Computer Science Students Association
unc chapel hill

Life, the Universe, and Everything!
(a.k.a What you're expected to know during Orals)
The 3rd guide in the Hitchhiker's trilogy
home  about us  students  guides 
About this advice
Quotes [top]
"Any man who reads too much and uses his own brain too little falls
 into lazy habits of thinking."
 - Albert Einstein

"One must learn by doing the thing; though you think you know it,
 you have no certainty until you try."
 - Sophocles
 

Who should read this document?[top]
The object of this paper is to discuss the Oral exams and how to prepare for them.  It is aimed at the doctoral student getting ready for Orals, in the hopes that he or she will avoid some mistakes that those of us who already went through it made.

Orals are not required for pure M.S. students, so they need not read this document.

What makes you an authority on this subject? [top]
Nothing, other than having taken and passed the Orals, discussing the subject with several other students and professors, and having a desire to share my experiences with others.  As any reader of the previous Hitchhiker guides should know, this is an entirely unofficial guide, not sanctioned by the Department or guaranteed in any fashion to be correct.  You assume all risk in following the suggestions in this guide.  But if it works, I want a 5% cut in your pay raise after you pass. :-)

Ok, I understand.  Now just one question.
Sure, name it...

Just what the heck are "Orals?" [top]
The Doctoral Oral Exam is the last chance the Faculty has to kick you out of the Department.  (There's the Defense, but I've never heard of a student failing the Defense.)  You must take them and pass as part of the requirements for the Ph.D.
When do Orals occur? [top]
The current schedule is to hold the orals after the start of the Fall semester. You and your advisor and can determine when you are ready to take the orals. Usaully you should take the orals at the start of your second year.
Don't I need to have a topic approved first? [top]
You need not have passed your Ph.D. proposal before taking Orals, although that's how it's usually been done in the graphics group.All you need is a committee and a basic idea of the area where you want to do your dissertation.  The test might be easier if you take it before proposal rather than after, since the topics will be broader and simpler.  It's something to consider.
Ok, I've decided to take it.  What am I up against? [top]
The Oral exam consists of your committee questioning you for up to 3 hours on several topics related to the classes you have taken in your first year. As the name suggests, both the questions and answers are given orally.  You should not depend on written notes. For each topic, one professor acts as a primary griller, but any other professor can jump in and ask supplementary questions. (It's important to remember that although one professor interrogates you at a time, five or more are listening, so don't tailor your answers solely to the one who asked the question.)  Generally, you're expected to stand during the test and answer the questions or work out problems on the white board.  The length of each session varies, depending on the number of topics you have.  After the test, your committee kicks you out of the room to decide whether or not you passed.  At least two-thirds must vote in your favor for a pass (although the vote is generally unanimous).  You learn their verdict within a few days of
completing all 3 hours.


What if I don't pass? [top]

If you don't pass, you get another chance.  After a minimum three month wait, you can retake the test.  If you fail the second attempt, you have to petition to get a third shot, and you better have a champion on the Faculty willing to fight for you.


What is the pass rate? [top]

Orals are generally a weeder. However, recently the pass rate on the first attempt has only been around 50%, which seems rather low.  I know one person who failed on the second attempt and unfortunately had to leave the Department, and another who took it three times and passed on the third attempt. So while most people eventually pass, it's not a cakewalk.


What are the topics? [top]

The topics will be related to the classes you have taken. These classes should have a distribution determined in the department web pages (???). Picking class you have taken. Might need to take classes you have not taken here. That is more difficult as the material might be covered in a different way.
Oh, great.  What can you help me with, then? [top]
(???) more advice on classess.
Ok, I have my topics settled.  How long should I study? [top]
Some people walk right into the test with no preparation. A few have passed that way.  I don't recommend trying it, though.  I have heard rumors that some members of the Faculty, when faced with a student who waltzes in with no preparation, will ask harder questions with the specific purpose of keeping the student's ego in check (since you have got to be pretty damn cocky to do such a thing!)

    (???) One month

Obviously, the preparation period is going to vary greatly, but two months seems to be the typical timeframe.  Some mix in regular work with that, and others withdraw completely and do nothing but
study, but two months is typical.

Unless your background is particularly strong before you start studying, you're going to need time to acquire and read the important and representative papers in each topic.  Factor that into your estimate of preparation time.


