GRE scores:your opinion

Evan W. Steeg steeg at csri.toronto.edu
Mon Feb 7 17:24:57 EST 1994


In article <1994Feb7.085608.3662 at tower> afc at gnv.ifas.ufl.edu (Andrew Cockburn) writes:
>In article <94Feb4.132100edt.324 at neuron.ai.toronto.edu>, steeg at cs.toronto.edu ("Evan W. Steeg") writes:
><deleted>
>>   This common-sense assessment is in accord with my (admittedly limited
>> and "anecdotal") observations.  Having spent some time in and around
>> some very good graduate departments (Computer Science and Biochemistry
>> at Cornell, Computer Science at Toronto), I have encountered very few
>> students who did not do exceedingly well on GREs (and SATs, etc.).
>> I have likewise known very few who had any real intellectual difficulty
>> with the standard first- and second-year graduate courses.  However,
>> I have known *many* who ran into serious trouble when asked to do
>> original research, and many who eventually dropped out of graduate
>> school for essentially this reason.
>> 
>>   Perhaps the kind of "success" for which GREs are such a good predictor
>> is not the most important kind of success in postgraduate science
>> training?
>> 
>The reason that most of your fellow students in good departments seem to
>have little problem with either GREs or academic classwork is that the
>selecting committee eliminated those that would have problems, in part 
>by selecting students with high GREs.  

  Yes, that's exactly what I was saying.  The students were chosen
on the basis of their ability to answer shallow questions on deep
topics, under time pressure and nervous stress.

> The typical applicant to such
>departments would have a great deal of difficulty with the course work.
>
>In my experience, whenever we lower our standards to let in a marginally
>qualified applicant, that person will almost always fail to become a 
>productive researcher.  Raw number crunching ability (which is more or
>less what the quantititative part of the GRE measures) is a necessary
>but not sufficient condition for success in science.
>
>Your argument is rather like saying that because not every 7 foot (215 cm)
>tall person does not become a basketball star, professional basketball
>teams should not preferentially hire 7 foot tall players.
>

  A facile, even rude, parody of my argument, Dr. Cockburn -- but
let's continue your metaphor for a moment.  It is in fact the case
that height and some other attributes (quickness, agility,
ball-handling skill) are somewhat negatively correlated.  So, sure, if
I as an NBA coach could draft a team of 7-footers who run, dribble,
and shoot as well as the best 6-foot players, I'd do so.  But I can't.
And therefore if I draft my team using *only* height as a criterion,
or even overemphasizing height among other criteria, I will not put
together the optimal team.

 More to the point, is "raw number crunching ability" really as
important in scientific *research* -- as opposed to scientific
test-taking and course-passing -- as so many so unquestioningly
assume?  *Especially* when this ability is traded off against
qualities that are more obviously related to research ability but
which perhaps may be more difficult to measure (or, at least, more
difficult to measure by the narrow methods of the SAT/GRE monopoly)?

  Your faith in the GRE approach is clear from your paragraph above in
which you equate "standards" and "qualified" with high GRE scores.
However, I question whether raw number crunching ability (at the
levels recommended for acceptance into good science graduate programs)
is sufficient *or* even necessary for *long-term* postgraduate and
career success.  As I said, I have in fact known students who aced
short time-pressure exams easily but who couldn't research their way
out of the proverbial paper bag.  Such students have invariably run
into trouble, sooner or later.  Sooner, when professors actually take
the time to devise good courses and good exams; later, if the high-GRE
weak-research students slip through a set of course exams of the same
shallow, time-pressure kind as the GRE, and then hit the wall when
they try to do research.  When students quit or fail before finishing
their advanced degrees, this is a sad waste of personal and
institutional resources.  I have also, by the way, known students who
don't do well on number-crunching exams but who compensate with
impressive amounts of imagination, diligence, and persistence.  Some
of these people were given a chance -- they were allowed in to good
grad departments despite their bad scores, and are embarking on
successful research careers.  Others never got the chance -- their
loss, and society's loss.

  Look, I actually sat an exam, for a graduate CS course on compiler
design, that required us to *parse strings*. We were not required to
analyze grammars, discuss the merits of different grammars, state or
prove results about the computational complexity of parsing strings;
we had to parse strings.  We had to simulate simple automata, under
time pressure.  This is like the GRE.  This is also like lining up a
group of people and asking them to count down from 5000 by 27s, as
fast as they can.  If I asked you, Walter Gilbert, Jim Watson, Lee
Hood, and other prominent scientists to do this, and I timed you, I
would get a distribution of times, some better, some worse.  Some of
you I'd "fail" and others I'd allow to "pass".  How much would these
results say about your ability to do science?

 If departments want to use GREs simply because it's cheap and easy 
and doesn't require a lot of thought by the admissions committee 
members, they ought to just admit it.  But they shouldn't have any
illusions -- many potentially good researchers are kept out of
science, and many "good students" drop out of grad school without
ever finishing a thesis or publishing a paper.

  Evan





More information about the Bioforum mailing list