Author: James

1. When experiments have to be done quickly because you are using living things or reacting chemicals which will begin to behave aberrantly if left too long, then you still shouldn’t rush. You are much more likely to have a negative effect on the experiment if you rush. Just move as quickly as you can without focusing on speed.
2. If you are doing cell culture or playing with small animals, rushing can be particularly detrimental. Your speed could hurt or damage them.

By the start of the third year of my PhD, the thought of weekend work always filled me with a cold dread. This wasn’t because I preferred to sit at home watching Red Dwarf and eating cheese puffs – though this remains unquestionably true – it was because whenever I worked at weekends I rushed it.

As consistent as my aversion to yellow snow, this resulted in me making obvious and highly avoidable mistakes, which meant I might as well have stayed at home (except for the health benefits of not eating the cheese puffs).

Not labelling things to save time was a key error. Most times I got away with it, but when I didn’t the feeling that I’d basically flushed my weekend down the toilet was not something I’d recommend.

1. When you are doing more than one procedure at a time, plan out the steps so that they do not overlap at any point.
2. Planning the time it will take you to complete protocols requires practice. At the start you will inevitably underestimate it. Therefore, if you are trying to fit several things all in the same day, allow extra time to do each bit.
3. If you need to do multiple procedures at once, organisation is the key. Don’t rush in to starting the first one or you might as well not bother. Plan them all thoroughly before starting, so that you know exactly what to do before you have to do it.

1. If you are planning to do multiple samples all at once, the first ones to be put in treatment might remain there for significantly longer than the last ones. This can be avoided by always starting with the same sample and staggering them, so that you only do a subset of samples at a time.
2. Use a lab timer. If you don’t have one, then ask for one. Don’t use your phone for this. While you might have a great app, you might also get something nasty on your phone. Then on your face.

I can remember weeks when my planner looked like a code book with the page lines cramming multiple sentences on top of each other. I had fallen victim to the fallacy that because I could fit the words describing the tasks in my calendar, it meant I could fit the time it would take to do them into the hours of a day.

When people were using equipment I needed, or it did not immediately work at the required speed, one experiment started to eat into another. Sometimes I was there well into the night, and generally the quality of the work was lower than if I had done them on separate days.

1. Even whilst you are performing the same technique over and over it is a good idea to read through the protocol before and after to make sure you are doing it right. If you find you are forgetting steps then highlight them, and specifically remind yourself about those steps before the next experiment.
2. Never assume that steps in established protocols are not necessary. The manufacturers have not added things just to make their procedures look more detailed. Of course, if you understand the protocol and you know that for your purposes specific steps are not needed, then that is different.

Once I’d organised myself and stopped trying to cram as much work into each experiment as possible the quality of my data improved. My main problem became that I paid less attention to the protocol because I assumed I knew it.

I rarely missed anything big – I was not such a fool to attempt a protocol I didn’t know pretty well without looking at it. Usually, the timing was a bit off or I didn’t shake mixtures, or I didn’t pre-prepare things that took time to prepare, but small things can still make big differences.

Fortunately, I’ll never know how much extra data I would have got if I’d followed this guideline, but I imagine it isn’t small.

Bob has designed a machine for detecting the number of aliens on other planets. He looks at Mars and as expected finds no aliens. Then he looks at two other planets and finds no evidence of alien life there either. Disappointed, he publishes that there are no aliens on either planet. It is only after the invasion that Bob realises the machine wasn’t turned on.

Bob had a negative control which was Mars, where he knew there were no aliens, and when his machine showed no aliens he knew he was not getting false positives. However, Bob had no positive control to show that his machine was actually capable of detecting aliens to rule out false negatives. Your repercussions will probably not be as bad as Bob’s, but if you have any desire for publication, even the lower impact journals will insist on these controls.

• Don’t just think about your controls on the day you carry out the procedure. You may well need extra components or additions which have to be ordered or pre-prepared. Like the rest of your protocol, this should be planned well in advance.

• Sometimes the most obvious controls are not the best. Try to avoid using substances you are unfamiliar with, or are unreliable. Use whatever substance or object is most likely to work. If your experiment worked but your controls did not, then your experiment is still void, and you will have no way of knowing what needs changing.

In the last section of my thesis I had a few experiments which I included to flesh it out a bit.

This was a costly and pointless decision simply because in the rush to complete them I had not used the relevant controls. The result was that my viva took an additional (fairly harrowing) hour where both examiners drilled me on what I could possibly conclude from these data. My answer, irrevocably, was that there was nothing. What else could I say?

No controls = no data.

I was forced to delete the entire section from my thesis. What a waste of time that could so easily have been a solid part of my project.

1. If you are collecting data from computer files then you could get someone in your lab to rename all the files for you with random letters, and note down which file corresponds to which letter. If you do this, be sure to have another version of each file saved somewhere else, in case your friend loses the key or makes a mistake. Always check afterward they have not made a mistake.
2. Not all experiments can be blinded. Sometimes you will remember which group the data comes from when you are analysing it, or some groups might be observably different as a result of the treatments used. It is inaccurate to say these data are blinded even if you tried.

I once blinded a study by wrapping bits of foil over the labels of my slides then lettering the foil and saving this letter as the file name when I analysed the data. I then went back to match the letters on the foil to the name of the treatment written on the slide beneath. I picked up the tray and it snapped, sending everything to the floor. Needless to say, my loosely attached bits of foil did not survive the journey unscathed. I was left with a bunch of broken slides, some lettered foil and the haunting certainty that I was going to have to start over.

