Last changed 6 July 2003 ............... Length about 5,000 words (29,000 bytes).
(Document started on 25 Apr 2003.) This is a WWW document maintained by Steve Draper, installed at You may copy it. How to refer to it.

Web site logical path: [] [~steve] [resources] [postgrads] [this page]

Creating the skeleton for a thesis

By Steve Draper,   Department of Psychology,   University of Glasgow.

This page is about the need, in writing a PhD thesis, for working on the logical skeleton: the argument, the structure around which it all should fit. It presents my opinions, and not everyone (not all supervisors) will agree with all of it.

An exercise on this.


Why does this matter? What's the problem?

I've seen some shocking PhD theses recently, where it just looked as if the student was still thinking like a school pupil: just do all the bits teacher requires, without ever thinking about whether it means anything. So there would be a literature review clearly written in the first year, and no longer relevant; studies described, but not made to add up to any overall argument. A PhD isn't about turning up to a factory on time every day for the compulsory period, but about producing some research that is useful to other scholars. A thesis needs a thesis.

This argument is also made by Chad Perry in two paragraph form here, based (it claims) on a review of 139 examiners' reports. (His whole document starts here.)

Overview of the solution proposed here

A PhD thesis, like most organisms, needs a skeleton. In fact there are four aspects to this, which a doctoral student in the process of writing up should cover:
  1. The main message: the conclusions you present: what the essential elements are, and the relationships they have to the final message. That is, it should not be just a description of what you did, but lead to some definite conclusion(s). This is your "contribution to knowledge".
  2. A reasonable argument. That is not just asserting your conclusion(s), but showing a reasonable discussion, an argument, reasons for accepting your conclusion. An argument consists of several parts, where some parts support others.
  3. It is not just an argument, which is a structure in the mind, but also a structure for the writing. That is, the structure of the dissertation (what the parts are and how they relate to each other) should be strongly related to the logical structure of your argument.
  4. Put in lots of "glue" statements to show readers how the parts relate; and to support readers who just dip in, or jump around. Not just a tree structure of subsections to organise the writer's activity, but a set of signals that readers can easily use to organise theirs: signs to structure the reading.

I call all this a "skeleton" because, while it spans the whole of your thesis, it can be separated from it; because without one, the whole thing is just a big, amorphous, floppy mess that doesn't work; because it structures the material, your work in writing it, and the reader's work in understanding it, and in finding the bits they want at any given moment; and because you can think about it and work on it separately from working on the content parts (which is just as well, because you probably can't really hold the whole thing in your head at once).

The skeleton of a thesis

A PhD by research produces, and is examined by, a dissertation. It is commonly called a "thesis": i.e. it is supposed to present an argument, and one that advances knowledge. This should report on work according to the conventions and standards of its discipline (whether mathematical proofs, engineering design and construction, historical research on materials, reporting experimental work, or whatever). But it should also present an argument. One of the hardest bits of work for most of us, who haven't written anything so large before our PhD, is creating the organisation for this.

You can equally see this from the viewpoint of the other scholars and researchers who might read your thesis. What they want is to take away your conclusions, plus your reasons for reaching them, and some estimates (their estimates, but heavily influenced by you) of how reliable and sound those conclusions probably are. So in writing, your job is to support this aim of your readers. Thus the structure of the dissertation should be strongly related to the structure of your argument. Each part is there (only) to contribute support to the argument.

To achieve this, then, you have to decide what your message, your thesis, your main conclusions, your contribution is; then what the parts of your thesis are going to be (to support this) and the order they should go in; and then signal this to your readers. Because you will find this difficult (it's too big to fit in your head easily) you also need to build a scaffolding to keep it under control while you write it. These are the four aspects listed in the introduction. The next four sections suggest how you might do each.

Deciding on your thesis message

The first thing is to decide what your most important message is. Unless you discovered nothing in your research, this will not be identical to what you expected when you began. A thesis is not a history of what you did, nor a justification of what you planned to do when you got funding. It is what the best rational assessment is now, with hindsight, of what can be concluded from your research.

