power and authority. In every organization whose members are professionals or semiprofessionals, there is an inherent strain between the principles of bureaucracy and colleagueship, which are shorthand terms for very different kinds of power relations among the members of an organization. And this is true in education as well. Anything which sheds light on the sources and weaknesses of professionalism within a group, as the Mason study does for teaching, also sheds light on the actual workings and administrative processes of the organizations in which members of that group work, and is thus a contribution to organizational theory.

But while the Mason study, done on a nation-wide basis, has broad implications for the recruitment and retention of teachers, and thus for the profession as a whole, it is at least as suggestive of additional research on the local level. For this nation-wide study does not tell us anything about the conditions under which the flow of teachers out of the classroom is especially low. Many school administrators would like to retain a larger proportion of their men teachers, and especially the better ones, in their classrooms. How can survey research contribute to this end? First, we would wish to compare schools and districts which differ markedly in their retention rates. There are certainly schools and districts where the turnover of male teachers is very high and others where it is much lower. When we make these comparisons we would, of course, have certain obvious questions in mind; we would certainly look into the possibility that in the high turnover situations average salaries and job satisfaction are lower than where the turnover is less. This may or may not be the case; in any event, this kind of comparative research can subject our hunches about the bearing of salary on teacher retention to some empirical examination. Further, we can explore the data for the operation of forces bearing on teacher turnover about which we now scarcely even have hunches. For example, the study by Mason and his colleagues supports the general impression that men teachers are especially concerned with the pay and working conditions of teaching, and that deficiencies in this regard lead many to plan to leave the classroom. But through comparative studies of schools and systems we may discover that, within limits, proportionately as many men leave teaching where the pay is good as where it is low, and that the appraisal of their own situations is based not so much on the absolute levels of their pay as on their comparisons of their own situations with that of others. For many male teachers, the "others" with whom they compare themselves are likely to be school administrators, who are largely recruited from male teachers and whose pay and status are very often much higher than that of teachers. It is an hypothesis worthy and possible of test that loss of men teachers from the classroom is related to the size of the differences in pay and status between them and their own school administrators. If this is so, the policy implication is that if it is desired to retain more of the most able male teachers in the classroom, the difference between rewards of teachers and administrators must be kept low--not necessarily by holding down the pay of administrators, but perhaps by providing an alternative channel of advancement within classroom teaching. While this is at the moment largely speculative, it is the kind of hypothesis with important policy implications that survey research could illuminate.

I have not thus far said anything about asking teachers why they leave teaching or whether they are satisfied with their salaries. A very common error in survey research in education has been to assume that the individuals in the situation could adequately account for their own behavior; and moreover, that by summing up individual answers to such questions we understand a trend or pattern of behavior. But a basic principle in survey research is that the question "why did Smith quit his job? is fundamentally different from the question "why is the quit rate at School A 30% annually while it is only 10% at similar School B?" These are two fundamentally different questions; it is the latter kind of question to which survey research properly directs its attention, and such questions cannot be answered by simply collecting a great many answers to the first kind of question about why Smith quit his job. Smith may have quit his job for any one of a thousand reasons or combinations of reasons, very few of which we can learn much about through a survey interview. But if the turnover rate in one school is three times as high as in another, the difference very likely flows from some differences in the schools as social systems, or the nature of the local community or the administrative staff, or the kinds of people recruited to the given school, and these are things we can learn about through survey research.

Now these are not easy things to learn about, but they are to a large extent independent of the idiosyncratic circumstances surrounding each individual case. If a given teacher tells us that she is leaving her job because her husband is being transferred by his company, we may know all we need to about that case, but we have learned almost nothing about the forces which make for high or low quit rates. We are not likely to learn much more even if we ask a thousand teachers why they are leaving their jobs, and present a summation of their answers, as is so often done, as an explanation of the quit rates. It is a mistake to assume that the individual respondents--whether students or teachers-whom we question can give us a useful explanation of differences in distributions of behavior to which they contribute but of which they may not even be aware. This is asking the respondent to do the analysis for us, and it is probably the single most common error for people doing research through survey methods in education. Individual respondents can tell us about themselves how they perceive or feel about something, how they act or how they plan to act; survey research is concerned with the rates or distributions of these perceptions, sentiments, behaviors, statuses, and the relations of these distributions to other variables. Through the close and patient study of many such relationships, the analyst himself attempts to arrive at an explanation of the phenomenon, a phenomenon which always takes the form of a difference in rates--in this case, of rates of teacher turnover. 7

Let me illustrate what I have been saying by reference to another area of education-school counseling. Counseling, as you all know, has been the focus of considerable attention in recent years. In Conant's recommendations, counseling in the high schools is the major device for sorting students of different interests and abilities into programs most suitable to their talents.8 And the very large sums allocated to counseling in the National Defense Education Act indicates the interest of the federal government in the role of counseling in identifying and encouraging talented students, especially those interested