Recommendations for practice are routinely included in articles that report educational research. Robinson et al. suggest that reports of primary research should not routinely do so. They argue that single primary research studies seldom have sufficient external validity to support claims about practice policy. In this article, I draw on recent statistical research that has formalized subjective notions about generalizability from experiments. I show that even rather large experiments often do not support generalizations to policy-relevant inference populations. This suggests that single primary studies are unlikely to be sufficiently generalizable to support recommendations for practice.