[throwawayjava]

Yes, 2 full standard deviations would be the higher (ie, lower) extreme of my "within a std.dev. or two" ballpark.

I use that ballpark because I'm told it's (very roughly) the population that my students were randomly selected from. My methodology isn't perfect, to be sure, but I'm not pulling numbers out of thin air :)

The specific criteria is that none of my students had an IEP. And, by definition, everyone more than two standard deviations from mean had an IEP.

Now, probably there are a lot of students with IEPs who are less than two standard deviations from the mean. Hard to know who those students were in terms of the IQ distribution. Probably a lot of them are on the lower end, but possibly learning disabilities are more smoothly distributed than that. And I also don't know how many students were excluded, so even if we knew where to concentrate their mass in the IQ distribution, it wouldn't help.

It's worth noting that this means I also never worked with students who tested above two standard deviations from the mean. "Gifted" students had their own special education that ran concurrently with the period of the day when the enrichment activities were done. I guess the assumption is they'd be fine, but I'm told by some teachers that's not necessarily the case.