In addition to the authors of this paper, Professor Joseph M. Allman, Mr. Fred Terbrusch and Mr. James R. McCoy have participated in the overall project and, consequently, have contributed to the intellectual antecedents and data collection upon which this paper is based. Special thanks must go to the chairmen and graduate student contacts in the departments studied. Their interest and cooperation were essential to the success of the project. Finally, thanks are due to Mrs. Cari Gabiou for her critical reading of the first draft of the paper, and to Mrs. Linda Mushkatel for her assistance in preparing the final manuscript. Any faults are the responsibility of the authors.

¹ Our response rates varied between 37% and 65% for graduate students and between 23% and 87% for faculty. We believe these rates are acceptable since we worked from lists of students in each department that undoubtedly contained much “slack” and were less than perfect as sample frames. Many did not distinguish students who were on leave or whose enrollment in the department was pro forma. There were also, of course, many students Included who were absent from the department working on dissertations and even some who had graduated. Wherever possible our contacts in the various departments helped us identify such cases. We feel that the response figures are a conservative reflection of our actual response from students taking courses full time in the department. As usual in surveys, we have no reliable way of knowing the characteristics of those who did not respond — although a reasonable guess is that their involvement in the department was more marginal than those who did respond.

To make a systematic exploration of departmental structure the study had to encompass a range of departments, preferably representative of the variety in the discipline. Our problem at the outset was that we could only guess at this variety. Accordingly, we used two sources of information which did not identify the dimensions of internal structure we were interested in but which gave us some data on very broad differences. The first source was a brief questionnaire sent to “the departmental secretary” in forty-five departments across the country. The second was the data used by Luttbeg and Kahn in their analysis of departmental differences. (Most importantly, Luttbeg, Norman and Kahn, Melvin, The Making of A Political Scientist: An Empirical Analysis of Ph.D. Programs, Public Affairs Research Bureau, Southern Illinois University, 1969.)Google Scholar These were sufficient to distinguish a “behavioral-traditional” dimension and a “departmental size” dimension. From the information available to us, we crudely grouped departments into four types defined by these dimensions and selected those to be studied from these types. Our financial resources only permitted a sample of fourteen departments, and within these we attempted a complete enumeration of faculty and graduate students.

Departments were selected from the four types according to the availability of faculty contacts there. Such contacts, and the student contacts who helped us with the administration of the questionnaire, were important if we were to have the active cooperation of people in the department. Although we will not report the findings by department name, we feel that the face validity of our final list as a representative sample of graduate departments is persuasive. Our sample of students, therefore, may be regarded as a crudely stratified two-stage sample in which the first stage was of departments and the second was an attempt at a complete enumeration of the relevant population (students and faculty). We argue that our sample is likely to be broadly representative of both departments in the profession and individual members, both students and faculty. We believe it is the best we could do with our limited resources.

² We used the “elementary linkage” technique designed by McQuitty, Louis, “Elementary Linkage Analysis for Isolating Orthogonal and Oblique Types and Typal Relevancies,” Educational and Psychological Measurement, 17 (1957), 207–29.CrossRefGoogle Scholar This technique defines a “type” as a subcategory of n variables (or, in the case of Q analysis, n cases or individuals) of the nature that all variables in the subcategory are more like each other than they are like any other variable in any other subcategory. In this preliminary analysis we clustered variables not departments; thus, our clusters indicate how certain departmental characteristics go together, not how similar various departments are.

³ One department was excluded because its faculty decided that research conducted on political scientists — rather than by them on other people — violated personal privacy. Another was excluded because it had such a small student body and such a small number of faculty that, after the response rate was taken into account, there were too few cases to deal with. These department-level variables based on mean scores can be classified “analytical” variables in the scheme developed by Paul F. Lazarsfeld and Herbert Menzel. They are distinguished from “global” variables which are based on properties of the entire organization — for example, its decision-making structure. Analytical variables are based on properties of the individual members of the organization, and these values are then aggregated to form a summary measure for the whole unit. Analytical variables (as in the present case) characteristically measure the distribution of some property among the members of the organization. See: Lazarsfeld, Paul F. and Menzel, Herbert, “On the Relation Between Individual and Collective Properties,” in Etzioni, Amitai, Complex Organizations: A Sociological Reader (New York: Holt, Rinehart & Winston, 1961), pp. 422–40.Google Scholar Other writers have discussed the problem of measuring such properties of educational units, most relevantly: Barton, Alan, Organizational Measurement and its Bearing on the Study of College Environments (New York: College Entrance Examination Board, 1961)Google Scholar; Lazarsfeld, Paul F. and Thielens, Wagner, The Academic Mind (Glencoe: The Free Press, 1958).Google Scholar

⁴ The means and standard deviations of these questions, all of which were measured on a five-point scale, were:

These figures are, of course, the means and standard deviations for mean values on the questions in the twelve departments.

