Writing a PhD Thesis


Writing a Ph.D. thesis is a lengthy, complicated, and sometimes difficult to understand enterprise I am, therefore, trying to gather sources that will help people to see the light at the endof the tunnel. I will post useful material here whenever I find it...

Currently this page contains:

  • Notes for the talk: Useful things to know about Ph.D. Thesis Research
    by H. T. Kung (CMU 1987) as text (below), as text-file


by H. T. KUNG

(Prepared for "What is Research" Immigration Course, Computer Science Department, Carnegie Mellon University, 14 October 1987) 

   Presentation Outline

   1. Introduction

   2. Why Ph.D. thesis could be really difficult for a student

   3. Types of Ph.D. theses (from Allen Newell)--not a topic of this talk

   4. Growth  of  a star (the transformation process that some students go

      through to become a mature researcher)--which stage are you in?

   5. Stages of Ph.D. thesis research

   6. Methods to  get into the depth of a topic (or how to come up with

      good ideas)

   7. Breaking myths

   8. Pitfalls to avoid (easy ones to avoid listed first)

   9. Some other general advice

  10. All the effort is worth it (believe it or not)



   1. Introduction

         - Ph.D. thesis is treated very seriously at leading universities.

              * Expectation is high.

                   - Ph.D.  thesis represents a substantial work.  Faculty

                     often tell other  people  that  "We  have  a  student

                     working  on  this  area for his or her Ph.D. thesis."

                     Amazingly  enough,  this  is  usually  sufficient  to

                     convince  people that the problem is somehow going to

                     be solved.

              * Ph.D. thesis research is a task to ensure that the student

                can   later   take   on  independent,  long-term  research

                commitments.  (If a Ph.D.  student does not intend to be a

                researcher,  the Ph.D. thesis work is not worth the effort

                in general at least at CMU.)

              * Through  the  Ph.D.  thesis   process   the   student   is

                transformed into a professional researcher.

              * Faculty are judged by the theses of their Ph.D. students.

              * High  standard  Ph.D.  thesis  is probably one of the most

                important  factors  that  contribute  to  the  success  of

                graduate education at leading American universities.

              * Ph.D.  thesis  is  probably  the  only  real challenge for

                getting a Ph.D.  degree.

                   - Ph.D. qualifier is seldom  a  problem  for  motivated


         - Ph.D.  thesis  research  is probably more mechanical than a new

           graduate student would think. (Of course the process  is  still

           too complex to be automated.)

              * Knowing  this  mechanism can be more important than thesis

                results themselves.

              * Some information presented here may be  relevant  to  your

                whole  research career, i.e., it is not just for the Ph.D.

                thesis per se.

         - This talk consists of pragmatic advice.

              * The talk is based on my  personal  experience  (i.e.,  not

                based on any serious research)

                   - I  happen  to have research experience in both theory

                     and system areas.  We will compare thesis research in

                     these two areas.

              * This  is  a  common sense talk and will have down to earth


                   - "I wish someone told me this before."

   2. Why Ph.D. thesis could be really difficult for a student

         - Most likely this is your first, major research experience.

              * A big challenge for most students

         - No simple recipe

              * Different talents

              * Different kinds of theses

              * Different approaches

         - The work is judged by thesis committee (mostly advisor).   This

           produces anxiety.

              * Unlike   other   research  you  will  do,  the  evaluation

                mechanism for thesis research is very unique.

              * No clear contract

              * No clear standard (we only know it is high)

              * Recall the Stanford murder case (the former student  said,

                after  he had finished--he did finish something-- his jail

                term,  that  he  might  do  it  again  under   a   similar


   3. Types of Ph.D. theses (from Allen Newell)--not a topic of this talk

         - Opens up new area

         - Provides unifying framework

         - Resolves long-standing question

         - Thoroughly explores an area

         - Contradicts existing knowledge

         - Experimentally validates theory

         - Produces an ambitious system

         - Provides empirical data

         - Derives superior algorithms

         - Develops new methodology

         - Develops a new tool

         - Produces a negative result

   4. Growth  of  a star (the transformation process that some students go

      through to become a mature researcher)--which stage are you in?

