, in the traditions group seminars that the ideas about informal social structure and (as McPhee put it) the dynamics of masses began to emerge, ideas that strongly influenced my Community Conflict, the study that Herb Menzel, Elihu Katz, and I did (Medical Innovation ) on diffusion of a new drug through the medical community, a paper on relational analysis that described ways of capturing the effects of informal social structure with survey-type data, and other methodological innovations, such as snowball sampling. These and others were the ideas that emerged four or five years downstream from the watershed that substituted mass communication and social psychology for community studies and social structure at Columbia. This orientation toward the study of behavior in loosely structured social systems is not all of sociology; but it was an important direction, and one that emerged not only from the combination of inputs we had received from the faculty and the research at Columbia but also from the further development of those ideas through our interaction.

This intellectual direction, which I had when I left Columbia, has not, of course, remained unchanged over the years. It has taken at least one abrupt turn, along with other, more minor ones. That, however, is another story, for which there is no space here.