This year I have read seven PhD Thesisiseeses (well, you tell me the plural of thesis!). Yes, that is too many to read in one year, no, I did not plan it that way.
The task was made far more manageable by the fact that each student understood the need to support their thesis reader through the process. Reading a thesis is a significant task and a challenging one, whether as supervisor or examiner. A clear structure is an absolutely essential part of the thesis but not one that is necessarily easy – certainly not intuitive – at which to arrive. So based on my large quantity of recent experience, and encouraging sounds on twitter, here’s my contribution to the ever-growing body of advice on writing a thesis.
I should briefly note that all the thesisae that I read this year were psychology ones. I am going to try to make this advice as general-purpose as possible, but if you’re working in a fundamentally different field – e.g. not collecting new data from human participants – I think it may not apply.
Starting in the middle
This blog isn’t about thesis writing per se, but I am going to follow the order I suggest you use for writing. This means starting in the middle, not writing the thesis in chronological order from start to finish. The middle is the easiest part to write and will define to some extent what you put in your introduction and discussion. This isn’t about changing your hypotheses to fit your data, it’s about setting up the actual discoveries you have made – which might be different from the discoveries you planned or hoped to make.
The middle part of your thesis will almost certainly look like a series of journal articles. I call these “results chapters” but each one in fact needs a short introduction, methods section, results and discussion. In an ideal world, your thesis will have over-arching research questions (more on these below) and each results chapter will address a different one of these questions. For example*, a thesis might ask:
- how can we measure happiness in people with dementia?
- how does happiness compare between people with dementia living with a spouse versus living in residential care?
- what are the key predictors of happiness in people with dementia?
This thesis could have three results chapters, developing and validating a new measure, applying it cross-sectionally to two groups and comparing them, and examining longitudinal predictors of scores on that measure
Including journal articles
Results chapters won’t always just look a bit like a journal article. In fact, some of these chapters might actually be submitted or even published manuscripts. You should decide which ones will be submitted to a journal before the thesis is submitted in consultation with your supervisors – and check your University regulations for how a submitted manuscript should be incorporated into a thesis. There are two important structural differences, I think, between a Results chapter that is written purely as a chapter, and one which is written as a paper:
First, this piece of writing will need to have a “stand alone” status because the journal reader won’t have the rest of your thesis to enrich their understanding. This will inevitably make the chapter a bit repetitive (more on repetition below) because you’ll be citing a much-condensed sub-set of the literature which will also be covered in your literature review chapter. Don’t worry at all about the repetition. Do make it completely clear to your examiners that this is a journal article though – include the abstract and key words, as well as a note of the citation so they can see where it is submitted / published. They will understand the need to repeat some key issues and will evaluate the chapter accordingly.
Second, this piece of writing will almost certainly want a couple of “bookend” sections. These will sit at the start of the chapter (or maybe at the very end of the previous chapter) and at the end. The bookends provide a bridge between the chapter content – i.e. your journal manuscript – and the rest of the thesis. This is because the manuscript will not only repeat some stuff from elsewhere in the thesis, it also might not flow perfectly in the stream of argument you are trying to construct. To publish your manuscript, especially if it has already been revised in response to reviewer comments, you might have a different emphasis than you would if the writing served only as a chapter. So these bookends are a really useful way to explicitly connect the manuscript to the wider argument and purpose of the thesis as a whole.
Some thesisees may want a stand-alone methods chapter, while others will incorporate all of the methods reporting into the results chapters.
Good reasons to include a methods chapter might be
- all of your results chapters draw on the same methodological framework and source – such as a longitudinal cohort study, a routine data source, or a complex new sample that you collected yourself
- a significant part of your thesis work concerned the careful and systematic selection of measures for use in the subsequent analyses
- you undertook significant preparatory, pilot or proof-of-concept work such as translation of measures, development of a coding scheme, or testing and adapting your data collection protocol
In all of these cases you might not have a results chapter per se, but your thesis would benefit from a dedicated space to be able to describe in detail this methods development process. If you spent a significant portion of your time on this work then take a moment to do it justice, and get the credit
What about repetition?
If you are writing a methods chapter, worries about repetition can creep in again. These are unfounded. In the first place, repetition is not an inherently bad thing and your examiners will be grateful for a certain amount of familiarity as they navigate through the thesis.
Certainly, writing a separate methods chapter does not negate the need to report your methods in your Results chapters too – whether these are written as journal articles or not. For example, if your Methods chapter describes a large longitudinal cohort, each individual Results chapter will need to describe the participants within that cohort who actually contributed data to that specific analysis – it will almost certainly be a different subset each time, depending on eligibility, drop out and missing data.