         - Knowing everything stage

              * Student: "I have  designed  a  supercomputer  even  before

                graduate school."

              * Faculty: speechless

         - Totally beaten up stage

              * Student: speechless

              * Faculty:    smiling   at   the   student's   progress   so

                communication is possible now.

         - Confidence buildup stage

              * Student: "I am not stupid after all." (student thinks)

              * Faculty: "Uh oh, she is ready to argue."  (faculty think)

         - Calling the shot stage

              * Faculty:  "I   am   going   to   design   an   n-processor


              * Student: "You are crazy, because ..."

   5. Stages of Ph.D. thesis research

         a. Selection of area--not a topic of this talk

         b. Selection of advisor--not a topic of this talk

         c. Becoming a researcher in the area

               - Building up general knowledge, experience, and confidence

               - Knowing issues and important questions in the area

               - Capturing research opportunities

                    * Don't  let  any idea or question go by without first

                      giving it careful thought.

                         - Be alert and diligent.

                    * Pay attention to new technologies

                         - Examples

                              * VLSI, networking, and new  chips  such  as

                                the   Weitek  floating-point  chips  three

                                years ago which in  some  sense  gave  the

                                initial motivation for the Warp project

               - Some useful things to do (from Dave Gifford, MIT)

                    * Read recent proceedings of the best conferences, and

                      ask more senior people what were  the  best  papers.

                      Try to figure out what makes a great paper (and thus

                      what makes great research).

                    * Keep a notebook that contains your  research  notes.

                      Put  all of your empirical data and initial ideas in

                      the notebook.  Make notes on a paper as you read  it

                      and  think  about  the assumptions of the author and

                      the importance of the results.

                    * Follow references from one paper  to  another  until

                      you  know  an  area  extremely well.  Don't count on

                      your advisor to hand you all of the relevant  papers

                      out of his file drawer.  He doesn't have them all!

         d. Thesis proposal

               - It  is the most crucial stage in the sense that the basic

                 concept is worked out here.

                    * To get important results you need to  ask  important


                    * This is the time you need your advisor most.

                    * Problems  in  later stages are usually rooted from a

                      weak thesis proposal.

               - Purpose

                    * A research plan

                         - A serious attempt to get  an  overview  of  the

                           whole research course

                         - Not really a contract

                              * Need  some  flexibility  because  research

                                always has uncertainty.

                    * Forming the committee

                         - Varies a lot

                         - Choose people for your  thesis  committee  that

                           can  help  with needed expertise.  For example,

                           it is useful to have a relevant theory  faculty

                           member on a systems committee and vice-versa.

                         - However,  there  is usually no need to optimize

                           too much on  the  selection  of  the  committee

                           members--advisor still plays the most important


                         - However it can be very important, when

                              * you have a "questionable" advisor, or

                              * you have an interdisciplinary topic.

                    * A review

                         - If there is any serious doubt,  it  had  better

                           show up now.

                         - Proposal  could  sometimes  be viewed as just a

                           forcing function for  taking  care  of  certain


               - Some  of the difficult questions always asked in a thesis


                    * What is your approach and what is new?

                    * What is your secret weapon?  (Herbert Simon)

                    * How do you measure your own progress?

                    * What are the success or completion criteria?

                    * How will the expected results  change  the-state-of-


               - The  grand  challenge for a thesis proposal is to come up

                 with an approach or an experiment.

                    * It is easy to identify a general problem  area,  but

                      setting  up  an approach and designing an experiment

                      can be difficult.

                         - Need ideas

                              * Just need one good idea, really

                              * Unfortunately,  there  is  no  magic  here

                                (however  see  some hints below).  This is

                                the hard part of any research project  for

                                everyone (not just for students).

                    * Need independent thinking

                         - You should be good enough to start arguing with

                           your advisor on technical issues  and  research


                    * Need  to  elaborate  on focus, approach, experiment,

                      and potential impact

                         - For theory research you may  propose  some  new

                           models of computation.

                              * Examples:  area-time  complexity (new VLSI

                                model in theory), parallel algorithms (new

                                cost models)

                         - For  system research you may design experiments

                           and argue their relevance.