It may be that you cannot see what your conclusions are until you have written your thesis (or so it feels). If that is true, then write a full draft of your work (e.g. writing up the experiments or whatever), and then see what you want to put in the discussion of each study. At that point, you will for the first time be able to start deciding on your overall story. How hard you find it will vary greatly from person to person. Some will have a had a good plan from the start, and though it will need updating, it may not need radical change. Others will have been reviewing it regularly, so there is not a lot more change needed when it is time to write up. However for others, some creative effort is needed. One reason is if a study came out completely differently from what was expected. The skillful person will swallow their shock, and turn it into something important. After all, if it surprised you, perhaps that is the important finding that others need to know about. Another reason is just that you never thought about where your research was going, and now it's time to put it together for the first time.

If you are really stuck about what your story essentially is, try this: I often do it when writing a paper, though I don't remember doing it for my thesis. What I do, is try to summarise it for someone orally e.g. over coffee. It's obvious that they will lose patience within a minute or two, so I have to pick out what my absolutely most important point is, and state it simply in a way someone can understand. Once I'm clear about that, I can go off and write the long version, organised around what I've discovered is my main message.

Any fool can do a brain dump of all the details they know (pick up any computer manual): the hard bit is to organise it around a single consistent theme/message/story. This takes a bit of work for a paper. For a thesis, it's harder again because it's longer; but once you've got your story straight, the rest is just work, filling in chunks, and writing "glue" to keep the (inattentive) reader straight about what each bit is doing with respect to the whole.

Deciding on your argument structure

Once you know what your message is, you need to decide on how to express it, and convince the readers. Logically, there are two types of decision here: selecting the content, and deciding the order of the pieces. The first job is selecting the content and the relationship of each element to the conclusions (your message) -- this IS the argument structure (the parts plus their relationships to each other).

Selecting the content is about deciding what pieces of evidence and argument support your message, and leaving out other things you know and perhaps did, that aren't now relevant to the argument. If research is a journey, you probably wandered around quite a bit looking for the best path. Now you know it, you only need to tell the reader about the best path, not the wandering around. A thesis is in some ways a kit of parts, and it's up to you to decide how to fit them together, what bits you are going to reinforce, what just sketch a connection for. So for example, you might slim down the literature review to those things still relevant, not everything you have read. This may still include mention of irrelevant things that reader might want to know about e.g. "most work in this field has been based on the kind of study introduced by Bloggs. However it turns out that a quite different approach produces better results...."

More generally, what you are doing in constructing your argument is demonstrating "critical thought" (which is often stated as a requirement in the paperwork for PhD exams). It is true that a defining characteristic of science is that things are published ("facts") which others then interpret in different ways from the original author. This is crucial to the way work cumulates (builds on what went before): things do not depend on what the first observer thought. And certainly one valuable thing you do in your thesis is report facts or observations, or what you did. However a PhD that only reported facts, what was done, measurements taken, would be unlikely to pass. You are expected to discuss what may and may not reasonably be concluded from your work, and the arguments for and against your conclusions. Although what you present is in effect in the form of an argument, you will not mainly be judged on whether or not you persuade others (as you might be in marketing, and politics). But you are expected to help others decide how much to believe your conclusions by providing them with your considered judgement and the important issues and evidence for and against. In discussing not only evidence for, but evidence against, alternative theories, and obvious objections, you are helping the reader form their own judgement, and simultaneously showing you have a good and thoughtful grasp of the issues involved in your work. Even if some people don't value your conclusions, they will value this discussion.

In deciding the elements to select and your argument structure, you are deciding this. Note that to help the reader, you need constantly to consider what questions and issues may be in their minds. Not mentioning obvious objections, or the rival explanation to yours will annoy them, and they will justifiably think poorly of you. So in deciding your argument structure, don't only consider favourable elements, but also things you must discuss because many readers will be thinking them anyway.

This exercise is meant to help and support postgrads in this stage of getting their argument structure clear.

