¹ The 26 texts are hereafter identified by their authors; see Table I.

² In this connection, see Tables 2 and 3 below.

³ Cantor, Robert D., Introduction to International Politics (Itasca, Ill.: F. E. Peacock 1976)Google Scholar; Cohen, Ira S., Realpolitik: Theory and Practice (Encino, Cal.: Dickenson Publishing Co. 1975)Google Scholar; Finlay, David J. and Hovet, Thomas Jr., 7304: International Relations on the Planet Earth (New York: Harper & Row 1975)Google Scholar; Isaak, Robert A., Individuals and World Politics (North Scituate, Mass.: Duxbury Press 1975)Google Scholar; Pettman, Ralph, Human Behavior and World Politics: An Introduction to International Relations (New York: St. Martin's Press 1975)CrossRefGoogle Scholar; Rosen, Steven J. and Jones, Walter S., The Logic of International Relations (Cambridge, Mass.: Winthrop Publishers 1974)Google Scholar; Sterling, Richard W., Macropolitics: International Relations in a Global Society (New York: Alfred A. Knopf 1974)Google Scholar; and Sullivan, Michael P., International Relations: Theories and Evidence (Englewood Cliffs, N.J.: Prentice-Hall 1976).Google Scholar

⁴ The estimated figure of 2,000 was given to us by several publishers' representatives.

⁵ The role of publishers in the processes whereby knowledge is passed on to new gensider, for example, the following account, told with considerable enthusiasm and pride to the senior author by the responsible editor, of how one large publishing firm brought an introductory text in anthropology into being. First the firm circulated a questionnaire to those teaching introductory courses in the field, asking them to indicate the main topics they covered in their courses and offering them three books from the publisher's list in exchange for the information. The returned questionnaires were then content-analyzed, computerized, and an average chapter outline, along with the main themes of each chapter, derived therefrom. At this point the firm scoured the field to find an author to write the book that the market research techniques had, in effect, created. As the editor put it, “We needed to find someone to flesh out the chapter outline!”

⁶ Rosenau, James N., “Assessment in International Studies: Ego Trip or Feedback?” International Studies Quarterly, XVIII (September 1974), 339–67.CrossRefGoogle Scholar

⁷ For some stimulating essays in which IR texts are insightfully assessed, see Boulding, Kenneth E., “The Content of International Studies in College: A Review,” journal of Conflict Resolution, VIII (March 1964), 65–71CrossRefGoogle Scholar; Neff, Charles B., “The Study of International Relations: New Approaches to a Familiar Subject,” World Politics, XV (January 1963), 339–53CrossRefGoogle Scholar; Snyder, Richard C., “Toward Greater Order in the Study of International Politics,” World Politics, VII (April 1955), 461–78CrossRefGoogle Scholar; Sondermann, Fred A., “The Study of International Relations: 1956 Version,” World Politics, X (October 1957), 102–11CrossRefGoogle Scholar; and Zinnes, Dina A., “An Introduction to the Behavioral Approach: A Review,” Journal of Conflict Resolution, XII (June 1968), 258–67.CrossRefGoogle Scholar

