What is "generalizable?"  In short, can it be applied to a population beyond the group studied?  A study that is to be applied universally should include an appropriate representation for age, race, gender, health, socioeconomic status and education level, among other factors.  Testing a group of white men in their 40s may not prove anything for an African-American women in her 70s, for example.  To have the best results, researchers should put together a study group that can be "generalized" across a population.