                              * Examples:   multiprocessor   architecture,

                                compiler for a parallel machine

               - Useful things to know when preparing a thesis proposal

                    * Be  honest.    There  is  no need to exaggerate your

                      claims!  If you point out  the  weaknesses  in  your

                      approach you will disarm your critics.

                    * Pick  a  project that is manageable so you can do an

                      excellent job - things are always harder  than  they

                      seem.   It is far better to do an outstanding job on

                      a moderate size project than a  moderate  job  on  a

                      large project.

                    * Include  a  tentative  thesis outline and a month by

                      month schedule in your thesis proposal.

                         - This may be difficult to do but  it  is  better

                           than no plan at all.

                         - This will also help gauge the total size of the

                           work you are committing yourself to do.

         e. Producing results

               - Lots of work--what else do you expect?

                    * System--be inside an active project  without  losing

                      sight of thesis

                         - Need  to  be  a  worker as well as a conceptual


                         - Your work depends on other  people's  work  and

                           vice versa

                              * Opportunity to see real problems

                              * Getting     good     support,    including

                                encouragement and demand, from the group

                                   - It seems that this arrangement really

                                     works in all cases.

                         - Be   quick,   because  you  don't  want  to  be

                           overtaken by the environment (this  is  one  of

                           the pitfalls to avoid, as described below)

                    * Theory--be lucky!

                         - Be flexible

                              * It is hard to insist that you will prove a

                                theorem before you go to sleep.

                         - Be  quick,  because  theoretical  results   are

                           totally  portable  and  so  competition  can be


               - Keep the committee  informed  (at  least  those  "trouble


                    * You can get real help sometimes.

                    * Committee members are obliged to talk to you.

                         - Sometimes  finding  a  qualified  person beyond

                           your  advisor  to  discuss  your  work  can  be


                    * Don't  want  surprises  in  the  later  stage of the


               - Ways to finish a thesis

                    * Incremental and adaptive approach

                         - A sequence of incremental results

                    * Big-bang  approach  (this  is  not  recommended   in


                         - One big theorem

                         - A big piece of software or hardware

         f. Writing

               - Why  some students find that Ph.D. thesis writing is very


                    * First major document

                    * Writing  is  time-consuming--part   of   the   .9999

                      perspiration (Satya)

                         - Think  how many good sentences you can write in

                           an hour.

                         - Fighting with fonts, figures, references, etc.?

                              * Please don't be too picky.

                    * When results are not totally solid, writing  can  be

                      really difficult even for an experienced writer (now

                      you know another reason why proposal writing is  not


                         - Can't  say  too  much and don't want to say any


                         - Writing about  flaky  results  can  be  a  real


                              * In  this  case  you  should  improve  your

                                results first.

                    * Writing has to  do  with  presentation  rather  than

                      finding  new  results.    So  writing  may not be as


               - However, thesis writing is useful in the  sense  that  it

                 helps  reveal  possible  problem  areas  and provides new


                    * Help get a large picture on what you really have.

                    * Help organize the concepts

                    * Completeness is forced.

                         - You must take care of things that you have been


                              * For  example,  you  need  to do comparison

                                with other results

                    * Correctness of the results is checked.

                         - You had better  have  the  proof  now  for  any

                           plausible   "theorem"   that   you   have  been


                    * New insights on how things really work

                         - New ways of looking at you results

               - Recommendations

                    * Get some practice--write some papers before thesis

                         - Write some joint papers with  people  who  have

                           substantial writing experience

                    * Need to know the theme of the thesis very well

                         - Outline first

                         - Write the conclusion first (try it at least)

                         - Start writing chapters which are more settled.

                         - Write the introduction last

                         - Iterative process

                    * Make the writing as precise as possible, so that you

                      know exactly what you are talking about.  This  will

                      save lots of rewriting.

                         - Precise   writing   usually  also  yields  good


         g. Getting final comments from the committee

               - Not too early or too late

                    * Getting some committee members  to  read  can  be  a


                         - They are busy people.  You want to give them an

                           "optimal" version to make comments.

               - How much to ask for comments varies a lot

               - Should not have any surprises now.

                    * You had better know what you have been doing by now.