⁸ In addition to the 178 syllabi that were usable for our purposes, we also received scores of syllabi for more advanced courses, partly because the inclination to cooperate was high and partly because some institutions do not differentiate between introductory and advanced courses. We are grateful to colleagues in the following institutions for their help in creating the sample of 178 syllabi: University of Alabama; Amherst College; Appalachian State University; University of Arizona; Barnard College; Boston University; Brandeis University; Brown University; University of California (Berkeley); University of California (Los Angeles); University of California (Riverside); University of California (Santa Barbara); California State University (Hayward); Carleton College; Case Western Reserve University; Catholic University of America; University of Cincinnati; Claremont Men's College; Colorado College; Columbia University; Cornell University; Dartmouth College; Denison University; DePauw University; Douglass College; Drew University; Eastern Kentucky University; Emory University; Franklin and Marshall College; George Washington University; University of Georgia; Hamilton College; Harvard University; Hobart & William Smith Colleges; University of Illinois; University of Illinois at Chicago Circle; Indiana State University; Indiana University; University of Iowa; Johns Hopkins University; Kansas State University; Kent State University; University of Kentucky; Lake Forrest College; Lehigh University; Livingston College; Louisiana State University; Marshall University; University of Maryland; Massachusetts Institute of Technology; McGill University; Miami University (Ohio); University of Michigan; Michigan State University; Middlebury College; University of Missouri (St. Louis); New Mexico State University; State University of New York (Albany); State University of New York (Binghamton); State University of New York (Buffalo); State University of New York (College of Arts and Letters at Geneseo); State University of New York (Stony Brook); New York University; University of North Carolina (Chapel Hill); Northwestern University; Ohio University; University of Oklahoma; Oklahoma State University; University of Oregon; University of Penn sylvania; Pennsylvania State University (Capital Campus); University of Pittsburgh; Pitzer College; Pomona College; Purdue University; Queens College (New York); University of Richmond; University of Rochester; Roosevelt University; Rutgers University (Camden); Rutgers University (Newark); Rutgers University (New Brunswick); Scripps College; University of South Carolina; Stevens Institute of Technology; Syracuse University; University of Texas (Austin); University of Texas (El Paso); Texas Tech University; Union College; Vanderbilt University; University of Virginia; Virginia Polytechnic Institute & State University; Wayne State University; West Virginia University; Western Kentucky University; College of William & Mary in Virginia; University of Windsor; University of Wisconsin (Madison); University of Wisconsin (Milwaukee); Wittenberg University; College of Wooster; Xavier University; Yale University; Youngstown State University.

⁹ Of the 178 syllabi, 55 percent required reading in only one of the 26 texts, 14 percent required chapters in two of them, and 7 percent assigned three or more.

¹⁰ Of the 42 syllabi that did not assign a text in our sample, 19 were among the 67 requiring reading in other texts, most of them in the fields of American foreign policy, international organization, or international law. All in all, 155, or 87 percent, of the 178 syllabi contained assignments in one or another text—a proportion that is quite consistent with the results of another survey of 553 IR professors, also conducted in 1972–1973, in which some 94 percent were found to rely on texts. Cf. Grundy, Jerilee, “LEP: International Relations Survey,” International Studies Notes, 1 (Winter 1974), 19.Google Scholar

¹¹ Comparisons of the figures in Table r must be made with caution, since there was substantial variation in the number of months and years for which the texts had been available for adoption. Indeed, several that appeared in none or in only one of the syllabi had been published in the year in which we conducted our survey, and thus could not possibly have been used in a number of courses.

¹² In addition, 50 syllabi listed a mean of 26 nonfiction books as suggested readings.

¹³ That this trend toward multiple-book courses is not confined to IR and has become sufficiently widespread and pronounced to trouble textbook publishers can be readily seen in Vanderhoef, Ray W., “College Textbook Crisis of the 1970's,” Publisherk Weekly, Vol. 200 (October 4, 1971), 24–25.Google Scholar

¹⁴ To some extent, of course, this gap may be narrowed through the verbal presentations instructors make in class. Yet it seems doubtful to us whether many teachers report extensively to their students on what they read in preprints and journals or hear at professional meetings; else they would be inclined to make more assignments from these sources.

¹⁵ These comparisons were based on a chi square test using a .05 level of significance, with the 32 syllabi for which class size was not known being omitted from the analysis. Of the remaining 146 courses, 19 had 25 or fewer students, 41 had between 26 and 50, 43 had between 51 and 75, 17 had between 76 and 100, and the remaining 26 had over 100, with 10 of these exceeding 200 students.

¹⁶ At least one page of 38 percent of the syllabi was devoted to specifying the purposes and/or procedures of the course; most of the rest devoted only a paragraph or two to such matters.

¹⁷ Kirk, Grayson, The Study of International Relations in American Colleges and Universities (New York: Council on Foreign Relations 1947).Google Scholar

¹⁸ Interestingly, all but one of 23 of the syllabi listing this requirement indicated that it should be performed through the printed rather than the electronic media.

¹⁹ Further evidence that IR teachers have some sense of working in a discipline can be found in the proportion of courses in which movies or items of fiction were listed among the assignments: only 3 percent of the syllabi listed movies, and only 4 percent assigned items of fiction. On the other hand, the limited extent to which the behavioral revolution may have had an impact on the teaching of IR is perhaps indicated by the fact that only 3 percent of the syllabi assigned, or allowed for, the writing of a term paper in which the student was required to develop a research design, and only 8 percent called for a paper that involved the processing and/or analysis of quantitative data.

