My initial thought was "Yes, you're absolutely right. That must shift the results", and I think that selecting for any particular subset of the overall population will shift it, but given that the researcher is already using a convenience sample (i.e., easily available college students) is it really going to change it more than the convenience sample already does?
Like, I can see there being a difference between a truly random sample from the overall population vs. students who are required to participate, but how much of a difference are you going to see between local students who volunteer vs. local college students doing this 'as part of their homework'.
Yeah, the biggest problem seems to be taking a bunch of extremely hormonal 18-20 year olds in an awkward and idiosyncratic social situation (university in general, and often forced dorm life for undergrads), and then calling that a universal study of human nature. Over and over and over again.
Like, I can see there being a difference between a truly random sample from the overall population vs. students who are required to participate, but how much of a difference are you going to see between local students who volunteer vs. local college students doing this 'as part of their homework'.