What you can do, however, is use cross-referencing in your Results to signpost the full detail in the Methods chapter. So for example, in the Measures section of your Methods, you might list the questionnaires that are used in the upcoming analysis, but also cross-reference with the Methods section where the reader can find complete detail on why they were chosen, their reliability and validity, and how they are scored.
NB: this cross-referencing won’t work if your Results chapter is in fact a journal manuscript of course. In this case, you will include the manuscript exactly as submitted / published which means no cross-referencing to other parts of the thesis
Just as we’ve seen with the methods, your individual Results chapters will include some discussion of the findings. These will want to focus on using the findings you’ve just presented to answer the research question posed within the chapter, and draw out implications which will – in a perfect world – set up the chapter that comes next: “so we’ve seen that people with dementia are happier when living with a spouse than in residential care, but there is also large variability in happiness in both groups. The next question then is to investigate what explains this variability in happiness.”
For students then, this often begs the question of what goes in the overall thesis Discussion chapter, to avoid a simple repetition of the smaller discussions that have preceded it. I’ve developed some pretty strong views about this… and here’s the structure I have recommended for every student of mine who has submitted this year:
A concise recap of results.
This could follow the structure of the chapters, revising the results in chronological order, but it can be more interesting and more enlightening to take a cross-cutting approach to summarising what you found. For example, (still working with my fictional dementia and happiness thesis) you might start by recapping what you’ve discovered about the role of a significant other in contributing to happiness (Results chapter 2) and in measuring happiness via your new observer-report questionnaire (Results chapter 1). Then you might go on to examine the effect of dementia progression on happiness, drawing on both Results chapters 2 and 3. And so on… Whatever you do in this part, keep it concise. I would aim for a page, and two at the most.
Consideration of limitations
I think lots of people would recommend that limitations go nearer the end of your discussion, but I’m going to argue for putting them up front. First, you probably have quite a few because thesis research rarely goes to plan and you often have to compromise along the way. Second, you may be worrying about how you can draw any conclusions at all from your thesis, given all these limitations (and / or null results – these happen a lot too because, guess what, all those papers you’ve read are subject to publication bias and not a representative sample of what happens in research!) The solution to both of these problems is to get the limitations out of the way, up front, and then move on. Then you can give yourself the freedom to examine what your thesis results might mean, safe in the knowledge that you and your readers are both aware of the solidity, or otherwise, of the foundations upon which you base those explorations
Importantly, the goal of the limitations section, then, should be to determine the strength of the case you can make for one thing or another. In other words, focus in on the limitations that determine how robust your conclusions can be. Don’t spend time apologising for choices you made – for example, discussing how a quantitative method with a bigger sample would have been “better” than the small-sample qualitative research that you carried out. Stick to your guns on this one. Instead, consider the limitations that applied within those methodological choices – for example, if you set out to examine gender differences, did you get gender diversity and balance in your sample? if not, this might be a limitation on your ability to draw robust conclusions about gender.
Implications: for research, theory, practice
Having established your limitations, you now have a wee bit more freedom to discuss the implications of your results. Personally, I think it is okay to position yourself as saying something like: “let’s presume that the patterns reported in this thesis are robust and replicable – in that case what would the implications be?” You aren’t pretending the work is more substantial than it is, but you are giving yourself permission to undertake the intellectual exercise of considering what it means.
I’d start with the Implications for Future Research, which is one of the easiest to write. This is a good place to put all those worries about “maybe I should have done a completely different study”. If you are stricken with doubt about your decision to measure the happiness of people with dementia via observer report, this is where you can call on future researchers to find more direct or objective ways to capture happiness. A good way to approach this section is to consider what you would want to tell a brand new PhD student starting in the same topic area today: “whatever you do, don’t forget to capture household income as a co-variate” can easily turn into a great recommendation for research in the field. Do make sure you do more than just call for replications and bigger sample sizes too – we want to see some concrete recommendations informed by your specific discoveries, please.
Next up I would think about Implications for Theory – what does your research tell us about possible mechanisms: the reasons why X happens, or A relates to B? Try to lift yourself out of the nitty gritty of your measures and findings to consider the bigger picture. Go back to your introduction and think about the theoretical models that influenced your work – do your data tend to align with one model more than another? Do they support or contradict the dominant expectations in your field – and if so, could you even propose an alternate? This doesn’t have to be a fully fleshed out theoretical proposition. You might simply state that existing models have failed to take account of the important role of social support.
Finally, the fun part (I think) is the Implications for Practice. This might be about healthcare, about education or criminal justice. Or perhaps you’re simply talking about day to day decisions – implications for parents, or job seekers, or tourists, or pet-owners. Remember in this part that while we have established that you have permission to assume that the patterns in your data are robust, that’s not the same as permission to extrapolate wildly. Rather than “all people with dementia should be supported to live at home for as long as possible” maybe you want to say “the importance of social support should not be underestimated when making decisions about the best place to care for people with dementia”.
