Okay, In this video, we're going to talk about blinding as an aspect of experimental design. Before we do that, I want to tell you about a famous study that was conducted in the early 1960s by Rosenthal and food. This experiment was performed by 12 graduate students. And those 12 graduate students were given two sets of rats. It was explained to the graduate students that the rats belongs to either a maze bright treatment or a maze Adult Treatment. And it was explained to the students at the difference between these groups was that the maze bright individuals had been selected to have particularly good abilities at solving mazes. Whereas the rats in the maze dole treatment had been bred to be particularly poor and solving mazes. And here's an example, maize that the rats encountered in this experiment. Okay? So the students were told about the breeding history of these rats. Students were also told which type of rat they were given. And then the task given to the students was to basically assess whether or not there really was a difference between the maze bright and the maze dull individuals as they performed in this maze. And the students had five days to assess the behavior or the performance of these two different groups of rats in, in mazes like this, okay, and here are the results. So we have data for five different days. And here the numbers that represents the rats performance and higher numbers are associated with better performance. And there are two things that we can see here. First of all, you can see that on average, the maze bright rats tend to have higher performance than the maze dull individuals, which is what we'd expect. And these differences are supported by statistical analyses which are reported in the original study. That's the statistics is not really what interests us here. And the other thing to notice is that if you look at the rate of change in performance over days, it seems that there is a trend that the maze bright individuals increase in performance faster than the maze dull individuals do. And that might be interpreted as evidence that the maze bright individuals were learning faster than the maze dull individuals. Now that's pretty cool. But that's not why the studies famous. This study is famous because this experiment, it was not actually performed on rats. This experiment was actually performed on the graduate students. And that's because in actuality, there was no difference between the rats. What actually happened was there was a group of rats that had been set aside for use in this experiment. And the rats were chosen from the group. Completely at random. And they were randomly allocated to this maze bright and maze dull groups. There had been no actual breeding of the rats, even though the students were told that the rats had experienced different breeding regimes. Okay? So the truth is that there was no difference between the rats for these two different groups. That raises the question of how did these results arise? Well, that is what's really so interesting from this study. Because what this study really hope helps to show when this is the first study to address this is it demonstrates how our expectations for the results of an experiment can actually lead to differences between the groups that we are comparing. In other words, just stay demonstrates how our expectations can influence the outcome of the experiment. How could that happen in this study? Well, there's at least two ways the study wasn't really designed to figure out exactly why these types of difference arose. That's not really the point. The point the studies to show that expectations can affect differences. But there are two potential differences, are two potential reasons for why these results could arise. At least. One mechanism could be that the students treated the mikes differently. So for example, the students might have treated the maze dull rats in a way where they might have caused the rats more stress. And as a result, the rats might have performed differently in the maze than the rats that might have been treated better. Okay? So in other words, it could be the way in which the students actually interacted with the subjects that could influence the outcome. Alternatively, bias could arise in the study when their students were actually collecting their data. So at the stage of the experiment, where the students had to assess how well the rats are performing that assessment, it might have been tainted or biased by the students prior expectations. This study really emphasizes the need for what's called blinding in experiments. And blinding involves hiding or concealing the treatment identity to which subjects are assigned from the researchers that are conducting that experiment. We can also have double-blind studies where not only are the researchers unaware of which treatments subjects had been allocated to, but the subjects themselves are also want aware of which treatment they've been allocated to. And this is particularly important in studies of humans. So we're humans are the subjects. Blinding occurs to various degrees among disciplines within biology. So within biomedical science, blinding occurs relatively infrequently. So it's not common. Even though there's strong reason to included, as we're going to discuss in some. And our next slides in ecology and evolution. Blinding is very rare. Even know, as we're going to see, it can be very wise to include blinding in our experimental designs. In some cases it may be impossible to include blinding and a study. For example, if your goal was to compare the behavior of female versus male deer in their natural habitat. Then the process of doing that might involve your Heidi and some bushes with some binoculars and looking out at deer. And you would need to know whether or not the deal you're looking at it was female or male in order to record your data appropriately. So in contexts like that, blinding is impossible. But in many other contexts, blinding is possible and can be a very wise components to include an experimental design. We can incorporate blinding at a number of different stages of our experiment. We can include blinding when the subjects are being allocated to their treatments. We can include blinding while the studies actually progressing. In other words, well our subjects are growing or well, they're performing some particular task. We can also include blinding