For the past two years National Public Radio reporter Ira Glass has been observing in, reporting on, and generally rummaging through Chicago public schools. First, he spent one year in a high school, and then he spent 1994-95 in an elementary school. His reports are wonderful for their clarity and humanity. Glass neither overplays the triumphs he saw nor obscures the problems and failures he found.
When he called recently seeking sources of data on in-grade retention, I took the occasion to ask him for a summary statement on what he'd seen in two years of close-up school watching. "Is there any hope?" I asked. After a pause, he responded with an answer that focused on the difficulty of reaching many of the children of poverty. He said that he didn't see any hope for the inner-city system without some major effort at reducing class size. Although they are physically present, "about 30% of the kids just aren't there," he said. He thought a much smaller class size would make it possible for teachers to find some way of engaging these psychologically absent children.
Glass' conclusion, based on his own surveillance, is in line with the largest and most systematic study of class size in the country. It is a study that Donald Orlich of Washington State University referred to in the pages of the April 1991 Kappan as the "most significant educational research done in the U.S. during the past 25 years." I'm referring, of course, to the Project STAR (Student/Teacher Achievement Ratio) study, conducted in Tennessee.
One wonders, then, why so little has been heard about class size in the reform discussions of recent years. Could it be that it is because it requires no fancy new equipment or teacher retraining (although the latter might help)? Could it be because it lacks the panache of "new standards" or "restructuring"? Could it be because it costs money? Could it be because the results arrived when the President at that time and his secretary of education (a former Tennessee governor) were both pushing school choice and privatization?
Could it be that certain individuals simply cannot assimilate these data? For instance, in the Brookings Review for fall 1994, Eric Hanushek of the University of Rochester claimed that, while class-size reduction is a "perennial cornerstone of educational reform ... econometric and experimental evidence shows vividly that across-the-board reductions in class size are unlikely to yield discernible gains in overall student achievement." At best, Hanushek is turning a blind eye to the evidence.
Educators' initial interest in smaller class size came about as a result of a 1978 meta-analysis of research on the topic, conducted by Gene V Glass and Mary Lee Smith. They found small improvements as class size decreased to about 15; they found accelerating achievement as class size fell below this point. This makes certain intuitive sense when one thinks about the ultimate reduction in class size: to a one-on-one tutoring situation. Tutoring can improve achievement by two standard deviations.
While the Glass and Smith meta-analysis has received strong criticism, other analyses have also turned up at least conditional support for the notion that small class size improves achievement in reading and math, especially in the early grades and especially for lower-ability students. Robert Slavin's 1989 analysis also found improved achievement for younger students, along with evidence that the effect might not be cumulative. In connection with class size, Hanushek says, "For every study that finds that increases in basic school resources promote higher achievement, another study shows just the opposite." This seems disingenuous.
Actually, in Project STAR, the effects of reductions in class size were not only "discernible" but also important. Project STAR showed reducing class size to be effective, especially for children in the inner city. Although the research was conducted over a five-year period from 1985 to 1990 and although the research has been reported in professional journals, it remains largely ignored by policy makers and reformers - at least at the national level.
Indeed, some reactions to this work have been actively negative. For his pains in describing the research, Orlich received "hate mail like you've never seen" and a denunciation from the NAACP (National Association for the Advancement of Colored People). That was because Orlich used the research to argue against busing for integration. He cited the work of sociologists who contend that busing has contributed to the destruction of the sense of neighborhood and community in the inner cities. Orlich reasoned that the resources used for busing would be better spent in reducing class sizes in neighborhood schools to those found in Project STAR.
Now the STAR research is being revived by Frederick Mosteller, a well-known statistician at Harvard University, in an article to be published in the journal The Future of Children. A research assistant of Mosteller's stumbled onto the study as part of a search of the literature on ability grouping. Mosteller calls Project STAR "one of the greatest experiments in education in the United States." He continues, "Its importance derives in part from being a statewide study and in part from its size. But more important yet is the care taken in the design and execution of the experiment."
In the spring of 1984 the Tennessee legislature adopted an education reform package known as the Better Schools Program. Although the career ladder for teachers received the most media attention, the program contained a modest class-size experiment that the legislature was persuaded to take statewide in 1985 in what became Project STAR.