What should I study?  I don't have any practice questions! [top]

This is an important dilemma.  To give a thorough answer, we need to look at the philosophy behind the Orals. (Much of this is based on conversations with Dr. Brooks, Dr. Fuchs, and other professors.)

Your committee wants you to succeed, but will not give you a free ride.  This is the basic mindset of your executioners going into the exam.  Deep down, they are on your side and want nothing more than to pass you.  If this isn't true, you need to pick a new committee!  But you must achieve a basic standard of performance before they can do so with a clean conscience.

What does they want out of Orals, besides fulfilling the requirement of the 2nd doctoral test mandated by the Graduate School?  Two things:
 

  • Does this student have the proper background to successfully complete his dissertation research?
  • Is this student able to field and answer questions smoothly and competently enough that he or she won't embarrass the Department after graduation during interviews, presentations, etc.?
    1. You may get a wide distribution of questions, ranging from extremely simple to devilishly tricky.  But based on these two goals, the ones you have to watch out for are basic questions and questions relating directly to your dissertation.  If a professor asks you some highly specific question on something only superficially related to what you're doing, then you have a chance to impress your committee, but if you can't answer it, it's not a big deal.  In fact, the other committee members may get after that professor for asking such an esoteric question.  But if you get asked a basic question, or one that is directly related to your dissertation topic, and you can't answer it, that looks bad.  Really bad.  This is the kind of thing you must avoid to convince your committee that you can satisfy requirement #2.  So the Two Commandments of orals preparation are:
    1. Thou shalt master the *fundamentals* of thy topics before studying the esoterics.
    2. Thou shalt know all of the problems with thy dissertation and be prepared to answer questions on them.
    If you take Orals before passing proposal, then Commandment #2 becomes more general: knowing the important problems within the area you want to do your dissertation in.

    If you do a good job in following these Two Commandments, you stand a decent chance in anticipating the most likely questions that will get thrown at you.  I think a big factor in my success during the Orals was giving good answers to several questions -- because I had thought of them before the test and spent some time beforehand thinking them through!


    Great.  But I still don't have any practice questions! [top]

    Get a study group together with other students who share similar topic areas.  Ask professors to come in and grill you with questions in a mock oral exam.  That will help expose you to the style of questions.  (Note: most, but not all professors are willing to do this.)  It's usually ok to ask members of your committee to do this, although I had mine grill me on different subject areas from the ones they would cover in the real exam.
    Previous orals are loosely transcribed and kept as part of a student's record.  Ask a student in a similar field who already took the test if he could describe the questions he had to answer.
    Questions usually follow a typical trend.  The professor generally has a basic plan of attack for the areas he wants to probe you on. It starts out with a broad and simple question, and eventually narrows down to a specific area or problem.  Along the way, the professor will think of other questions, based on your responses, and may pursue various aspects of such side issues before returning to the main thrust of the question.  It's not unusual to run out of time before he can get to the ultimate point he wanted to ask you about!
    Unfortunately, terminology is part of what you must know to convince your committee that you are competent.  This leads to what I call "read my mind" questions -- you talk about some phenomenon and then a prof asks you, "And what is the word we use for this?" I hate those.
    I'm not used to oral exams.  What can I do? [top]
    If your education was like mine, you had very few oral tests. Written tests are simply more efficient; it's tough for a professor to individually quiz each student in a class of 200!  Through high school, undergrad, and grad school, I had a total of two oral exams before the big one.  This lack of experience leads to justifiable anxiety.

    Oral exams are fundamentally different from written tests.  On the DWE, you can flounder on a question for 40 minutes, then get a flash of inspiration, and madly scribble down the answer in 5 minutes.  Or you can write an answer, come back to it, realize that you were an idiot, erase it and write down the correct answer. Neither of these options are available on an oral test. Once a question is posed, you've got between 10 and 60 seconds to think before a professor will try to prod you in some way to help you find the answer.

    Then again, an oral test offers advantages that are unavailable on written tests.  If you're not sure what a question is asking, or you need to make assumptions, you can get immediate feedback from the questioner.  You can communicate in ways other than words -- by using hand gestures, body language, voice inflection, and drawings on the white board that you can modify while the professors watch.

