<p>Actually, I just came across a study of participants of NSF REUs. In terms of grades, the BOTTOM quartile were students with a GPA of 3.4 or lower (top quartile was 3.9 or higher). Looking at this, the average graduate from many selective institutions (e.g. Berkeley, Harvey Mudd, Princeton) would be in that bottom group. So, it is completely understandable if these students get upset about the grading and throw in the towel. The irony, of course, is that while much of scientific progress depends on collaboration, the system discourages cooperation among students.</p>