                    * However, if there is any problem, it had better show

                      up now.

         h. Defense

               - Mostly  a  formality and a happy occasion (should be like


                    * You know that your results are  good  and  you  will

                      present them well.

                         - You  should  know  the answer to the question -

                           "What  are  the  three  main  ideas   in   your

                           thesis?".    You  should be able to rattle them

                           off and relate them to previous work.

                    * Getting a date set can be more  difficult  than  you


                         - Committee  members  do  not necessarily stay at

                           CMU as long as you do!

                         - Weekend defense is not really desirable.

                              * May be difficult to get audience.

               - However defense is still very important:

                    * Opportunity for final improvements for the thesis

                    * Formal presentation to the community

                         - Many people form their opinion of your n-years'

                           work from this presentation

                    * Presentation   material   can  be  used  for  future


                         - Used in recruiting presentations  if  you  have

                           not settled on a job yet

                    * Psychologically important

                         - Once in a life time occasion--you will remember

                           it always.

                    * Don't want to blow it.

                         - Absolutely no surprises

         i. After defense

               - Usually there is still some minor work to be done for the

                 thesis (too bad)

                    * Defense was moved early for various reasons

                    * New comments from defense

                    * Did  not  have  time  or  did not want to polish the

                      thesis before defense

               - Publication

                    * Articles, books (or give the thesis to your parents)

                    * Very important to publish the results in journals

                         - This is the only reliable way to  archive  your

                           results.    (You  don't want to lose them after

                           all these efforts, do you?)

                         - Publication is important for academic career.

                         - May break the thesis up  in  several  articles.

                           When  appropriate, some articles may have joint

                           authors such as your advisor.

                         - Do it right away before you get on to the  next


                    * Books can be good too.

               - Follow-on work

                    * Keep mining the thesis--why not?

               - Finally you are free!

   6. "Methods"  to  get into the depth of a topic (or how to come up with

      good ideas)

         - No magic, but we will still try ....

         - How to develop initial ideas

              * Study other work and do comparison

                   - What are similar issues and solutions?

              * Look at examples

                   - Generalization and abstraction

              * Make hypothesis and validate it formally  or  informally--

                keep trying

                   - You will discover issues at least.

              * Do modeling and abstracting

                   - Get the essence

              * Just do something--be active

                   - Implementation--details reveal issues

                        * Join a project to do some real work!

                        * Handle a smaller case

                        * Implement   a  throw-away  simulator,  language,

                          design, etc.

                   - Start proving "theorems", even if they are  known  to

                     be difficult.

                        * Quick way to understand issues

              * Work  with  good, experienced researchers (don't forget to

                use your advisor!)

                   - They might have deep insights on similar problems.

                   - They  can  help  calibrate  the  difficulty  of   the


                   - You  learn  the subject matter from them more quickly

                     and directly.

                   - You learn their techniques

                        * Every successful researcher has his or  her  own

                          bag of "tools":

                             - Calculation,      synthesis,      analysis,


                   - If they also get stuck once in a while, you know that

                     you are not that bad after all.

         - How to develop existing ideas further

              * Exploring problem and solution spaces

                   - Enumerate   parameters  individually  (and  do  quick


                        * To see where your current ideas sit in the space

                   - Correlate results

                   - Generalize ideas and results to other points  in  the


                   - Produce phenomena and explain them (Herb Simon)

              * Brainstorming your ideas with others

              * Presenting your ideas in papers or/and seminars

                   - Ideas    will    be   checked   out   carefully   and

                     systematically (see above on thesis writing)

              * Example steps that can be used to get some  depth  from  a

                simple result such as a speed-up curve

                   - Explain the curve

                   - Look at the problem and solutions spaces

                   - Do some comparisons

                   - Change the assumptions

                        * How stable is the result?

                        * How   will   results  vary  or  correlate  under

                          different assumptions?

                   - Derive some general principle

                        * Similar curves for other situations?

         - General comments

              * Thinking is the key

                   - Thinking is more important than reading

                        * Books are not always right.

