Dissertation Do's and Don'ts

Advice on preparing laboratory dissertations in the Medical and Physiological sciences

by Dr Jan Schnupp

The dissertation is a very important component of the Final Honours Schools in Medical Sciences, Physiological Sciences and PPP, and it is a rather different animal from the essay based exam papers that you will be used to. Here I describe so1me of the things that you should watch out for if you have to prepare a lab project based dissertation in one of these final honours schools. This is all based on my recent experiences as FHS examiner, but it reflects my personal opinion and experience, it is NOT an "official communication" of the exam board. Nevertheless I think may students will find the following comments to be helpful advice as they embark on a one-off journey into rather new territory. It is written specifically with lab dissertations in mind. Library dissertations are quite different, but some of the advice below may still apply.

In 2006 the dissertation weighs equally to other parts of the exam for the medical students, but there is a large fraction of the faculty who would wish the dissertation to weigh more heavily than, for example, the extended essay, and it is likely that there will be a recommendation for regulations to change so that the dissertation becomes more heavily weighted. For the Physiologists the dissertation already counts "double" in the calculation of the final marks. Furthermore, when examiners discuss borderline cases, it is not uncommon that a particularly good or bad dissertation mark is used as an argument to decide whether to push a student up or not. This reflects Oxford's desire to be a leading research University, so the research project should weigh heavily. But while this is a perfectly proper attitude to take for the University, it does mean that students embarking on the FHS find that the part of the examination that they are least familiar with, and the one part where more things can go wrong than in any other, counts most heavily. To put it bluntly:

A bad mark in a dissertation is a real millstone around your neck with respect to your final mark!

So should you be worried? No! But you do need to smarten up about what is required and play your cards right, because you only get one stab at this, and if it goes well you will find this to be the most rewarding part of your studies. So what do you need to do to get your dissertation off to a flying start?

Start with the end in mind

Hopefully you will choose a dissertation project that interests you, and you will do it for the science, and not as a purely marks-oriented compulsory exercise. Nevertheless, even if you do it "for the love of it" you still want to get decent marks at the end of it. Which is why you need to understand from the outset how your project will be assessed. If you want to optimize the number of marks you get out of your project, then you must undertand one thing very clearly:

The examiners do not mark your project as such, nor are they supposed to. Essentially, they mark your project REPORT, and to a lesser extent, your viva voce presentation.

This is a very important difference! It means that you can have a phantastically succesful run in the lab, find a cure for cancer, regenerate amputated limbs in adult mammals, make a living organism out of inanimate mud, solve the mind-body problem, but still end up with a very low mark if you don't write your work up in a way that describes these achievements effectively and makes it quite plain to the examiners that you have done these things in an insightful way .If the project is a great success but badly written up then the examiners are likely to beleive these achievements to be those of your supervisor, not yours, and they will probably mark accordingly.

You may think this is unfair, after all you worked very hard and everything worked out so well. But bear in mind that this cuts both ways: imagine your project does not go well at all, despite your best efforts none of the experiments worked as planned, everything went pear shaped, and you end up with zero data. Not a good position to be in if you can avoid it, but with over 150 projects done every year, there are always more than just a few unlucky students who find themselves in that position. However, if they write a thoughful and clear report that describes what they were trying to do and why, what they would reasonably have expected to find and why, make sensible suggestions as to why or where things did not work out, and show that they have learned a lot from the experience then they still have a chance to get a good mark.

If your supervisor has supervised a number of other projects in the past, then he or she should have copies of the previous years' candidates, and he should have been fed back the degree class awarded to those projects. Ask if you can see any of the previous years' writeups. Ask what marks they got. Ask the students in the year above you to show you theirs. Get a feeling for what is expected of you write-up wise and try to work out why some write-ups might have scored higher than others.

So what do examiners look for?

Examiners tend to like 'hypothesis driven' research.

In a nutshell, a lab dissertation should have the following structure:

Intro: "Inspired by previous work X, I hypothesize that Y might happen if I do Z."
Methods: "this is how I performed Z, and how I attempted to observe Y."
Results: "this is what I saw when I tried to observe Y."
Discussion: "my observations do / do partly / do not at all conform with my original hypothesis. Therefore the hypothesis :"

"has been confirmed entirely" (possible but unlikely: don't oversell your project, even if it went well!)
"may have to be modified in the light of the new findings" (good, be creative!)
"should be discarded" (why not, 'falsification' is thought to be a powerful tool in science, but it does beg the question whether it was a sensible hypothesis to go for in the first place. It may not have been, and if in retrospect you think it was a bad hypothesis to start with, say so! The examiners are likely to blame the supervisor for sending you off on a wild goose chase and will like the fact that you are honest about it.)
"will have to be investigated further in such and such a manner in the future" (most likely outcome, and therefore not a terribly original ending, but if it's the best you can come up with then it will do fine)

Note that not all research is hypothesis driven, but most of it is. Consequently the examiners will expect your writeup to conform to the above structure, and you should deviate from it only if there is a very good reason to do so. If you work on something extremely unusual it may be very difficult to predict what you might find. If your project is a fishing expedition, fine, but be clear and upfront about it in your write-up. Also: it is not uncommon for the aims of a project to change. You may set off investigating hypothesis A but hit a brick wall and then end up going after a rather different hypothesis B. You won't have space or time to describe these trials and tribulations. They are part of research, but make for lengthy and boring naratives, so at the end of the project, apply for a "change of title" and write it up as if you'd been after hypothesis B all along.