²⁰ To some extent, of course, the simple-count data set is more reliable than the thematic data set—partly because it does not involve sampling procedures, and partly because intercoder reliability is bound to be higher when the mere appearance of a word or format entity is counted than when complex themes have to be recognized and aggregated. The simple-count data set is free of sampling problems because it consists of the entire population; i.e., the analytic units were tallied each time they appeared on every page of every text. High intercoder reliability was sought with respect to this set by working closely with the coders to insure common use of the few categories that were subject to multiple meanings (such as whether the mention of a city constituted reference to the country in which it is located), and by then spot-checking their work to prevent systematic error.

In the case of the thematic data set, on the other hand, we did not have the time to content-analyze each paragraph of every text (as it was, roughly ten minutes were required to code the 49 variables for each paragraph), and thus we were compelled to focus on a sample of them. This we did by consulting with statisticians, who assured us that an analysis of 10 percent of the paragraphs was sufficient to achieve a representative sample, and by selecting the paragraphs through a table of random numbers (with transition paragraphs in which the author locates moves from one topic to another being skipped over in order to maximize the analysis of substantive content). The selection process was carried out within each chapter so as to insure coverage of the entire text. Four procedures were used to enhance intercoder reliability in the use of the thematic categories. First, we spent a full twenty weeks (approximately eighty hours) during the winter and spring of 1973 (at Ohio State University) meeting together as a group, identifying, clarifying, pre-testing, and revising the thematic variables. In so doing, we considerably narrowed our differences over the meaning of concepts and the rules for tracing them in the paragraphs. Second, for none of the texts were the paragraphs coded by only one or two of us; rather, the paragraphs for each book were distributed among us for coding, thereby minimizing the degree to which any one of us might introduce systematic error. Third, so as to further prevent systematic bias and achieve uniformity in the coding, the senior author checked the coding of every paragraph and altered those codes for particular variables that appeared to stem from one or two coders holding conceptions of the variables that deviated from that shared by the rest of the group. Fourth, because two aspects of the paragraphs—their “prime units” (i.e., their central foci) and six cause-and-effect variables—proved especially resistant to convergence during the pre-testing, only the senior author identified the prime unit in every paragraph which served as the basis for assessing the other variables, and alone coded the variables that required judgments about cause and effect.

Ideally, of course, we ought to have also used a fifth technique to insure reliability: that of having a substantial sample of the paragraphs coded by all the coders and then calculating the degree of agreement or disagreement among us on each variable. Early in the coding process we did carry out this procedure on twelve paragraphs. By the time we realized that we had not subjected a sufficient number of paragraphs to this test, several of us had left Ohio State; it proved impossible to reconvene the group in order to carry out further intercoder reliability checks. The results on the initial twelve paragraphs were not discouraging, however. Several variables were found to be troublesome, but most yielded high intercoder reliability scores. (A score exceeding .90 was recorded on 38 of 43 variables, and the mean score for these 43 variables was .95.) In checking the coding of each paragraph, the senior author was especially attentive to those variables that had earlier failed to yield high scores.

In sum, we are confident that our procedures minimized the presence of systematic bias in the thematic data set. Any such bias that remains is that of the senior author who explicitly exercised it in the context of the consensual judgment of the other authors.

²¹ See, for example, Flesch, Rudolf, How to Test Readability (New York: Harper and Brothers 1951)Google Scholar; and Gillen, Barry, “Readability and Human Interest Scores of ThirtyFour Current Introductory Psychology Texts,” American Psychologist, XXVIII (November 1973), 1010–11.CrossRefGoogle Scholar

²² It is noteworthy that a similar pattern, perhaps also stemming from the license which the absence of a dominant paradigm affords, was discovered with respect to the inclination of textwriters to provide their readers with lists of additional sources to consult for more thorough treatments of the major topics covered by each chapter: while 15 of the texts contain such lists, n did not facilitate learning through the use of this device.

²³ Our coding rule for the identification of explicit values was to ask, “Does the paragraph contain any explicit indication of the textwriter's values?” A simple affirmative or negative response was recorded in each case; any doubt was resolved in favor of a negative answer.

²⁴ Our coding rule for differentiating between explicit and implicit recourse to the materials of other disciplines read as follows: “The paragraph or footnotes must indicate that it is drawing on other knowledge bases, rather than simply using or citing these bases (e.g., a discussion of international trade would be coded as [no recourse to another discipline] unless it is noted that the discussion depends on or draws from economic theory or data).”

