For pragmatic language, much of the difference is in expectations. The range of NT development is quite large in children and preadolescents, while a middle/late adolescent is expected to have attained some of the more sophisticated pragmatics skills, so delays in doing so are more significant. You see this happen quite typically in middle school or early high school, when kids with social skills delays get left behind, because their age peers make the leap, and they haven't yet.
On an even simpler level, some of the classic tests for pragmatic language (like the Social Language Development Test) have two forms--an elementary-age form, and an adolescent form, or, start around age 10 (like the Test of Language Competence, now revised as the CELF-5 Metalinguistics, which starts at age 9). The tests given at 10 and 15 may well have been different tests, with more sophisticated and subtle skills assessed on the older form, or the first administration may have been at the bottom of the norms, with a possible floor effect that obscured deficit areas.
The reason the second situation occurs, of course, is because of the first.