Examiners tend to look for "properly controlled" research.

Note that the discussion is an intepretation of your data, and the examiners will want to see that this is done thoughtfully. Be very mindful of the possibility that there may be alternative, equally valid interpretations. Particularly if your experimental design misses some important control experiments, then the interpretation of your data may be ambiguous, and it may not be possible to draw firm conclusions! Is that a problem? Yes! But it's a salvageable one if you notice it during the write-up phase, and, assuming that it's too late to do the controls, your discussion is up-front and honest about the fact that your data cannot be interpreted unambiguously. Ambiguous results are sometimes the result of a flaw in the initial experimental design, and it might have been able to avoid this flaw at the outset by careful planning. However, the examiners may put at least part of the blame for that on the supervisor. After all, at the outset of the project you are an inexperienced rookie. BUT: by the end, as you write up and (hopefully!) think hard about what your data mean and why, you really ought to notice if there is a problem. Rushing into a project with a flaw is a minor, forgivable blunder for an undergrad who, after all, has to get stuck in and initially trust the supervisor that this project is a good idea. Writing up a flawed project without noticing anything wrong, however is a major blunder. Also, examiners may be prepared to accept that you didn't have the time or the resources to do all the controls that one ideally would like, but they are unlikely to be impressed if you don't notice that important controls are missing, or if you 'pretend' that these controls may not be required. No study can control for everything though, and you may find it hard to judge for yourself whether a particular control experiment would most likely be considered essential, highly desirable, or perhaps an overkill. If in doubt, discuss it with your supervisor and/or your college tutor.

Examiners like to see students take some initiative in their project

Depending on the project you do, you may not have much scope to contribute to the design of the experiment. There maybe limitations in animal laws, access to patients, ethics, data protection, health and safety, limited access to expensive equipment or consumables, etc, which may severely restrict how much you can do yourself when it comes to collecting data. This happens A LOT, and examiners appreciate that, but this is no excuse for you not making at least some sort of an effort to make the project 'your own' in a real and meaningful way.

Whether you can collect the data yourself or whether you have to rely on more competent or qualified members of your supervisor's research group to collect the data for you, either way the data 'belong' to your supervisor, but you MUST make sure you get your own (if necessary suitably anonymized) copy of the data, and become intimately familiar with them. You MUST make an effort to analyse the data yourself as much as possible. If members of your supervisor's lab offer to help with the analysis, try to get them to teach you how to do it yourself instead if it is at all realistic to do so. If you simply accept some summary graphs prepared by a post doc in the lab to paste into your report you may think you've got an easy ride, but not only have you not learned anything, you also run the risk that the examiners will see through this in the viva, and they will not be impressed.

If your supervisor won't let you take copies of the data away with you to play with and analyse at your leisure then complain to your college tutor and consider changing project!

Also, consider doing more than simply following your supervisor's lead with respect to how the data should be analysed. More than likely he or she will have some good suggestions as to how this should be done, but you will only know how good these suggestions really are once you have considered whether there might be other ways of analysing this type of data. If you have friends in College who know some statistics, ask them how they would analyse this type of data, and why. And your supervisor is likely to be mighty impressed if you come into the lab with some fancy non-linear analysis of covariance of the data that he or she had never heard of before. (And if you had to bribe a stats finalist in your college with buckets of beer to help you get your head around this and do this analysis, well, that's what college bars are for, right?)

