To the Editor:
Class-size reduction merits both new research and continuing analyses of Tennessee’s Project STAR class-size data (“Class-Size Reductions Seen of Limited Help on Achievement Gap,” Feb. 27, 2008). The STAR (short for Student/Teacher Achievement Ratio) experiment, conducted from 1985 to 1989, is old enough to be replicated as well as represented accurately.
The principal investigators of the study, of whom I was one, are pleased that STAR and its database are used. Its results could serve as a basis for future study, and should not be omitted from a discussion as a rationale for a new study.
Along with Jeremy Finn, I have researched small classes from 1982 to 2008 using both STAR data and studies in other settings, such as North Carolina. The constellation of STAR-related class-size studies provides much data that researchers often omit; perhaps it is easier to obtain test scores than dig into the STAR report.
Spyros Konstantopoulos, whose review of STAR data is discussed in your article, is correct that “manipulating class size” doesn’t appear to reduce achievement gaps. His comments are apparently based on his examination of Stanford Achievement Test outcomes, only one of several measures used in STAR. The STAR study identified other key elements, and a careful reading of its report might have softened his call for a randomized experiment to “include observations of teaching practices and interactions between students and teachers in small classes,” as you report.
The STAR researchers seldom say that “manipulating class size” by itself made all of the differences obtained. Results similar to STAR outcomes will not be obtained if researchers omit key experimental elements: random assignment, or heterogeneity; early intervention; intensity; duration (three or more years of the intervention); and a cohort.
Considerable data on teaching practices and interactions are detailed in the STAR technical report, “The State of Tennessee’s Student/Teacher Achievement Ratio (STAR) Project,” which includes information on the following: descriptive data and teacher effectiveness; effects of class size on classroom processes; teacher grouping practices; and effects of reduced class size on curriculum, instruction, and teacher-child interactions. A detailed appendix on teacher-effectiveness practices includes 25 pages of observation instruments, data, and rater agreements. Principal investigators of STAR also conducted separate studies on teacher-student interactions and teaching practices.
Much can be learned from a careful reading of the STAR report, articles about the experiment written by investigators, reanalyses of the data, and comparative analyses of STAR data with other databases. A new study, however, probably is overdue.
Charles M. Achilles
College of Education
Seton Hall University
South Orange, N.J.