²⁵ Our coding rule for distinguishing an explicit recourse to such techniques read as follows: “The technique must be specifically identified. Code as [reference to technique not made] if the results of the use of a technique are given, but the technique itself not specifically identified. Simply to use a technique is not to refer to it.” Separate subcategories for generating (content analysis, simulation, survey research, etc.) and analyzing (correlational methods, factor analysis, etc.) techniques were included in the coding scheme, but so few paragraphs were coded in them that the results did not seem worthy of presentation in Table 6.

²⁶ The prime unit of each randomly selected paragraph served as the basis for coding virtually all the variables discussed hereafter. It might be a person, a role, a process, a concept, a country, an event, an international organization—that is, whatever constituted the main subject or concern of the paragraph (as determined by the senior author on the basis of a close reading of the paragraph and, where necessary for clarification, the preceding three paragraphs).

If a paragraph contained a definition of a phenomenon not identified as the prime unit, it was coded as not having a definition. Similarly (to give one other illustration of the centrality of the prime unit to our analysis), in our effort to trace the degree to which the textwriters were concerned with the outcomes of behaviors or situations, we included a “consequences” variable and a paragraph was coded as citing a consequence only if the prime unit was posited as giving rise to it. Consequences caused by entities other than the prime unit were not coded. Originally we sought to develop a scheme in which our variables were applied to every actor or entity mentioned in a paragraph, but this proved much more unmanageable and unreliable than focusing on what the textwriter seemed to treat as the main subject of the paragraph.

²⁷ The data on this variable were created by affirmative or negative responses to the question, “Is the prime unit of the paragraph [explicitly or implicitly] identified in and of itself in time or space?” Four coding rules were employed to reach a decision in each case: (a) References to times and places that were not part of the prime unit were ignored, since the coding judgment pertained only to the prime unit “in and of itself.” (b) “For the unit to be located in space, mention must be explicitly made of an event, person, or place that a narrative historian or political geographer could specifically identify in an account or on a map.” (c) “For the unit to be explicitly located in time, mention must be made of a day, month, year, decade, century, etc.” (d) “Implicit time involves specified events without specific time frames (e.g., if the unit is linked to the Korean War without the fact that the conflict occurred in the early 1950s being mentioned).”

²⁸ For an extensive attempt to elaborate this approach to knowledge, see Rosenau, , The Dramas of Politics: An Introduction to the Joys of Inquiry (Boston: Little, Brown 1973)Google Scholar.

²⁹ Our coding rules cited the appearance of the following phrases as examples of contents being treated as self-evident: “obviously,” “quite clear,” “there can be no doubt that …” and “In fact, of course,….”

³⁰ In order to maximize the chances of finding alternative explanations, our coding rules required an affirmative response “if the same prime unit is coded in the three previous paragraphs and alternative formulations of it offered.”

³¹ Following the practice used in coding all the variables, each sample paragraph was coded only once in response to the question; explicit rules favored the extreme alternatives being employed whenever a paragraph contained wording that was applicable to both the middle (unspecified probabilities) alternative and one of the extremes. If, for example, a cautious term such as “possibly” was used along with an incautious one such as “certainly,” the paragraph was coded in terms of the latter.

³² The other substantial difference between the two groups on a methodological variable concerns the types of evidence on which they relied to develop their observations: the More Concrete group used historical examples in 53 percent of their paragraphs, and hypothetical or generalized examples in 20 percent; the comparable figures for the More Abstract group are 27 and 31 percent. These differences can also be easily understood in terms of the distinction between concrete and abstract analysis. Interestingly, again the two groups did not differ by more than 3 percentage points in the extent to which they cited quantitative findings in support of their observations.

³³ The unit of measurement used in the analysis and in Table 15 is mentions per 25,000 words. This ratio is based not on a sample of paragraphs, but rather on every mention in every paragraph (except those in book titles, prefaces, appendices, footnotes, or tables). Mentions of capital cities used as a shorthand for countries (such as “Washington” or “Moscow”) were coded as a country reference. The geographic categories used to code the data were those developed in Russett, Bruce M., Singer, J. David, and Small, Melvin, “National Political Units in the Twentieth Century: A Standardized List,” American Political Science Review, Vol. 62 (September 1968), 932–51CrossRefGoogle Scholar.