Creating a scaffolding structure for your writing

Because you probably won't write your thesis in order from start to finish, nor in one session, you need to make a plan, and record this somehow to refer to while writing goes on. A thesis is about 80,000 words or half an average book, while a typical journal paper is 6-10 thousand words. So the thesis is about ten times bigger than what you are used to writing, even if you have published some papers.

The plan isn't just a list of the order of the chapters. In fact what order you put them in is only the second aspect, part of choosing how to help the reader; while the first aspect of the plan is making your selection of parts and getting clear how they relate to each other. So the most important aspect of the plan is a statement of how each part relates to other parts. You need this to get it clear in your head, and later to tell the reader how it fits together.

So, for every piece of your thesis (every chapter, section, subsection; every table and figure) write down what this element is meant to do in the thesis, what it relies on and/or refers to, what later bits rely on it. Write this out, put it in the text at (say) the start of each piece, put it in italics or something to mark it as a message to you not to the reader. Then it acts as a placeholder for bits not yet written; as a plan for each bit as you write it; as a specification so that when reading a draft you can test each section against its own specification. For instance: "The main business of this chapter is to describe and present expt.2. It relies on the motivation having been already discussed at length: partly in the introduction (overall motivation for the topic), partly in chapter 4 in the results from the preliminary experiment. The method is also only stated here, but justified in the discussion of methods in part of ch.3 (literature review). The various possible conclusions are also given here, many of them only here. However conclusion D motivates the third study in the following chapter, and G is discussed again in the final chapter."

I suggest putting it in the text in italics, and deleting it (like taking down scaffolding when a building is finished) only when you finally submit the thesis. You may copy some of the scaffolding into the main text as "glue".

Signposting the structure for readers: "glue"

The other part of the planning is deciding the order in which to put things. An argument is not naturally a line or sequence but a branching structure, just as a stool may be supported by three or four legs, not one. There isn't a single logical order for the legs, yet in writing you have to choose a single order, so you have to invent one. You also need to tell the reader what this order is, and how the parts relate. If there is a convention and you follow it (e.g. literature review first) then you can worry less about telling the reader because they will assume it; while if you do something they aren't expecting then you must warn them clearly, and preferably say why you are doing it. However do NOT stick to a convention that doesn't fit your particular argument: do something sensible and warn the reader. Similarly, if you have several experiments (or bits of software) to describe, you may think it obvious that you should present them in the order your did them. But that doesn't follow: perhaps the order required for the logic of the argument is different. For example, you might do an experiment that shows the result your whole thesis is based on, but then did supplementary experiments to check that other factors don't have an effect, or to prove your measures are valid, etc. In that case, it might make a better argument to present those calibration studies first, then the "first" experiment, whose results are now much more believable.

You also have to make this structure clear to the reader. This is for two reasons: firstly to communicate the structure of your overall argument, which is actually the whole value of your thesis; and secondly because few people read even a short paper much less a thesis from start to finish: instead most of us jump about using its structure (see for example Hartley; 1999). Hartley implies that a fixed conventional structure is best, and once everyone expects it, it speeds up reading and also means authors don't have to spend time wondering what structure to choose. Against that, is that not all papers and arguments are essentially similar. Furthermore, there may well be more than one kind of reader e.g. someone who wants your results, not how you got them; someone else who wants to use your methods, not necessarily your results; etc. On the other hand again, his standard headings may represent expectations and questions in many readers' minds, so that answering all of them explicitly (even if only by saying why this heading doesn't apply here) may be a useful system for authors.

In any case, to support readers in their non-linear reading, they need signposts; and you have to write the thesis so that (like web pages) when readers jump into the middle, they can (unlike in a novel) make sense of it. There are three general kinds of technique for supporting this jumping about:

  1. Organising the thesis into parts: dividing it into chapters, sections and perhaps subsections, and giving these names and numbers.
  2. Things such as contents pages to tell readers where to jump to.
  3. Things such as titles and opening paragraphs that tell a reader who has jumped in as quickly as possible whether they are in the right place for their purposes.