The last part of your Discussion chapter will be the Conclusion. In this case, make sure you do what it says on the tin – this is about drawing a final conclusion from your work, and Not about repeating your results. If an old school friend asked you what you had been doing these past few years, and specifically asked what you found, what would you be telling them? This is a good place to start when deciding what to put in your Conclusion, which should be as short as you can get it – one paragraph is fine, 2 or 3 at the most
The literature review(s)
The hardest part of your thesis structure to get right is the literature review. I’m not talking here about a systematic review – if you have one in your thesis this should really be considered a results chapter, where your participants are the papers you recruited, and from which you collected data. I’m talking about a structured argument that draws on a comprehensive array of relevant evidence to:
a) demonstrate what is already known in your field of study
b) demonstrate what remains to be discovered
c) thereby set-up the research questions you are asking, and (to some extent at least) the methods you’re using to answer them
A regular literature review should always follow a “funnel” structure from more general to more specific. In the case of our sample thesis you might be doing something like this:
dementia (definition, prevalence, impact) -> caring for people with dementia (principles, at home vs in a residence) -> quality of life and caring for people with dementia -> happiness as a factor in quality of life when caring for people with dementia -> your thesis questions
However, imagine that this thesis was also exploring how the happiness of people with dementia varied by the context in which they lived specifically for immigrant communities in the UK. In that case you also need to cram a literature on immigration (definition, prevalence, impact) plus cultural differences in caring traditions, quality of life and markers of happiness. Not to mention you need to find a place to examine any work that has already examined parts of the intersection of these two topics – such as differences between ethnic groups in dementia prevalence.
So this is now a very hard job, and I’m afraid I don’t have an easy solution for you. Basically, you just need to choose one funnel and follow it almost to the narrowest part, then stop, zoom back out again, and embark on a second funnel. Make sure you signpost this explicitly for the reader too – “Having considered the evidence on quality of life and dementia care in general terms, we now need to examine how these factors operate in a particular population: immigrants to the UK. To do so, I will first provide some basic information about the nature, prevalence and impacts of UK immigration according to the latest research findings”. This makes it clear that having considered some quite specific findings we are now zooming back out, to take a broader via on a new topic, before we connect those dots back up again later
Do you need a separate introduction?
Some thesis’s that I’ve read have also had a separate Introduction chapter, distinct from the background literature. This tends to be a short overview of the topic area, providing some key details and definitions, and may include a preview of some methodological aspects too. I don’t think this is always needed – your thesis abstract serves much the same purpose after all. However if you are weaving together a lot of different literatures, or taking an approach which might be quite innovative or counter-intuitive for the field, an Introduction chapter can be a good idea. This is chance to set out your stall, making sure your reader understands the definition of – for example – happiness that you will be adopting thereafter. You might also want to get them on board with the idea that you are conducting design-based, or ethnographic, or action research – especially if most of the literature you are drawing on is clinical trials. There may even be some myth-busting to do if the constructs you’re working with are persistently misunderstood. The purpose of the Introduction, if you have one, should be positioning and it will be especially valuable if you think the default position in the literature, and your position in this thesis, are quite different
The research questions and the thesis spine
This blog post has become veeerrry long – as befits a post about thesis structure, I guess. But before I finish I just want to say something about the concept of your thesis having a spine – or maybe a golden thread, that runs all the way through what you write. This spine should be constructed from your research questions and they are far and away the most important part of your thesis to get right. You should have something in the region of 2-5 research questions that guide your entire thesis. Your literature review should expose these as important questions which need to be addressed to allow the field to move forward – in terms of knowledge generation, theory development and / or real world practice. Your methods should be tailored to addressing these questions. Don’t ask about the impact of an intervention if you don’t measure outcomes robustly – instead ask about its potential, or acceptability, or feasibility. Your results chapters should answer the questions – sometimes one at a time, and sometimes collectively, approaching the same groups of over-arching questions from a few different angles. And your discussion should settle on an answer to those questions, albeit contextualised by limitations, and with inevitable new questions raised.
This is my proposal then, for how to structure a thesis. It won’t work for everyone – especially methods that fall far from what I am used to – but it might be a useful guide. In any case, having a clear sense of the structure of your thesis will help you write. If you’re sitting down to write section 3.2.1 that’s far less intimidating than sitting down to write A Thesis.
I guess the only thing left to say, is Good Luck!
* I will use this example throughout and it is entirely made-up, as are the hypothetical findings – any relation to a real life thesis is purely coincidental!