    Part of the test is how well can you present yourself in front of a group of people.  It's not just a measure of whether you know your stuff, it's how well you can communicate. How you answer a question is about as important as the content of your answer.  If you're not a confident speaker, it's worth some effort to get practice before the Orals.

    If you're not used to oral tests, how can you prepare for them? One way is to teach a class.  You have to do that as part of the Ph.D. requirements anyway.  It gives you practice thinking on your feet and on clearly explaining difficult subjects.  Another is to give presentations, either informally within the Department or better yet, externally at a conference.  Answering the questions after your talk will give you practice under fire.  Running demos to technically sophisticated audiences also proved to be helpful practice for me.  This is especially true when the audience is formed of big shots, since they ask blunt, difficult questions, and they're smart enough to reject a bad answer from you.  Finally, get professors in to grill you in mock orals.  (I don't recommend having fellow students do practice grilling, since they tend not to probe deeply enough, which can inspire a false sense of security in the victim.)

    What else did you learn about oral exams? [top]
    Use the white board.  Bring in your own magic markers; do not rely on the ones at the board (they may be dried out).  It helps both you and your professors if you write down part of your answer on the board.  You can also buy a few seconds of time to think by sketching something out on the board.
    You will probably get to pick which professor asks which topic, and the order in which the topics get covered.  Other times, the profs will kick you out of the room right before the exam to settle those two questions.  Be psychologically ready for that to happen.
    Assigning topics to professors is usually easy -- match the expert to the particular subject.  The order of topics is a more interesting issue.  Multiple strategies exist.  You can place the subject you fear most at the end, in the hopes that you do well in the previous ones and run out of time during the last one.  On the other hand, if you're borderline until then, the anchor subject could decide your fate.  The first subject should be one that you think has the least likelihood of derailing you by presenting a question that you just can't handle.  It's important to get at least one topic under your belt, to gain some confidence and build momentum.  I put my best subject right in the middle, in the hope of snowballing early momentum.  Decide what strategy works best for you.
    You get at least one bathroom break, so if you're running into real difficulties, call time out and regroup.
    In learning the basics, don't forget the basic basics!  I had several pages of notes on basic trig, calculus, linear algebra, physics, probability, statistics, and the like.  Since you can't rely on written notes during the test, I spent more time remembering how to rederive formulas rather than memorizing them.  I might not have remembered what the Taylor expansion of sin(x) was, but you can rederive it if you remember the basic expression for a Taylor series.  During a practice session, I forgot what the Fourier transform of a triangle function is, but was able to rederive it. At least one professor I know is infamous for making his students work out math on the white board.
    When you get a question, especially at the start of a session, make sure you understand the question before blithely spouting out an answer.  Initial questions are often vague and depend on unstated assumptions.  If Steve Weiss asks you how to do sorting, you probably should ask him what you're sorting.  If Fred Brooks wants you to draw the graphics pipeline, first find out what rendering technique he has in mind!

    Such interchanges to find out what the professor really wants will impress the committee, but make sure you're doing it to find out what he/she wants, not to steer the question in the direction you want! The professor is in charge during the exam and will cover the areas he is interested in.  This can be frustrating at times, since they tend to quit probing you on a subject that you're doing well on and move on to something else that you might not know so well, but that's the way the game is played.  If you attempt to sidestep or steer away from the posed question, your committee will not be pleased.

    If you're not sure of an answer, qualify it.  If you really don't know the answer, say so.  An inability to say "I don't know" at the proper times will get you into trouble, especially if your professor is trying to determine how deep your knowledge is! Trying to bluff your way past the limits of your knowledge is not going to work when your committee has far more expertise than you do. A much better response is something like "Well, I'm not really sure, but my intuition tells me..." followed by your educated guess.

    At the same time, I've also seen a tendency in students to give up too easily, to say "I don't know" far too quickly.  There's nothing wrong with shutting up and thinking for a minute, or in taking 30 seconds out to work something out on the board before giving your answer.  If a question needs some CPU time, don't be afraid to crunch on it for a while.  Not every question that the professors give to you has an obvious and immediate answer -- they expect you to think about some of them.  Being able to recognize those takes practice and experience.  If you eventually can't find a solution, then say so, but don't give up without a fight.