                             - Note that  in  the  system  area  with  few

                               exceptions  people who build systems do not

                               have  time  nor  need  to  write  up  their

                               experience--it  is  too  bad  but  it  is a


                   - Be alert on all sorts of opportunities

                   - Do the thinking right away while you have it.

                        * Ideas and interest may be lost more quickly than

                          you like to believe

              * Talking to people

                   - Don't  over  do  it  (you  still  need to do the work


   7. Breaking myths

         a. "Advisor is a stronger researcher than you."

               - It is true  that  advisor  is  experienced,  wise,  smart

                 (maybe), and knowledgeable in general.  Advisor also sees

                 a bigger picture, and has contacts in the area.

               - However, advisor is not always right.

                    * Advisor is not as focussed as you.

                    * Advisor does not have more time or energy  than  you


                    * Advisor is not as innovative in general.

                         - They know too much.

                         - They are more conservative.

                              * They know too many horror stories.

                         - Aging does not help.

                    * Advisor's  knowledge may be obsolete (don't say this

                      in front of him or her!).

               - You must believe that you can do better than advisor  for

                 some research areas.

         b. "System theses take longer than theory theses."

               - The  most  difficult  part of a thesis is to come up with

                 some good, new ideas.   The  difficulty  in  getting  new

                 ideas is the same for theory or system research.

                    * Theory  thesis  is in general not about solving open


                         - Actually good theoreticians always work on  new

                           problems,  models  and methods so that they can

                           solve the problems that are "solvable"  in  the

                           first place.

                              * Greatest contributions are ground breaking

                                ones, such as new models.

                              * New approaches give new  insights  to  old

                                problems.    This is the way open problems

                                usually get solved (e.g.,  the  four-color


                    * For  systems  theses  it is important that the major

                      ideas  in  the  thesis  are   independent   of   the

                      implementation--the  goal  is to have the ideas live

                      on in other systems as well.  A good systems  thesis

                      usually  has  a  new  algorithm or new method at its


                    * Few theory students  who  finish  really  early  are

                      likely  those  who  have  prior research experience.

                      (Recall that theory results are highly portable!)

                    * Incompetent theory students are more noticeable than

                      weak  system students.  So we don't often see theory

                      students who drag on for a long time.

               - There are some differences in systems and theory research

                 however,  but they should not have too much impact on the

                 thesis research time.

                    * System needs implementation,  whereas  theory  needs

                      more background study.

                    * Theory   research   is  self-sufficient  and  system

                      implementation may depend  on  other  people's  work

                      (you should not get into a situation where you don't

                      have control).

         c. "Ph.D. thesis research follows some standard guidelines."

               - Yes, a Ph.D. this must represent a substantial result  in

                 a very high standard.

               - But  there  are  many  ways to leave a mark in a research

                 area.  As long as you have come up with some  good  ideas

                 and  pushed  the  frontier  of  knowledge,  you  will  be

                 surprised sometimes how flexible your committee could  be

                 in  terms  of  the research approach, acceptable results,

                 and thesis presentation.

               - There is a small percentage of Ph.D. theses completed  in

                 unusual manner.  Don't give up too early if you belong to

                 this class.  Try it or you will never know.

   8. Pitfalls to avoid (easy ones to avoid listed first)

         a. The goal is too big to reach.

               - Theory

                    * Proving P /= NP

                    * Proving P = NP is even  worse  (likely  this  thesis

                      will never finish!).

                    * Deciding  whether  P = or /= NP is best of the three

                      (i.e., be flexible)

               - System

                    * The initial effort is  so  large  that  real  issues

                      never get a chance to be looked at.

                    * It is important to size the project and evaluate the

                      total effort carefully based on past experiences.

         b. Ideas cannot stand without  an  implementation  that  competes

            with commercial products.

               - Chess  machine  implementation is OK, because there is no

                 commercial competitor.

               - In this sense,  Warp  hardware  is  more  difficult  than


               - Floating-point  designs  that  require a high-performance

                 chip implementation to  validate  the  concept  would  be


               - Never need to implement another vector processor!

         c. The thesis area is overtaken by technology and environment

               - Technology advances have solved the thesis problem.

                    * A  clever  operating  system using no more than 128K

                      memory is not very interesting today.

