This is one question that has been bugging me for a long time now. Why do universities consider GRE/TOEFL scores at all? Perhaps it is fine for master's degrees, where there will be an enormous number of applicants whose language abilities cannot be otherwise established. But what is the necessity for a PhD degree? Why not zero in on a small group of candidates based on their profiles and then conduct a Skype interview? That way the professor gets to know both the verbal and the research potential of the student. Why isn't this the case for PhD admissions?
GRE/TOEFL scores are used in a number of different ways (some of which are alluded to above)
- GRE scores are sometimes used as a university-level filter (if your GRE score is < X, then you'll need a strong support letter from your department to get admitted)
- TOEFL scores are used as a filter for giving people TA funding (if your TOEFL score is too low, you can't be assigned to be a TA, and if it's lower, you can't even get RA funding)
- More informally, the GRE/TOEFL scores are used as a "do you even care' filter: for a CS program, a quantitative score less than 600 might be considered to be a warning that the candidate doesn't even care enough to prep for it.
But for Ph.D programs, the GRE/TOEFL are either used as a high-bar disqualification filters to prune applications (in top schools), or as low-bar disqualification filters to prune the non-serious applications.
On the admissions committees I've been on, GRE scores have been used primarily for one purpose, namely dealing with students from out of the way places. Every year, we see a number of applications from students at not very prestigious schools but with perfect grades and letters saying they are the best in years. We suspect that the courses are easy and the competition is not impressive (being the best is meaningless if we doubt the second best is very good), but we don't want to reject someone unfairly. GRE scores give a simple, consistent way to compare these students with those from other schools. Most of them have unimpressive GRE scores, but occasionally they do very well on the GRE, and in those cases we investigate further.
TOEFL scores are another case in which consistency is very helpful. Skype interviews would give more information, but different interviewers would be more or less demanding (plus the interviews would be a lot of work). If you want to set a consistent cut-off, for example for TA support, then a standardized exam may be the right approach. Note also that the administration may not trust faculty interviewers not to exaggerate the English abilities of students they want to admit.
If it weren't for its usefulness in screening the applications from out of the way places, I'd be in favor of eliminating the GRE entirely. However, even making someone jump through a meaningless hoop can actually be a useful filter. In practice, having a successful career requires occasionally doing things you don't care about, for reasons that may not be clearly explained (and might or might not turn out to be justified if they were explained). Some students run into serious psychological issues here. Maybe they can't overcome their disorganization if they don't feel motivated, or maybe they just refuse to participate in anything without a clear justification. It may be unfair, but these students are not likely to be successful in the long run, and it's a waste of time and energy to prepare them for a career that isn't likely to work out. Jumping through hoops like the GRE is a mild test in this direction. If you are too disorganized to sign up for it in time, or if you aren't willing to jump through such a hoop at all, then that's a bad sign for your future career.
My (computer science) department does not require GRE scores at all. A few faculty still use them to evaluate applicants—there's no accounting for taste—but there is certainly no official cutoff (as Suresh suggests). In practice, they're only useful if they're really low.
However, by state law, an international student at my university cannot be hired as a teaching assistant if their TOEFL spoken English score is below 24. (Too many courses were being taught by foreign students with thick accents.) Since my PhD program has a TAship requirement, we must filter out applicants with low TOEFL scores. We can still admit applicants with scores below 24 if someone offers them an RAship, but they have to bring their score up to 24 by the end of their first year. So in practice, you have to have at least a 20 to be considered at all. (And we do conduct phone interviews with borderline cases.)
Firstly, not all schools place weight on GRE/TOEFL for PhD admissions. Some do, Some don't.
A publication and good grades but bad GREs/TOEFL does not immediately guarantee rejection. Nevertheless, I had done some research on this and here is what I learnt :
- GRE (I talk about the old pattern, I have no clue about the new one) tested both Quantitative and Verbal aptitude along with the ability to write and analyze arguments.
- Quant section is essentially a sanity check. A 780/800 or above indicated that the student has well founded basics and he is able to solve everyday math problems. This was meant to test only those parts of mathematics which everyone applying to graduate school does encounter. Thus, no Calculus or Linear Algebra. Universities find this score useful since it is a good indication of the mathematical basics of the student.
- Verbal. This initially irritated me since I saw no benefit in testing my vocubalary for a technical PhD. But later, I learnt that the verbal section actually varies with IQ. There were numerous high IQ societies which gave out invitations on the basis of GREV marks. This is a comparison of the different high IQ socities with their GRE marks requirement. The GRE (together with Quant) was an instrument for checking IQ without making it too blatant. Here are a few citations:
from 5/94 to 9/30/01
(quantitative + verbal + analytic) 1875
Graduate Record Exam (GRE), combined verbal, quantitative, analytical (June 1994 through Sept 2001) : 2180
GRE (verbal + math + analytical scores) 2180 (6/94 till 9/01)
GRE (“old”): a score of 1610 on the “old” GRE (taken before October 1, 1981)
Personally, I think of all this as pseudo-nonsense but c'est la vi
- The argument section is (arguably) the most useful for PhD admissions. The 2 parts (writing and analyzing) are along the lines of how one writes research papers. The writing sections teaches/tests for rigor in ones' arguments and requires that each statement be backed up appropriately. The analysis section tests the students ability to read "actively" just as how one would read a research paper. The question requires the student to point out flaws in the argument presented and the more the number of (realistic and nontrivial) flaws you find, the more you score. GRE tests students for critiquing general arguments such as building a new library since "the old one is far away from campus" and similar flawed arguments while in research a paper could state something similarly inane in that field.
Among Reading, Listening, Speaking and Writing, I think none of them are useful. Reading gets covered in the Comprehension questions in GRE (with a notch higher difficulty). Listening and Speaking are almost useless since it is fairly simple to memorize generic things to say and still get above the 24/30 marks threshold. Writing gets covered in the GRE as well (At higher difficulty). If the university needs to know if the candidate can speak good english, I think an interview is a good alternative or (as some universities seem to be tending to) IELTS is a good option since you actually interact with a examiner rather than recording answers for "The city you love most" to be evaluated later.
But, I Think TOEFL does filter out students with extremely poor English skills (beyond repair for the university), so it does help in spite of having overlaps with GRE and being inherently flawed (IMO).