    The ability to tailor the length of the response based on the difficulty of the question seems to be something that certain professors look for.  This takes judgement and educated guesswork. During the exam, I think I was able to sense which questions were simple and which were difficult.  For the simple ones, they expected concise and direct responses.  But on the difficult ones, they expected some thought and analysis, which requires more time. Giving a long answer to a simple question or a brief answer to something that demands several minutes of analysis first will make you look bad.  I found this out the hard way when I hesitated on a simple question on transforms (fearing that it was a trick question).  That earned me five more minutes of grilling on that subject before my professor moved on to something else.  You should also avoid "core dump" answers where you spew out everything you know on the question and everything related to it. Only give the relevant information to the question at hand, not everything you know about related fields as well.

    In general, your professors would like to see evidence that your knowledge of the topics goes beyond sheer memorization and booklearning.  If time and energy permit, your study program should not be just reading papers.  Work out some problems, try implementing some of the algorithms, or play with some applications.  Discuss the area with other advanced students in your field.  Go to conferences and exhibitions. Supplementing your reading with hands-on experience is the best way to build intuition.  Some of the answers I gave in my exam came from attending SIGGRAPHs and actually building things, rather than anything I read.
    Work out research problems on your own.  Play "what if?" games by changing some key parameter and then seeing how that will affect design decisions.  While the odds of picking a question that you'll get asked during the real thing is very low, the practice will sharpen your skills.
    Keep attending research meetings within the Department.  I didn't think it a good idea to completely cut myself from Department meetings while studying for Orals, since the problems that rise from the meetings can give you food for thought.  They can also inspire a question that one of your professors might ask you (it has happened!) so you can think of it as self-protection as well.
    Cultivate estimation skills.  The ability to do back-of-the-envelope calculations is very useful.  Let's say computer or human vision is one of your topics.  You better know the resolution of the human eye, in terms of degrees of arc.  But let's assume you forgot that fact.  Could you estimate it?  Can you think of a situation where you can barely recognize something at a long distance?  How about looking at an 18-wheel truck from an airplane?  If you know typical cruising altitudes (around 35,000 feet) and the length of the truck (perhaps 40 feet?) you can get a decent estimate of the eye's resolving power.  Doing ballpark estimation is helpful in doing quick analyses on whether something is feasible or not.  Try some of the following as exercises:
    If you're struggling with a question, your professors will try to help you by giving a hint or prodding you in the right direction. There's nothing wrong with needing a hint or a push now and then, so don't be afraid to ask for help occasionally.  But some students prefer to contemplate a problem without outside interference until they reach a solution.  For those students, these well-intended hints only serve to break their concentration and destroy their train of thought.  If you are one of these, discuss this with your committee before the exam.  If you request it, your committee will probably agree not to give any hints or prodding until you explicitly ask for them.
    What does it take to pass?  The unofficial grapevine says that you should do very well on at least one topic.  Screwing up one topic is ok if you perform adequately on all the others.  Screwing up two topics will sink you.  If you fail your first attempt, be absolutely certain that you are bulletproof on the questions that were asked in that first test.  According to one professor, there's no shame in doing badly the first time, but if you can't handle the same question on the second attempt, your professors will be very disappointed in you.


    Any other comments? [top]

    Scheduling the exam can be one of the harder parts of the test, if
         you have committee members who travel often.  It's not always easy
         getting five professors into the same room for three straight hours!
         If you think you'll run into such problems, schedule your exam well
         in advance.

         I brought in a bike canteen, since I get thirsty when talking too
         much.  I also had a calculator, which I was allowed to use.  While
         I brought in my binder of notes, I never referred to them during
         the exam.  You may be allowed to look at written references during
         the test, but ask permission first, and do not do so very often.
         You normally don't have such material available during interviews
         or when answering questions after presentations, so don't rely on
         them during Orals.

         The notes on mental attitude and on avoiding burnout from the DWE
         guide also apply to the Orals.

    Thanks for the tips.  But what was that pay raise you mentioned earlier? [top]

    After you pass both Orals and proposal, your stipend increases by  $500 per semester.  It's not much, but any little bit helps.


    Good Luck !


    University of North Carolina at Chapel Hill
    Computer Science, Sitterson Hall
    Chapel Hill, NC 27599-3175 USA
    Server Manager: webmaster@cs.unc.edu
    Content Manager: cssa@cs.unc.edu
    Last modified:
    .