The STAR researchers sought to reduce the size of participating classes to 15 students, the point at which Glass and Smith found that reduced class size began to produce large effects on achievement. Schools willing to participate were located in about one-third of the state's districts. Those schools had slightly higher per-pupil expenditures and teacher salaries than the state average, but they had very similar pupil/teacher ratios and percentages of teachers with advanced degrees. Standardized test scores for participating schools were slightly below the state average.
Students in the STAR schools were randomly assigned to one of three classes: a regular class (22 to 25 students), a small class (13 to 17 students), or a regular class with an aide.(1) Teachers were assigned randomly, and apart from class size no other matters - such as curriculum, materials, or scheduling - were addressed.
Cross-sectional analyses showed that students in small classrooms outperformed those in larger classrooms in all grades. The effect size in grade 1 was .34 for reading and .33 for math. Longitudinal analyses showed that the gains were sustained even after the children moved into regular classes in the upper elementary grades. The effect sizes for minority students were consistently larger than those for white students. Jeremy Finn of the State University of New York at Buffalo, who was a statistical and research design consultant to the project, advised me by phone that a measure of school engagement at grade 4 also favored students who had been in smaller classes.
Finn and Charles Achilles of the University of North Carolina at Greensboro wrote in the fall 1990 issue of the American Educational Research Journal:
This research leaves no doubt that small classes have an advantage over large classes in reading and mathematics in the early primary grades.... Although this experiment yields an unambiguous answer to the question, Is there a class size effect?, other related questions remain unanswered. The issue of whether the benefits offset the costs involved is well beyond the scope of this paper. However, it should be noted that few well-defined interventions have shown as consistent an impact as this one on the performance of minority students in inner-city settings, not to mention both their minority and nonminority peers in other settings.
It is worth noting in passing that Finn's initial interest in the project was in seeing that it was done right from a methodological standpoint. He did not anticipate that the study would find any class-size effects.
Barbara Nye, director of research on STAR at the Center for Excellence in Basic Skills at Tennessee State University, informed me that benefits of smaller classes have now been sustained through seventh grade. Data for the eighth grade were still embargoed at press time. Moreover, advantages have turned up in other core subject areas as well as in reading and math.
Reducing class size to the levels of Project STAR is not cheap. If one assumes 14 pupils for a small class and 23 for a regular class, then a school of small classes would require five teachers for every three needed for a school of regular classes. At the national average teacher salary of about $35,000 a year, the school with small classes would require about $1,000 more per pupil per year. On the other hand, scores go up, and retention rates go down. Since both variables are linked to dropout rates, some savings to society would be realized in later years. Nye also reports that the smaller classes reduced the number of referrals to special education - another saving.
So why hasn't Project STAR had more impact? Well, it has had some. Eleven states have cited it in passing legislation either to reduce or to cap class size in the early grades. Still, media coverage has been limited in comparison to coverage of, say, the reports of international comparisons of mathematics achievement. Nye thinks that the lack of attention stems partly from the fact that people don't generally think that children in Tennessee are "at risk" in the same sense as those in New York or Los Angeles are. (They are.)
Moreover, a study issuing from a regional school, such as Tennessee State University, which conducted the research, lacks the cachet of a study coming from Harvard or Stanford. Mosteller senses that the people managing Project STAR have been so busy with that task that they have lacked the time to devote to serious dissemination efforts. Tennessee, at least, has taken the study seriously (as have such other states as Nevada); it has mandated classes of 17 in grades K-3 by 2002. And it has funded the mandate.
1. The teacher-with-an-aide condition also produced positive results, but they were smaller than those of the small classes. The data reported here come from a variety of sources. "Student/Teacher Achievement Ratio (STAR): Tennessee's K-3 Class Study, Final Summary Report, 1990" is available from Elizabeth Word, Tennessee State Department of Education, Nashville. "The Lasting Benefits Study" is available through Barbara A. Nye at Tennessee State University, Nashville. "Answers About Class Size: A Statewide Experiment," by Jeremy Finn and Charles Achilles, appeared in the fall 1990 issue of the American Educational Research Journal. And Mosteller's article will appear in an upcoming issue of The Future of Children. Kappan readers who have kept back issues for 10 years will find a brief review of Project STAR in Helen Pate Bain and C. M. Achilles, "Interesting Developments on Class Size," in the May 1986 issue.
GERALD W. BRACEY is a research psychologist, writer, and executive director of the Alliance for Curriculum Reform, living in the Washington, D.C. area.…