Is physics education research based on a representative sample of students? To answer this question we skimmed physics education research papers from three journals for the years 1970-2015 looking for the number of research subjects, the course the subjects were enrolled in, and the institution where the research was conducted. We combined this data with demographics data about these institutions to compile a profile of physics education research subjects, and compared the demographics of this population to those of all students taking physics in the United States. Our results suggest that physics education research subjects, as a whole, are better prepared mathematically and are from a narrow and unrepresentative subset of our intended target physics student populations. For this reason, findings from research may not be as generalizable to all student populations as we have previously assumed.