at the time at which we're actually collecting our final numbers. So this might be when we're actually counting the number of times something happens, counting the number of times that a certain behavior occurs, for example. Okay? And finally, we can include a blind you during the stage of data analysis. This is pretty rare. So this is something that's not done very often, but it can be very wise. I'll explain this a little bit more. Imagine that we were conducting a study that was comparing the effect of a drug versus a placebo experiment, or an experiment that had a drug and a placebo treatment. While we go about analyzing our data, the person analyzing those data might have some preconceived notions about how the experiment might turn out or how they might want the experiments turn out. And data analysis involves a series of decisions about how we would handle those data. And if we know in advance or if we can tell by looking at our data as we're analyzing it, which data points belong to, which treatments, then our interaction with the data in that way. So our interaction with the data, knowing how the results are turning out could influence decisions that we make while we're going through the data analysis process. And as a result, our knowing about which data points belong to, which treatments could influence how we analyze the data naturally, how the results turn out. We could include blinding in the data analysis by simply using ambiguous codes for our different treatments. So instead of calling something drug treatment and a placebo treatment, we could just call a treatment 1 and treatment 2. So that when we're analyzing that data, we would not know which data are sorry, which treatment our data belong to. To implement blinding in a study like this requires help from others. In particular, we would need to enlist the help of a colleague who could maintain a file that would allow us to remember which subjects belong to which treatment. Remember we want to hide. We want to hide from the researchers which treatments the various subjects had been assigned to. So the way in which we can so we could do that by assigning a particular code number to each of our subjects. And then we can have a file that keeps track of how that code number is matched up with a particular treatment to that subject is, has been allocated to. Okay? So blinding does require help from others, ideally from people who are not directly involved in the experiment except for this aspect of managing the blindness of the study. So it does require more work. But that work can be very worthwhile. As we're going to see a second. There have been a number of studies that have assessed how blinding or a lack of blinding can influence results. So I'm going to show you a series of studies that have analyzed results from papers that did use blinding and that did not use blinding. And by comparing those two types of experimental designs, we can determine whether or not blinding itself influences the outcome. In this first study, which was a study of review studies or systematic reviews, they found mixed evidence for the effects of blinding. I won't go into the details. I'll just say that whether or not they found evidence for blinding affecting the outcome, really dependent on how the analyze the data. So we don't have a very clear picture in this case. In this study which examined which examined other studies looking at multiple sclerosis, they found that there was a big difference in the effect size depending on whether or not the study used blinding or not. So what I want you to see here is here's the data summarizing the results from studies that did use blinding. And here's the outcome of a variety of studies that did not use blinding. And there is strong statistical evidence that these two groups differ from one another. And let's just look at the size, this difference. So the average effect size for the studies that did use blinding was around 130 percent. Whereas the average effect size for those that did not use blinding was around 40 percent. That means that when we do not include a blind in these studies, the effect size increases by a third. That's a huge change in the effect size. So not including blinding can greatly exaggerate the effects that we find. And that's not something that we want to be doing in science. We don't want to be exaggerating that results. We want to be reporting the correct results. I've just summarized the results from this study here where again, they were comparing studies at data to not use blinding. And they found that studies that did not use blinding were 1.7 to 7.7 times more likely to report a positive outcome, then studies that did use blinding. In other words, if a study did not include blinding, they were much more likely to report a positive outcome from their experiment. And I'll just say that these points here, I believe represent the endpoints of a 95 percent confidence interval, which is why I've given a range here. If I remember, the average difference was about five times. So the point here is once again, and we find that if we do not include blinding, we can be much more likely to get positive results. Add in this last study, this is one that looks at studies where humans are the subjects. And what they found was that studies that did not include blinding exaggerated the effect size by about 36 percent, which is really consistent with what we saw on the multiple sclerosis studies. So it's going to wrap up this video here. The point here is that blinding is highly recommended as a component of your experimental design. Wherever it's possible to include it. It involves more work. But including blinding can help us to have much greater confidence in the results so we do obtain, and that's what we want in science. And finally, just as an added note, blind IAM requires teamwork. And so that is more work. But once again, I'd argue that typically that additional amount of work will more than pay itself off in terms of the additional confidence that we can have in the outcome of the experiment. If you want to learn more about blinding and how to report, including blind getting your study. This is a great paper just to, to go to. You can also use this paper just to find some more sources of reading on this topic if you're interested. I'll end the video there and I'll say, thank you very much. And I hope this video has been helpful.