I really shouldn't comment, because I haven't read the article and have only skimmed the posts. However, I'm very curious now about how achievement was measured. Locally, we use the lovely state test, and the administration tries to make sense of beans where a perfect score is still really only "grade level". You simply can't sort kids with a test like that.
The authors of the paper do consider this issue (p19 of the pdf
http://www.class.uh.edu/faculty/simberman/bci_2011.pdf):
"One particular concern is that the lack of results may be due to top-coding of the exams.
Since GT students are high-achievers many of them may not be able to exhibit growth on
achievement tests as they are very close to getting every question correct. To address this, in Online Appendix Figures 7 � 11 we provide distribution plots of raw scores on each of the 7th grade Stanford Achievement Tests for students with Euclidean distances between -10 and 10. In all cases the mass of the distribution is centered quite far from the maximum score. For example in math the modal score is 62 out of 80 while for reading it is 67 out of 84 leaving substantial room for improvement."