A hypothetical example
A student signs up for a project which is intended to test whether a particular treatment reduces hand tremor in patients with MS. The idea is that MS patients will be asked to trace out a spiral shape with a stylus on a computer drawing pad, before and after the treatment. The success of the treatment is to be measured by whether the tracing of the spiral shape has become more regular after treatment. At the outset this looks like a sweet, clean, self contained little project.
However, after a two days it is becoming clear that the supervisor found only two patients with MS willing to take part in the study, only one of them has the strength to try to draw any spirals at all and the spirals are unrecognizable in both cases. The undergraduate ends up with a dataset consisting of two squiggly lines, and since he is neither clinically qualified nor well connected, he will struggle to get his hands on any more patients to collect more data. Yet he's supposed to put in six weeks lab work. He's had it, right? End of the road. Kiss the first class degree goodbye.
Well, no. Who said that this project should ONLY be about collecting data from MS patients? Wouldn't it be interesting, in fact necessary, to collect spiral drawings from healthy volunteers, ideally of different ages, to get some sort of 'healthy baseline' data against which the pathological data can be compared? So the student arranges to get his mates to volunteer as healthy subjects, collects their spirals, not once but twice on consecutive days, gets a friend from computer science to help him number crunch the digitized drawings and to calculate inter-subject and intra-subject variability. Being particularly enterprising he gets his college tutor to ask GB for a small research grant (Ł50) which he uses to buy a gallon of vodka, and he now repeats the spiral tracing "steady hands" test with his volunteers before and after consumption of a carefully measured and administered aliquod of ethanol. To make sure that this 'proper science' and not just a trick to get his mates some free booze, the student bribes a friend and colleague who happens to be well versed in statistics with the remaining vodka to teach him how to set up the appropriate ANOVA test on the data.
There may not be enough data in the project to say anything conclusive about MS, but it was worth the try. And by widening the project out, first by 'calibrating' the technique properly on normal volunteers, and then by testing whether the technique might (also) be used to measure the effects of ethanol induced cerebellar ataxia, the student now has more than enough data, and the examiners cannot fail to be impressed with the creativity the student developed when the original plan hit a brick wall. A first-class mark is suddenly well within the realm of possibility.

The moral of the story: when you set out on a project there are countless things that can (and usually will!) go wrong. First-class students are recognized forst and foremost by their attempts to find a way around them. If it means that you have to shift the goal posts a little, so be it: they are your goal posts, and you can shift them quite a lot simply be redrafting the intro to your project a little. If you are not making any progress can't see a way forward yourself, talk to your college tutor, and if all else fails, consider changing project. That last-resort option is of course only open to you if there is still time before the deadline. And because problems usually only become apparent some way into the project, you would be well advised to start your project as early as at all possible!

The role of the supervisor

The role of the supervisor in dissertation projects is perhaps one of the most 'schizophrenic' factors in your project. On the one had it's meant to be YOUR project, on the other hand you will typically do a project dreamt up by your supervisor, in your supervisor's lab, using your supervisor's resources, following your supervisor's instruction. Examiners often spend a not insignificant amount of time trying to 'guess' how much credit or blame for the good or the bad bits about your project respectively should be given to your supervisor rather than yourself. You want to make sure you get the credit you deserve, and deflect blame you don't.

Supervisors are expected to fill iin a little form at the end of the project in which they are supposed to comment on your performance, the difficulty of the project, etc. If your supervisor is fair-minded (most of them are) and you manage your relationship with your supervisor well (I bet nobody had told you that this is something worth doing before!) then there it is likely that the supervisor will write his report so as to deflect some of the blame away from you if things didn't go well, and will make sure that you get the credit you are due for your hard work. You would have though that that's only fair enough, and should be automatic, so what's this 'manage the relationship with your supervisor' business?

Supervisors often have too many other things to do to give you much time. For you the project should be the top priority, for them you come way down the list, and they would probably much prefer if you just got on with it and let them get on with the hundreds of other things they need to do. Supervisors are also often pressured by their Heads of Department or the Head of preclinical teaching to offer more projects than they would like to, so it does happen that not all projects are as well thought through at the outset and the supervisor may be somewhat overstretched. But even if your supervisor would dearly like to be available for you and is excited about your project, chances are he will have to rush off to a College meeting / clinic / interview panel / tutorials / conference etc, etc for much of the time that you are in the lab trying to collect your data. This may or may not create a big problem for you. If you are confident and your project is uncomplicated enough that you can do all the experiments yourself, great. Otherwise, if your supervisor leads a decent size research group, there is a good chance that much of your day-to-day supervision will be mostly done by some of your supervisors experienced grad students or postdocs. That can work very well. But you still have to bear in mind that it is the supervisors, and not their post-docs (!), who are expected to fill in a brief report form on the project and on your performance at the end, and examiners do very much take this into account. So we have this odd situation where the 'ideal' student, certainly from the supervisor's but also from the examiner's point-of-view, is a student who works very independently and requires only little instruction at the outset, but if the student is really that independent, then he can easily slip below the supervisor's 'radar', and when the time comes to fill in the report, the supervisor really has no idea how hard the student may have worked, or how imaginative or persistent he may have been when problems came up. So how do you make sure that you periodically re-appear on your supervisor's radar, but at the same time appear 'independent' and avoid hassling him unnecessarily? It depends a bit on the atmosphere in the lab and the individuals involved, but as a general rule if you don't get the chance to spend at least 15 minutes each week with you supervisor to discuss the progress of your project then you would be well advised to start sending brief progress report e-mails to him or her once a week. Discussing the project with your supervisor's deputies is great, but you still have to keep the 'old guy' in the loop if you want to make sure you get the credit you deserve. (And while you are at it, you might as well send a copy of the e-mail to your college tutor!) Keep it brief, like so:

