This paper analyzes the 19 studies presented to the National Institute of Education's (NIE) panel on the effects of school desegregation on black achievement and discusses the author's own findings. The author concludes that desegregation did not cause any decrease in black achievement generally, nor did it cause any increase in math achievement. Although desegregation increased mean reading levels, the distribution of reading effects appeared to be skewed, with a disproportionate number of school districts obtaining atypically high gains. Studies with the largest gains were characterized along a number of methodological and substantive dimensions (none of which could be isolated as causes of the atypically high reading gains) including: small sample size, two or more years of desegregation, desegregated children who outperformed their segregated counterparts even before desegregation began, and desegregation that occurred earlier, was voluntary, occurred in schools with larger percentages of whites, and was associated with enrichment programs. Because of the small samples in the NIE project, and the apparently non-normal distributions, the author states he is not confident that anything has been learned about desegregation's effects on reading on the average. Across the few studies examined, he found that variability in effect sizes was more striking and less well understood than any measure of central tendency. The paper ends with a review of the implications of the findings for various interest groups and a summary of the implications the NIE project has for theories of research synthesis. (CMG)