³⁴ Furthermore, Frankel, the author of the one text prepared originally for distribution in England, resists ethnocentric temptations and refers to the United States considerably more often than to Great Britain (again see Table 17).

³⁵ Perhaps the best way to summarize our coding rules for these categories is to give an example or two of actors coded in each: (1) “Winston Churchill”; (2) “the American presidency” or “the voter”; (3) “the Nixon Administration” or “the Labor Party”; (4) “bureaucracies” or “dissident factions”; (5) “Albania”; (6) “superpowers”; (7) “the United Nations” or “IBM”; and (8) “transnational professional societies” or “multinational corporations.”

³⁶ Although often the prime units were individual, subgroup, national, or international actors, a number of them were not: thus there is need for caution against treating the data in Tables 20–24 as exclusively descriptive of actors. They also depict the characteristics of situations, wars, crises, schools of thought, and the other topics that constitute the main focus of a paragraph. An insight into the way in which the texts conceptualize actors is best achieved by comparing the columns in Tables 20–24 rather than by interpreting the absolute figures.

³⁷ For an elaboration of this distinction between “in order to” and “because of” statements, see Snyder, Richard C., Bruck, H. W., and Sapin, Burton, Foreign Policy Decision-Making: An Approach to the Study of International Politics (New York: Free Press 1962), 144Google Scholar.

³⁸ The quoted words are taken from our rules for coding each variable.

³⁹ Our coding rules required that at least “two parties” had to be explicitly identified for a prime unit to be treated as an interaction sequence. If the existence of a second actor had to be inferred from the description of the sequence, it was not coded as interaction.

⁴⁰ The coding rule here specified, “by force is meant the use of arms.”

⁴¹ “Conflict refers to differences of opinions, goals, etc., among parties to a situation (including interpersonal and intergroup conflict as well as nonviolent and violent conflict).”

⁴² “Cooperation refers to compromise, accommodation, concerted effort, etc.”

⁴³ McClelland, Charles A. and Hoggard, Gary D., “Conflict Patterns in the Interactions Among Nations,” in Rosenau, , ed., International Politics and Foreign Policy: A Reader in Research and Theory (rev. ed.; New York: Free Press 1969), 711–24.Google Scholar

⁴⁴ These findings are consistent with the impressions of Boulding (fn. 7) and Snyder (fn. 7) recorded more than a decade ago. Both authors decried the imbalance between conflict and cooperation and urged textwriters to pay more attention to integrative processes. Apparently their advice has yet to be taken seriously by the authors of text-books.

⁴⁵ Our coding rules allowed for distributions of power and authority that “may involve superiors, subordinates, or equal parties,” as well as those within “domestic” settings (such as, for example, “the President is more powerful in foreign affairs than Congress”).

⁴⁶ Indeed, one survey reports a slow but continuous growth between 1921 and 1970 of the space that IR textbooks devote to diplomacy. Cf. Plischke, Elmer, “Treatment of ‘Diplomacy’ in International Relations Textbooks,” World Affairs, XXXV (Spring 1973), 328–44Google Scholar.

⁴⁷ Our coding rules defined “policy making” as involving “interaction among role occupants of a specified unit designed to chart a course of action.”

⁴⁸ These findings are also consistent with the impressions of Boulding and of Snyder, recorded a number of years ago. The former (fn. 7, p. 68) admonished textwriters for not concentrating more attention on internal sociological processes; the latter (fn. 7, pp. 473–75) stressed the need for more extensive consideration of foreign policy decision-making processes. Again their advice appears to have gone largely unheeded.

⁴⁹ The index was constructed according to the following formula:

where T is the sum of all the “tools” variables across all paragraphs, S is the sum of all the “stimuli” variables across all paragraphs, E is the sum of all “expression of knowledge” variables across all paragraphs, and N is the number of paragraphs sampled from the text. We chose to construct the index in this fashion for three reasons. First, we wanted the index scores for each of the three dimensions to range from o to 100. The index scores are therefore an average of percentages for each dimension relative to the number of paragraphs. Second, the index treats each of the three major components of an IR text as equal by dividing by the number of items that make up each category. Some may view this as a questionable assumption, but it is quite difficult if not impossible to make a case for differential weighting of the three categories. Third, we controlled for book length by dividing the index score by 3^N, thereby avoiding unintentional weighting of variables according to the number of items included in the index.