The most common technique for telling readers where to jump is the contents page. Notice the phrase "contents page". A lot of books and most PhD theses allow the "contents" to sprawl over many pages. This is much, much less useful than a contents that can all be seen at the same time with no page turning, because the whole point is to offer an overview or map, and many selections are made not because a chapter title is obviously the right one for that reader's purpose, but because it is the least bad match out of the whole set of chapter titles. Less conventional, but very important, for theses that include several experiments or other empirical studies is to have a list or table of these so that the reader can immediately see how many there are, what they are called in this thesis (e.g. expt. 1,2,3 or "the pilot" "the main study" "the followup study"), and where they are described (chapters and page numbers). Almost all readers will want to get this clear, and will refer to it regularly if it is provided. Similarly if the thesis describes more than one piece of software, these should be listed clearly with names, pointers, page numbers. In addition, there is nothing wrong with telling the reader how they might like to read the thesis. I included a "skimmer's guide" in mine, suggesting that readers who wanted one kind of approach should read chapters 1-5 and 9; while those interested in my other approach would want to read 1, and 6-9.

The main techniques for helping the reader decide if they have arrived at the right place are titles, and "glue" sentences at the start and/or the end of every bit. For example "This chapter described my second and third experiments, concluded X and Y, and discussed the degree to which these conclusions can be relied on. X is used in the next chapter to motivate the fourth study, and Y is important in the overall conclusions in chapter 9."

You might use your scaffolding/specification statements (that were first written to help you) for the glue statements put in to help the reader: certainly they would be much better than nothing. However don't make the mistake that what is optimal for you is optimal for the reader. There's a lot of stuff that is good for you to write out in tedious detail in order to get yourself really clear about it, but which you may decide not to bother the reader with: you should still write it for your own sake, but that doesn't mean you should make others read it. For example, you might note to yourself that fig.8 needs to tell the story at a glance, that your biggest result is right there, and that once you've got a powerful visual format for it you will use the identical format in fig.14 because everyone will want to compare the two results; but you will probably decide not to say this to the reader but let it speak for itself: except perhaps to say as part of the legend for fig.14 "see also fig.8 p.143".

Writing for yourself, rewriting for the reader

Many supervisors insist doctoral students write up chapters as they go along. In some ways, this is very bad advice. Some of the worst theses I have had to examine have obviously been done like this. The result is, you have early chapters that obviously are no longer relevant to the argument that has now appeared.

What is right about the advice is that you need all the practice writing you can get; and, independently of that, most of us find that writing is the best, sometimes the only, way of working out what you really think. And that is besides the big value of publishing papers as you go. So you definitely do want to be writing a lot right from the start. And the worse you are at it, the longer it takes, the more you need to practise. Writing is painful; avoiding it is career suicide.

However if you think you can write a chapter one year, and use it next year, you are daft. That is like thinking that every mile you run in training is a mile you don't have to run when the race comes. Obvious nonsense. Training is vital; but it isn't gaining exemption from the race, it just equips you to do it faster and better when the time comes.

Actually the truth is slightly better than that. Some of the time consuming things in a thesis only have to be done once right, and then copied / re-used: for instance typing up and checking references in the bibliography (though don't forget to delete ones you don't use after all); designing tables of data and results carefully. But most writing relates to an argument, and you won't know what your argument is until you have finished the research.

The other thing is, that you learn by writing: arguments only get clarified, or even only occur to you, as you try to write. This means the real beneficiary of your writing is you. This suggests that you should first write for yourself, then later rewrite for the reader. Just as your lecture notes, or notes from papers you read, may not make much sense to other people but are useful to you, so your first attempts to write up an experiment or present an argument may be better done privately with no worry about making sense to other people. I do this more and more: write things up for myself, and separately (later) for others. When writing for yourself, you don't have to explain background or motivation, just focus on the facts and above all on the argument. When that is straight you can later go back and put in what it needs to make sense to an outsider. In many ways, all that "glue" and structure is helping the reader to those things.

In summary, you need to practise.