               - Advisor (or student sometimes) has  changed  his  or  her


               - Other   new   projects   have   better   approaches   and


               - Other  people  have  published  similar   and/or   better


               - Advisor  has  a  better  job  elsewhere or the project is


               - Lesson:  You should always do your thesis as  quickly  as


         d. Totally isolated work

               - No  encouragement  and  support--no  one cares about your


                    * Can't even find an advisor sometimes

                    * Doing a thesis away from CMU is really difficult.

               - System research

                    * Lone ranger approach is almost suicidal.

                         - No software, systems  and  application  support

                           for evaluation

                         - Very  difficult  to  do  anything  real without

                           feedback from a community

               - Theory research

                    * At least global networking is needed.

         e. Not knowing when to stop

               - Thesis is not the last research you will do.

               - You can do the same  research  after  your  Ph.D.  thesis

                 (while making more money).

               - Learn  to  make  reasonable  assumptions  to restrict the


         f. Unhealthy competition between student and advisor

               - This is more likely to happen in the theory area.

               - The potential  is  always  there  (especially  for  smart

                 professors  with  lots of ego).  In general if both sides

                 try to be fair, things can always be worked out.

         g. Lots of numbers and hacking but no fundamental principles

               - System research has to have more than implementation.

               - Implementation for a thesis research is interesting  only

                 if it can be used to validate some theory.

               - This problem should be fixed as early as possible.

         h. Things  dragged  on--wonderful  general ideas in the beginning

            that never get  developed  into  a  coherent  approach  (i.e.,

            heading to a black hole--there is no output)

               - Wrong  areas  for  the  student (and perhaps the advisor)

                 with respect to ability and interest

               - Nightmare case--it does no good to anyone.

   9. Some other general advice

         - Stay away from areas that have been thoroughly  mined  by  your


              * Keep yourself at the very front of a research area so that

                you have a better chance to hit something big or at  least


              * After all in research what matters is the work that pushes

                us into new territories.

              * Make use new advances in other areas

         - Don't avoid thinking

              * Thinking is hard but there is no substitute for it.

         - Psych yourself up for this unique experience of doing  a  Ph.D.


              * Make  yourself  believe you are solving the most important

                problem in the world

              * Remember what worked for you before

                   - If you work best when you are competing with  others,

                     then create some confrontation.

              * Must be very alert about issues and opportunities

              * Thesis  process is sort of artificial (almost a torture in

                some way)

                   - The thesis is judged  by  a  committee  (mainly  your


                        * More subjective than exams

                   - Probably  one of the most humiliating experiences for

                     people of this age (advisors should all remember this

                     and be considerate.)

                   - The  process  is  not  a  typical research style--you

                     don't do anything similar to it  again  even  if  you

                     will be doing research after the degree.

              * The  thesis  process  can  be  long  and  treacherous. (Be

                prepared for it.)

                   - You don't want depression.

              * There are quite a few very competent people  who  just  do

                not want to go through this.

         - Use forcing functions well to speed up the thesis process

              * Competing with someone else

              * Family pressure

              * Financial pressure

              * A job is waiting

              * Advisor is leaving or project is over

              * Equipment is retiring

         - Never throw away advisor's comments

              * Cox-Denning case

         - Keep  good  relationship  with  your  advisor  (even  after you


              * Good thing to do--no exception almost

              * Relationship is unique.

                   - Advisor usually has lots of influence on you in  this

                     very  important  stage  of  your  life.  Advisor also

                     appreciates the good research you did with  him,  and

                     is in general interested in your well-being.

              * Advisor may be your mentor for your entire career.

  10. All the effort is worth it (believe it or not)

         - Experience  from  Ph.D.  thesis  research  is unique.  You have

           learned how to do research.  Future research  is  going  to  be

           more  interesting  because  you  will know how to do it, so you

           will have more freedom and fun.

         - Almost all leaders in research have this experience.  You  will

           have  confidence  in  your  research ability.  You will look at

           things differently than people who  did  not  go  through  this

           process.   It is very clear that Ph.D. thesis research is still

           the best way we know of in developing powerful researchers.

         - In summary, it is the best investment for becoming a successful