A comment from Diane Ravitch's blog:
Labor Lawyer January 20, 2013 at 8:36 pm Although “liberal” on most political issues, my personal experience strongly supports tracking — based on ability to do the work — as common sense and heterogenous classes as creating unnecessary obstacles to effective instruction. Common sense says that students in a class will learn much more if a teacher teaches a subject at a single level for 60 minutes rather than three similar subjects at three different levels for 20 minutes each.
If you’ll forgive a sports analogy, a basketball coach would accomplish much more in a 2-hour practice if all participating players were at roughly the same ability level — high, middle, or low — than if 1/3 were outstanding Division I college basketball players, 1/3 were barely competent high school players, and 1/3 were junior high computer nerds who neither played nor liked basketball.
I acknowledge research demonstrating that low-achieving students do better in heterogenous classes than in classes where everyone is low-achieving. My explanation for this result is that, in most public schools, a class where everyone is low-achieving will suffer from minor but endemic misbehavior that effectively prevents effective instruction and that generates strong peer pressure to join in the misbehavior. If this is so, the solution is to track by academic ability (to maximize teaching efficiency) while implementing classroom management reforms in the low-academic-ability classes (to eliminate the endemic misbehavior).
Heterogenous classes minimize the disruptive effects of the misbehaving students by limiting the number of such students in any given class. However, this approach to the problem (of high concentrations of misbehaving students disrupting classes) necessarily increases the number of misbehaving students in those classes that, if tracked based on academic ability, would have relatively few misbehaving students.
In the low-SES-area schools, particularly in the inner-city schools, there are so many potentially misbehaving students relative to the number of likely well-behaved students, that spreading the misbehaving students evenly among all classes has the effect of creating endemic misbehavior in all the classes. Hence the flight of concerned/functional parents from these schools to the charters (and the mediocre test scores in all the classes).
Bottom line: Track by academic ability, but simultaneously implement reforms in the low-academic-ability classes to minimize misbehavior. (Probably, the best way to minimize misbehavior would be to implement reforms starting in pre-K that improved reading/vocabulary for students from low-SES families, so that school would not be so frustrating — but that’s another long comment.)