Dear Dr Notindalabenuff
just a quick note to keep you up to date about the project. I finally managed to get the in-vivo anti-matter proton assays to work on Wednesday but I'm still a bit concerned that the calibration of the fluorescent spectro-plectometer might be out. I'm looking into that. I've collected data from four slime cultures this week. Three of them I had to start over, so I came back into the lab at 9pm on Thursday to finish that off, I hope it's o.k. if I work late (I understand health & safety can be a bit of a bore about this sort of thing). Anyway it's working now and I hope to finish the remaining seven next week. I also discussed the experiment with Dr Nerd, the stats fellow at my college, and he suggested that I should try a cross-balanced latin square design ancova analysis. I have no idea yet what that is, or why it's better for what we are trying to do, but I've just found out that the statistics department run a free 'stats clinic' and I've booked a slot with one of the statisticians there to find out whether this will really help us with our analysis, and if so, how to do it. I'll keep you posted.

Best regards,
Carlotta Ueberstudent

Remember also that a dissertation supervisor on whom you left a lasting good impression can come in very useful later in life when you need someone to write you a letter of recommendation for a job application. Finally a checklist

Plan to spend a lot of time on your project, but not too much, as it's only about 1/5 of your final grade.
Think about ways in which you might be able to develop initiative. In some projects that's very hard to do. Your supervisor won't be happy if you waste hundreds of pounds worth of restriction enzymes because you decided to show 'initiative' by not sticking to the experimental protocol. But in some small way you can probably develop some ideas of your own. It doesn't have to be much, just the smallest spark may suffice to distinguish you from a 'lab monkey' or 'robot', but if all you ever did is mindlessly follow a recipe handed to down by others you can't really expect to get a first.
Try to get the data collection off to a flying start. If there are problems (and there almost certainly will be!) try to fix them yourself if it's safe to do so, otherwise get help. If you can't get help and you've been stuck for a couple of weeks with no progress then you may have to consider a radical rethink. If you don't get any help from the supervisor, contact your college tutor. Better start a different project then spend three months banging your head against the wall.
Unless you see him or her on a very regular basis anyway, send weekly progress reports to your supervisor and cc your college tutor.
Take home copies of the data with you. Play with them. Bore your friends with them. (A true test of friendship).
Don't underestimate how long it takes to write up. Stop doing experiments at least 3 weeks before the deadline. Try to collect nice figures to illustrate your data early on if possible. Well constructed graphs, with appropriate axes labels, error bars, units,... take a long time to prepare but make a very good, professional impression. Use a big font size on axes labels: examiners are old farts with bad eyesight.

Follow the 'standard structure' given above unless you have a very good reason not to.
Does your intro make a clear statement of the question you are investigating?
You should show scolarship, but you don't have much space, so don't put huge number of references of only peripheral relevance into the intro or discussion.
Cut to the chase. Once you've drafted your intro, ask yourself, could I say this with half the words? The prose usually improves if you say it with fewer words.
Start with the Methods and the Results. Before you write your Results, decide what figures you want to use to illustrate the results, and how to order them, then write the text around them.
Leave the Intro until near the end and write the Abstract last, once you have really thought hard about what conclusions can or can not be drawn from your data.
Ask yourself whether your experiments are 'well controlled' and your results conclusive. It doesn't matter too much if you find they aren't provided you don't oversell your results or overlook major gaps.

Giving a professional 5 minute presentation is really hard, but it is a hugely important skill to develop.
Carefully trim the slides down to the essential take-home message. Don't overload them with detail. You can hold a slide or two 'in reserve' just in case the details become important during questioning in the viva but they rarely do.
When making slides, use big fonts because examiners are old farts with bad eyesight.
You MUST practice the talk at least twice to an empty room with a stopwatch, and then at least once (three times if you are shy) with an audience before the viva date. Giving the talk to an empty room, out loud, will feel strange at first but you will be surprised how good an exercise it is. If you feel embarrassed, blame me but do it anyway! If you are a St Peter's or Oriel student then this is not optional, but a required part of your course. (You wouldn't give a solo violin recital without first practicing aloud but on your own either, would you?)
On your first empty-room practice talk, simply try to say everything out loud that you think you need to say in the 5 minutes you've got. You will probably find it won't fit, and that you will have to trim and rearrange your presentation.
The 2nd time round, imagine your audience there, practice the gestures that you will use to draw the attention of the audience to this important detail on the crucial graph but make sure you then turn back to your imaginary audience and make imaginary eye-contact.

Try to enjoy it.