From Today’s Times: Grouping Students by Ability Regains Favor With Educators:

Though the issue is one of the most frequently studied by education scholars, there is little consensus about grouping’s effects.Some studies indicate that grouping can damage students’ self-esteem by consigning them to lower-tier groups; others suggest that it produces the opposite effect by ensuring that more advanced students do not make their less advanced peers feel inadequate. Some studies conclude that grouping improves test scores in students of all levels, others that it helps high-achieving students while harming low-achieving ones, and still others say that it has little effect.

To me it seems obvious that a judicious grouping and teaching by ability is a good idea, for precisely the reasons that advocates outline in this article. Yet I’m flummoxed that in a country that has been debating this issue for two decades as tens of millions of schoolchildren go through the system each year, we apparently lack reliable data on whether it does work, or what makes the difference between making it work and having it fail. I recognize there are advantages to having many different school districts try many different things. But I can’t help but think we’re hampered by the fragmentation of the U.S. education system and what seems a lack of national-scale programs that can run controlled classroom experiments in such a vast population. We’re running lots of little experiments but have no way to compare their results. We have tens of thousands of experiments with n numbers of 15 or 25, but nothing to show for it but anecdotes of the sort related in this article. As a result, every district hashes this out in its clumsy way.

The U.S. education system much resembles the U.S. healthcare system that way: massive numbers, but very little data on what works.