1. A detailed explanation of why you did everything.
2. The method you used to collect the data in as much detail as possible.
3. Your results with all the statistics.
4. What you concluded from the work you have done.
5. What are your thoughts for further work.
6. If you have any ideas for the long term direction of the project you could also include this.

The most important part to include is the methods you used and any results that went wrong. It will of course be embarrassing if you have made mistakes, but THAT IS WHAT THESE MEETINGS ARE FOR. The earlier you learn why things are going wrong the earlier you can put them right.

It is a mistake to think you can make these corrections on your own. It is human nature to bias our opinion of our work to protect ourselves from the depressing realisation that we are bad at things. Other people will spot your mistakes more easily, so show them.

1. Make sure you understand why you used the stats in your presentation.
2. Make the presentations look good. Don’t waste time spending hours changing fonts, but remember that a lot of what your supervisor thinks of you will be based on these meetings. They don’t see the hours you spend working out of hours. Make sure your data are well organised and coherent.
3. Label the graph axes thoroughly. Don’t use nondescript terms like increase or change, but describe what is increased or changed within the axis. If the description is too long, add a paragraph below the graph to show the explanation.
4. Show procedures that haven’t worked. There is little to learn from things that worked well, but where they failed there is probably a reason, and your supervisor or someone else in the group may know it.
5. Meet with your supervisor as regularly as they allow. If they are too busy or you don’t get on with them, then find someone else who will chat to you about your work.
6. When meeting one on one, have a plan of each point you need to raise pre-written on a notepad so that you can tick them off as they are discussed.

I used to hate presenting at lab meetings. As a result, I put little effort into making the data look presentable. The frequent and impenitent lack of stats or labelled axes earned me some harsh words from my supervisor. It took me too long to realise that just because these things were unimportant to me, didn’t mean they were to him.

I also only showed the data that I thought were interesting, and left off the data that didn’t work. Not only did my very brief presentations not impress anyone, but the experiments that were failing kept failing and I was forced to look for explanations elsewhere. When I finally showed the failed results, my supervisor and the rest of the lab made several very good suggestions that I tried.

It still didn’t work, but that’s science. At least I got to write that I tried those things in my thesis, which saved me from further scrutiny in my viva.

• Time allowance is consistently underestimated. Always allow yourself longer to do a task than you think you will need.
• People doing animal studies might be told they need 12 animals to make the study statistically robust. Using 12 would then fail to account for what happens if one of them dies, or if the variation is much greater than expected.
• Human studies should always follow the same principle. People can drop out or never turn up.
• If you have to deal with costs, then leave space in your budget for unexpected problems. Science is notoriously unpredictable.
• When using expensive (or even cheap) reagents, students often use just enough of it to completely screw everything up. This is an entirely false economy, as you will find out when you have to do the experiment again.
• When you are doing things in bulk make sure to add extra reagent or your last few samples might not receive the same volume of solution.

1. Take the time to actually calculate how much of something you will need and then add extra.
2. Ethical and safety approval is something that frequently takes far longer than students think. Make sure you allow for this in your experimental plan.

When you’ve just spent £500-700 on a tiny tube containing 3-4μl of liquid, your immediate inclination is to use as little of it as possible and make it last. I’ve done this over and over… and over and over. Then I probably went home and ruined dinner by not adding enough oil to the pan.

I would add just enough liquid to cover my slides so as not to waste reagents. This was frequently a disaster. The surface tension broke, the liquid spilled, or the slide dried still untreated. I might as well have poured that expensive reagent down the drain.

Secondly, I also frequently forgot that pipettes are not magic calculation devices. They carry a degree of error, and frequently take up more liquid then they are supposed to. Not a lot, but when I had 20 samples all receiving reagent from the same volume, if I hadn’t made enough for 22 there wasn’t always enough to cover the last sample equally.

• Don’t completely ignore them and do your own work. If they don’t like your ideas then discuss the scientific merits with them further. If they have questions, answer them, and if they find problems, find ways around them.
• If you don’t get along with your supervisor then that’s fine. Not everyone can get along. But that doesn’t mean you should deliberately antagonize them. Try and be as polite as possible, and when possible mend fences.
• Don’t form collaborations with other groups behind the back of your supervisor. This may embarrass them and they may not react well when they find out. Tell them what you are doing and what relevance it has to your project. If it doesn’t have any relevance, then you probably shouldn’t be doing it. Also state that you are doing it on the side, not at the cost of work they want you to do.
• If you are working hard out of hours, send them emails with updates to let them know. Otherwise you might be unpleasantly surprised by their impression of you.

1. Remember how easy it is to be biased towards your own work. Don’t just think about how good the data will be if they show exactly what you want them to show. Think about what else they might show, and whether that is still worth knowing.
2. Think whether you are doing the procedures the best way, and are you using the most precise measurement of the parameter you are testing.
3. Read literature in agreement and disagreement with your hypothesis, and consider the implications of the latter.

I was a big one for designing my own experiments. I was lucky enough to be passionate about what I was doing and my interest spawned ideas. In my head, all of them were ground breaking, but rarely did they survive a trip to my supervisor.

He was not being cruel; he had just spotted things about them that I hadn’t. Objectivity and experience allowed him to see the flaws in my work that I could not.

Whenever I tried to bypass his approval and run the experiments, I was always disappointed with my results. They all contained some flaw that I had not anticipated, or suffered from bad planning. I could have saved a lot of time and lab money by not bothering.

Others I have known have compounded this mistake by adding their own data unapproved to their thesis, and vehemently defending it in their viva. This almost led to failing their PhD.

Apply objectivity to your own work as you would to other people’s. Respect your supervisor and listen to their advice. It makes it all the more satisfying when they like your ideas.