Why do scientific experiments (mainly in neuroscience and psychology) need to be done on more than 1 person? I was under the impression that the most important part of determining the study’s accuracy was to ensure that the outcome had a less than 5% chance of occurring without the theorised variable.
Couldn’t a situation emerge where the outcome was almost certainly attributed to the variable in question even with one person. For example, something extremely random, like (stupid example) a blood clot forming in someone’s left pinky finger after being reminded of childhood trauma (and it was predicted beforehand).
In: 0
Sure. A situation like that could happen. Famous quote but I forget the author, “if I found a talking a pig, nobody would care about my N of 1.”
But statistically, n of 1 could be an accident. If I predict the clock will say 5pm randomly, there is still a chance I’d be right even if I wasn’t accurately predicting the time beforehand. That’s what we have to account for and that’s why we need more than 1 test.
The number of observations necessary depends on the true size of the effect. E.g., if I’m perfectly precise every time I guess the time, you wouldn’t need very many tests to be confident about it. If I’m a mess and only right a little more than random chance, you’d need many, many tests to tell it’s different than chance.
There will always be extraneous variables, but the larger the sample size the easier it is to tell what is or is not an extraneous variable. In one person you would have no way if determining cause and effect if a blood clot happened during some trial, but if only one person out of 100 had that happen, you can more easily determine it was a fluke. You would still have to mention in your study that that happened, at the very beat you can say like this may have a 1/100 chance that you’ll get a blood clot from this, but thats much more accurate than if it happened in a one person study where 100% of the people involved developed a clot during the trial
When you want to make statements about a population, you have to do experiments that test against that population. One person does not represent a whole population. There is a lot that can be said about the general human condition from experiments on individuals, but without including many many more people, you can not know if it really is universal, or if it is a quirk of the subject.
If you want to make statements about a person then you can perform experiments on that one person and make general statements about that, but those are rare outside of clinical spaces.
Things affect people in different ways and it’s important to find out risk factors and things that can potentially make you more vulnerable to negative side affects (if it’s a study about that sort of thing) but basically everyone is different and they need a large sample size to represent as many people as possible
Every measurement has error in it. Sometimes we know how big the error could be. Sometimes we don’t. The error in biological measurements can be really big. There are differences in living organisms based on their genetics and the environment they were raised in. If we only do a few experiments, we don’t know how much of the effect was what we were looking for and how much was just random chance.
As an example, there are a small percentage of the populace that is immune to HIV. If we were doing a trial with only a few persons in it and got one of these people, it could throw the results way off.
>Why do scientific experiments (mainly in neuroscience and psychology) need to be done on more than 1 person? I was under the impression that the most important part of determining the study’s accuracy was to ensure that the outcome had a less than 5% chance of occurring without the theorised variable.
>Couldn’t a situation emerge where the outcome **was almost certainly attributed** to the variable in question even with one person. For example, something extremely random, like (stupid example) a blood clot forming in someone’s left pinky finger after being reminded of childhood trauma (and it was predicted beforehand).
How would you quantify this “almost certainty”? When you test a sample of 1 you will only ever have two possible results for anything: 100% and 0%.
Experiments require controls. You can do observational research on a single subject, but you could never ever do an experiment.
Let’s consider a silly, but precisely defined, experiment. I have a red pigment. I hypothesize that if I apply this red pigment to a person’s hand, their hand will be red. I take a single subject, apply the pigment to their hand, and measure how red their hand is (using some objective measure of, like, wavelength absorption).
From this, I can conclude that the subject’s hand was a certain amount of red after application of the pigment, but I can’t conclude anything else. I need something to *compare* my measurement to in order to determine that the pigment made their hand *more* red. The most straightforward way (in terms of the scientific method) is to compare the treated subject to a control subject who did not get the pigment.
Another way to obtain a point of comparison is to measure the subject’s hand bothe before and after application of the pigment. In this case I still have one person participating, but their past self is serving as the control, and their future self is treated. This is not ideal because there other things that may be changing over time that could affect what I’m trying to measure.
Either way, you need a point of comparison to perform the experiment, and once you’re comparing two things, you need to be able to characterize the probability that they are different by chance rather than due to the treatment. That’s where statistics and p-values come in.
When you perform an experiment you may not be able to account for all variables, like your blood clot example. These unaccounted for variables can be known as noise. Performing an experiment again at a later date allows the study to occur with different noise variables. If you can perform the experiment again successfully with different noise variables then you know what you were testing is successful and not just a fluke from an outside noise variable.
I want to diverge from some of the other answers slightly and say that research on small samples, or even a sample of 1, can be the most appropriate choice in some contexts.
The most obvious example is when researchers or practitioners come across unusual one-off outcomes. Sometimes, a doctor might see something really strange in a particular patient that they have never seen or heard of before. Or an engineer might be investigating a structure that has failed in an extremely unusual way. They will often write this up and publish it as a “case study”. This might be helpful to someone who encounters exactly the same situation in future, or someone might be able to combine it with other information to help understand some underlying mechanism.
Also, there is an advantage of small samples in that you can study the individuals in the sample in more detail. Suppose, for example, that you want to understand people’s political views. You might design a survey and send it out to a representative sample of 1000 people. You can be confident that your results will be reasonably representative of the population, but unfortunately it’s hard to ask anything but extremely superficial questions (“Which party do you support?”, “Do you support the death penatly?”, etc.) to such large numbers of people. Alternatively, you could conduct detailed interviews with 10 people. These results will not be representative of the population, and some of what you pick up might just be unique quirks of these specific people, but on the other hand you can ask lots of searching questions and get an in-depth understanding of how each individual thinks about politics.
An extreme example of that is when people are essentially studying themselves, which happens quite a lot in parts of the humanities. Philosophers will often spend a lot of time focusing on their own thought processes, because you can never understand other people’s thoughts as deeply as your own. Anthropologists will spend a lot of time thinking about their own interactions with the population they are studying. And so on.
Latest Answers