Using the method outlined in Benjamin Fruchter, Introduction to Factor Analysis (Princeton, N.J.: D. Van Nostrana Company, Inc., 1954), pp. 61–73 Google Scholar, we arrived at loadings for each of our departments on this dimension and the subsequent ones. This allowed us to introduce the variable into the analysis along with other departmental variables, and also gave us some idea about the distribution of departments in these terms. Within our sample of departments — and we have reasonable grounds for thinking this sample is representative — the range of loadings was from 9.2156 to 13.1741 and there was a relatively even distribution of departments along that continuum. The mean score was 11.0933 and the standard deviation was 1.1354. In short, our data suggest that, in terms of this “organizational climate” variable, a wide range of departments exist.

⁵ Wheeler, Stanton, “The Structure of Formally Organized Socialization Settings,” in Brim, Orville G. Jr., and Wheeler, Stanton, Socialization After Childhood (New York: John Wiley & Sons, Inc., 1966), p. 82.Google Scholar

⁶ Tne means and standard deviations of the variables loaded on this cluster are:

Within the sample of departments the range of loadings was from 6.8974 to 9,4571 — somewhat less than for the organizational climate cluster — and the distribution was evenly spread along the continuum. The mean score was 8.0575 and the standard deviation was .7051.

⁷ Roose, Kenneth D. and Anderson, Charles J., A Rating of Graduate Programs, American Council on Education, One Dupont Circle, Washington, D.C., 20036, 1970.Google Scholar Twenty-two schools are ranked by this report, including five of those we sampled. In these cases we assigned an ordinal ranking by their order of appearance. Ten further schools were listed in alphabetical order as having a ranking in a second group, and three of ours were included. These three were assigned the same ranking (in this case seven). A further list of twelve beyond that contained one of our schools and it was given a rank of eight. The remaining schools we sampled were not given a ranking from this publication, and we assigned them the same score at the bottom of our list.

⁸ We should point out that not all of the student respondents were anticipating careers as professors of political science. In fact, 12.9% were planning careers in some other capacity. As might be expected, a higher proportion of Ph.D. students were planning professorial careers than M.A. students. Our data were collected in the fall of 1969 and the winter of 1970; we suspect that the proportion of students anticipating a career in a university post has declined since then as a realistic reflection of the job market — although this is based on limited impressions from a small number of departments.

⁹ See, for example: Haire, Mason, “Biological Models and Empirical Histories of the Growth of Organizations,” in Haire, Mason (ed.), Modern Organization Theory (New York: John Wiley & Sons, 1959), chapter 10.Google Scholar According to Alan Barton: “Size is a major but ambiguous attribute of the social structure of organizations. Size itself has certain necessary, formal consequences for the possible range of interpersonal relations, of communication links, and of levels of authority as conditioned by spans of control. In any given study, classifying organizations by size also classifies them by certain kinds of communications, authority, and social relations patterns which are its consequences and which in turn have other effects; it is by no means easy to say what intervening variables or incidental correlates size indicates.” Organizational Measurement, p. 39.

¹⁰ Argyris, Chris, Personality and Organization: The Conflict between System and the Individual (New York: Harper & Row, 1957).Google Scholar

¹¹ This is particularly true of communications between individuals at different authority levels in organizations. Harold Guetzkow has written that “A dominant feature of such nets (communications) is its directionality, in that orders usually flow vertically within the organization, from a few individuals at the top of the authority structure to the many individuals in its lower regions.” Guetzkow, Harold, “Communications in Organizations,” in March, James G. (ed.), Handbook of Organizations (Chicago: Rand McNally & Company, 1965), p. 543.Google Scholar See also, Kahn, Robert L. et al. , Organizational Stress: Studies in Role Conflict and Ambiguity (New York: John Wiley & Sons, Inc., 1964), pp. 190–92.Google Scholar

¹² Tannenbaum, Percy H. and McLeod, Jack M., “On the Measurement of Socialization,” Public Opinion Quarterly, 31 (1967), pp. 27–37.CrossRefGoogle Scholar

¹³ Maruyama, Magoroh, “The Second Cybernetics: Deviation-Amplifying Mutual Causal Processes,” The American Scientist, 51 (1963), pp. 164–79.Google Scholar

¹⁴ Maruyama, “The Second Cybernetics,” p. 177.