• Question: because your a scientist and you do experiments,well when you do an experiment/theory/test,do you do several tests to see if your results are coerrect or to see if you get an anomoly(an odd result)?

    Asked by charlotteharrison to Anne, Carolyn, Joe, Mariana, Nick on 16 Mar 2010 in Categories: .
    • Photo: Nick Bradshaw

      Nick Bradshaw answered on 16 Mar 2010:

      Yes, all of the time. It depends on the type of experiment, but normally I would perform any test at least three times to check that I got the same result.

    • Photo: Mariana Vargas

      Mariana Vargas answered on 16 Mar 2010:

      Yes this is a very good question. It is very important to
      – Repeat an experiment enough times to make sure that what you are seeing is not an “odd one out”. The number of experiments you do depends of many things such as how variable your measurements are.
      That is why in a research publication it is important to provide the statistics where you show what your mean and standard deviation are.
      – Also it is hugely important to do “control experiments”. For example if you give a drug to your subjects, you would have two groups: the first where you give them the drug and another grup where you do everything the same but the “drug” is just a placebo, so that you can discard any other effects that are not due to your drug. So you would need to compare the statistics of your control group and your experimental group, and show whether they are significantly different or not.

    • Photo: Joseph Devlin

      Joseph Devlin answered on 16 Mar 2010:

      Yes, there is a limit to how much any one experiment can show you and even assuming you did everything perfectly correctly, there is a 1/20 risk of a false finding (assuming normal statistics). So anything brand new that we find, we try to replicate at least twice more in two independent experiments before we start to really believe in it. It’s even better if another group replicates the result completely independently — that’s always very encouraging.

      A few years ago we found an odd thing about the organisation of human hearing in the brain. For basically all species of animals that have been tested, when you present a sound to one ear, there is a larger brain response on the opposite side of the brain. So, for instance, if you play a noise to the left ear of a monkey, rat, gerbil, ferret, etc then both sides of their brain respond to it, but the stronger response is on the right. When we looked at this in humans we got a really surprising result — we found that no matter which ear you presented a simple tone to, the stronger response was on the left. We basically didn’t believe this because it so didn’t fit with the existing literature so we replicated this in a second set of subjects and then started to wonder: was this a biological bias towards processing sounds in the left hemisphere and if so, could this be why language tends to be processed more strongly in the left hemisphere? I still don’t know the answers to these questions — they are things we are actively working on — but when other groups started to find similar results, it made me think there was probably something truly different about humans that is exciting.

    • Photo: Carolyn McGettigan

      Carolyn McGettigan answered on 16 Mar 2010:

      Good question. Most of our experiments have a thread i.e. we see how one experiment turns out and then we might run it again with a different condition included, or with a slightly different task for the participants. It is important to have a sense of how consistent our results are.

      Actually, when it comes to neuroimaging research, yours is a very important question. It’s very expensive to run an MRI scanner, and so it’s not so easy to repeat experiments just to check that the results are replicable. Each experiment costs thousands of pounds. Also, the research journals aren’t very interested in publishing studies that are a straight replication of something already published – they always want something new. So the costs and the bias in the publications means that studies don’t get replicated as often as they should. This is a problem. However, we always look at other published research for commonalities and differences between studies, and because there is such a vast amount of research out there, we can begin to identify the more reliable and consistent findings.

Comments