<p>
</p>
<p>Understanding material and generating it are very related, depending on field especially in theory. Indeed, understanding reflects the deep thinking needed to ask good questions.</p>
<p>At the EECS ugrad town hall, discussion turned from classes to research, and one of the professors flat out said “No matter what your grades are, you will not get into Stanford, MIT, or Berkeley for EE/CS if you have done no research. I’ve rejected a 4.0 from Caltech who just took classes and never did research.”</p>
<p>Now take this however you want, but it is pretty clear that research experience is necessary at least in EE/CS. </p>
<p><a href=“http://www.cs.cmu.edu/~harchol/gradschooltalk.pdf[/url]”>http://www.cs.cmu.edu/~harchol/gradschooltalk.pdf</a> confirms this. You can go to the part about admissions and read it yourself.</p>
<p>
</p>
<p>Nobody is disputing that the correlation between understanding material and generating new material is high. But what is even more correlated with generating new material? Actually generating new material, particularly new material of sufficient value to the academic community to merit an A-pub with prime authorship listing (i.e. first-authorship, or ideally sole authorship). If you have done that, then your grades matter little, provided they aren’t horrific.</p>
<p>
</p>
<p>The question is how high that point is, and whether the high grades people seem to have are merely a function of drive, so that people with top publications are also often people with strong grades. For instance I would wager once one is above 3.5 the publications matter most. I think if the publications were great and the GPA were low, then an explanation would be required, because it really should not be that bad to get mostly A’s and high B’s if one is publishing such stuff. A possible explanation could be a narrow undergrad focus on certain topics, and weakness in
others. A PhD program would then decide how much it cares, depending
on whether it has evidence the candidate would pass basic reqs like
Prelims.</p>
<p>@Sakky, in some fields publications tend not to be meaningful,
particularly ones where enormous breadth of understanding is important
to success. A simple, quite good piece of writing demonstrates the ability to generate, but a PhD program is also about generating in the long
term, and producing something meshing together lots of intuition
depending on the field. I suspect perhaps a field like CS is less like this
in some ways, which is why publications are more common in the undergrad years. A very theoretical field with a ridiculously lengthy history tends to be different. I think there, professor recommendations tend to be most important, beyond having a strong profile in terms of
mastering the important courses, since the professors have enough
experience to comment intelligently on what traits the student has,
beyond being able to solve hard problems from class, relevant to
research success.</p>
<p>On the other hand, in other disciplines conceivably the scenario you posted is more likely.</p>
<p>To be clear, I agree once you obtain top quality publications reflective of PhD level work, you are sort of done, but I think in theoretical fields, the percentage is much less than 2. In fact, this is clear from how few people get into the top theory grad programs. I know for a fact most people at those do not even have top publications, and some none at all. And there are only a few top programs, each taking a tiny number of students.</p>