Four Critical Years: A Book Review

9 lectual self-esteem was negatively associated with attending a selective Pro­ testant college and also with attending a public university (selective or nonselective)." I assume that Astin deliberately chose the approach of total disclosure rather than selectivity in presenting his findings, so as not to bias the reader by his own integration of the results. But the effect is overwhelming and we turn to Astin's final summary like a drowning person gasping for air. One final question about the present work, also follows as a consequence of his multiple regression approach. I feel it leads to findings that really do not provide a firm underpinning for the broad recommendations that Astin makes. In all his work Astin has stressed the importance of controlling on the characteristics of the entering freshmen. Indeed, one of his major con­ tributions has been to demonstrate how much variance in outcomes was ex­ plained by entering characteristics, and to caution us against overemphasi­ zing the impact of college. His present data confirm this view. Astin does not present systematic data on betas and partials, but the tables in the ap­ pendix indicate that, for most of the outcomes he studied, little is contri­ buted to the Multiple R once entering characteristics are controlled. He ba­ ses his recommendations on statistically significant factors, but with the size of his N's, statistical significance can occur with minute contribution to the variance. Instead of joining the general alarm about the disadvan­ tages of public education, large size, community colleges and commuter sta­ tus, it seems that a more appropriate stance might have been for Astin to caution us that all of the concern is exaggerated, and that these factors contribute minimally to the variance. On balance, the value of Astin's contributions far outweigh the concerns I have raised. I have stressed the concerns because I feel they are particu­ larly pertinent in the present work, where Astin has broadened the student outcomes with which he is concerned, to variables less amenable to the ex­ traction of clear meaning given the methodology he uses. FOUR CRITICAL YEARS: A Book Review by Arthur W. Chickering Director, Center for the Study of Higher Education Distinguished Professor of Higher Education Memphis State University The significance of Astin's Cooperative Institutional Research Program-its data base, longitudinal perspectives, and susceptabi1ity to diverse mul­ tivariate statistical analyses— is dramatically demonstrated by Four Critical Years. Moreover, this report documents the great value of research programs sustained through time and shows us how that value increases with each pas­ sing year. It takes little imagination to project from this study an array of future studies which can provide complex and fundamental information con­ cerning a wide range of policy decisions facing colleges and universities during the next decade. Higher education will be well served if the talent, energy, and resources which have characterized this program in the past con­ tinue well into the future. Of course there are limitations in the data, research methods,.and sta­ tistical analyses employed. In longitudinal research one always discovers in subsequent years omitted areas of inquiry that have become increasingly cri­ tical and, conversely, one finds that apparently significant issues addressed by the original data turn out to have died in infancy. If Astin and his as­ sociates could warp time and step back into the 60's to revise and readmini­ ster their questionnaire, I am sure the collection of items would be signifi­ 10 cantly modified; if that privilege were extended to us, most readers would do likewise, each according to his or her particular predilections. While large scale survey research is an effective actuarial tool which can contribute powerfully to general policy questions, it is not useful for individual persons or single institutions except as a most general guide. With such large numbers the very small differences which can be statistically significant may have limited practical significance for individual decisions. A rough scan indicates that most of Astin's correlation coefficients are be­ low .50, leaving much of the variance unaccounted for. Now that is not unus­ ual in this kind of research, but we need to recognize it. And Astin prop­ erly...