Using only college students as participants but wanting to make generalizations about the entire 18 - 23 year old population (not just students) is problematic due to lack of:

random assignment

random sampling

non-response bias

manipulation