Another way of expressing this

Here is a more chatty way of putting all this, that I wrote to a doctoral student I'd met. He said:
"I have great difficulty in believing one of my supervisors who, to me, seems to subscribe to the view that this [writing up a thesis] just happens!"

I replied:
I can recognise your problem, but may be a really bad person to advise you. That's because for me it just did happen. I showed no sign of wanting to finish for years (this wouldn't be tolerated in the UK today) and finally I got this attractive postdoc position. My new boss told me he wasn't legally allowed to pay me until he had my PhD certificate in his hand. This turned out to be completely untrue, but productive. I sat down, and invented a story to connect all the bits I'd done, and a jolly good story it seemed to me too, even though I'd never had a plan in advance: it just did happen. So I acted like your supervisor, but obviously I'm a glib, confident bastard and that did the trick. Your talk however suggested that you have done loads of stuff, but don't realise its value.

So I suppose my advice is: just start composing it. Sit down, and invent a story about it all. The "thesis" is the story: the overall argument you want to make. I've come across some poor students here who had real difficulties with this: they acted as if they just had to do the bits and never mind how they fitted, so they did their lit. review in year 1 and of course their work changed and it was no longer a good fit; they wrote up their experiments as if saying what they did was all that mattered; and had nothing much to say in conclusion probably because they never thought about their work having any significance. It gave the impression there was no point to it all, or certainly that they had never thought about that. So don't do that.

The point of making up the story is now to concentrate on how all your bits fit together; what you want to end up saying, how your work supports that conclusion, how the literature relates to the argument you are now putting forward. It doesn't have to be a true story of why you did that: but a self-critical and measured account of what you now think can be said about it.

If you are really stuck about what your story essentially is, try this: I often do it when writing a paper, though I don't remember doing it for my thesis. What I do, is try to summarise it for someone orally. It's obvious they they will lose patience within a minute or two, so I have to pick out what my absolutely most important point is, and state it simply in a way someone can understand. Once I'm clear about that, I can go off and write the long version, organised around what I've discovered is my main message.

Any fool can do a brain dump of all the details they know (pick up any computer manual): the hard bit is to organise it around a single consistent theme/message/story. This takes a bit of work for a paper. For a thesis, it's harder again because it's longer; but once you've got your story straight, the rest is just work, filling in chunks, and writing "glue" to keep the (inattentive) reader straight about what each bit is doing w.r.t. the whole.

I suppose one reason I didn't have this kind of problem myself, was that the first thing I ever "published" (though only in a magazine like SIGCHI bulletin) was a review of another PhD thesis. It was a moderately good bit of research from Stanford, but really, really badly written up, so translating it into a comprehensible summary was useful to others. But I guess it made it clear that I could do at least as well as some of the other work around. Nothing like reading some other theses to give you some strong opinions about what not to do, and that you're the boy to do it!

Although in many disciplines, PhDs have a fairly standard conventional format, this is less true for original work and interdisciplinary areas like HCI. The key thing is to construct something that other researchers want to read, and can find what they need in it. They are going to be interested, but without a lot of your special knowledge, and with some different assumptions. Construct a reasonable argument: why you think you can believe what you do, what it depends on, why some other people's assumptions aren't relevant to what you do, etc.


Bazerman, C. (1985) "Physicists reading physics" Written Communication vol.2 pp.3-24 [Interview study with 7 physicists on their reading of journal articles]

Burrough-Boenisch, J. (1998) Survey of EASE Confererence delegates sheds light on IMRAD reading strategies. European Science Editing vol.24 no.1 pp.3-5.

Dillon, A., Richardson, J. & McNight, C. (1989) "Human factors of journal usage and design of electronic texts" Interacting with Computers vol.1 pp.183-189.

Hartley, J. (1999) "From structured abstracts to structured articles: A modest proposal" Journal of Technical Writing and Communication vol.29 no.3 pp.255-270. [contains a good summary of literature showing that non-linear reading strategies for scientific papers are widespread.]

Web site logical path: [] [~steve] [resources] [postgrads] [this page]
[Top of this page]