 All right, so continuing where we left off this time around, we're going to talk about the highest and most rigorous level of scientific understanding experiments. If you want to know if you have a functional relation between two variables, you have to be able to do an experiment. When I say functional relation, I mean that one variable actually produces a change in the other variable. So when we say experiment, I'm going to be talking from a purely mainstream psychological perspective. An experiment is any deliberate alteration or manipulation of a variable in an environment to see if there's a corresponding change in behavior or some other phenomenon of interest. So if you are actually changing your independent variable to see if there's a change in the other deep-end variable, then you can actually determine if you have a cause and effect relationship. Experiments are a little bit like the Hokey Pokey, right? If you put it in and take it out, do you get differences when it goes in and when it comes out? So moving forward, in an experiment, you have an independent variable. It's that thing that the experimenter has control over. It's that thing that the environmental manipulation, it's the thing about the environment that you're changing. Let me give you an example here. So maybe we want to know if watching violent media makes people act more aggressively. In this particular case, the violent TV that we expose the participant to is the independent variable. It's the thing that we're either putting into place for them or it's the thing that we're not putting into place for them. So maybe someone is watching some other kind of TV. I'll come back to that in a moment. Next, the deep-end variable is whatever the phenomenon of interest is that we want to study. And so often in psychology, especially in behavior analysis, that's behavior. It's the thing that you're evaluating for a change as a function of that independent variable either being present or not being present. Now, there's a couple of different kinds of experiments. I'm going to go through just a couple of different variations here. Remember, we're talking about an experiment where we're evaluating aggressive play. So maybe I'm going to expose different clients to some violent television and then see whether or not they behave differently. My experimental group is going to get whatever my independent variable is. The experimental group is exposed to the independent variable that's being studied. You can actually have a control group that's going to receive a few different things. To begin with, a control group might be people who do not experience the independent variable. So maybe they get nothing rather than watching TV. Maybe they read a book, so no TV at all. They might receive a placebo, which for instance, if I'm saying that the violent TV is my experimental variable or my independent variable, then maybe I want to expose them to some TV. But TV that I know is not violent. So maybe we're going to let them watch Dora the Explorer. You can also have a control group who gets exposed to whatever the treatment, the best practice treatment of the time is. But essentially what we do is after the experiment, we compare the experimental group to the control group. We look to see if there's any difference between the group. If there's a difference between the group, we know that the only thing that was different between them, fingers crossed, if you had very good experimental design, the only thing that was different between them was the experimental variable. So we compare the experimental group to the control group and any difference between those groups is presumed to be a function of the independent variable. So this is how a lot of mainstream psychological studies are conducted. It's important, though, for you to use a between groups design, you absolutely have to have something called random assignment. And random assignment means that everyone in your participant pool had an equal chance of being in either the experimental group or the control group. So to give you an example, if I in this class said, who wants to come participate in a study with me? And then about half of the class said, ooh, ooh, I want to participate in a study. And then I run the experiment, the people who didn't volunteer are in the control group, the people who volunteered are in the experimental group. When I get to the end of the study, I can't really compare the experimental group to the control group because not everybody in there had an equal probability of being in either group. It's possible that the differences that I see between those groups are due to the independent variable, but it's also possible that the difference is due to something else, a confounding variable. Something like the folks who volunteered to participate may have wanted to make me happy, which means maybe they worked harder, or maybe there was some incentive for them to participate, like they really needed extra credit, and that's why they participated. So you have to have that random assignment to as much as possible rule out confounding variables that could account for why there's a difference between those groups. If you don't have random assignment, you cannot compare those two groups to one another. And this is probably the most important part about the between groups experimental design is making sure you have that random assignment. However, it's really important to know that being random doesn't necessarily mean that those groups are going to be perfectly equal. So notice what I'm looking at here is the group on the left in the experimental group of the seven little smiley faces, three of them are gray, four of them are yellow. In my control group here that I'm showing you on the right, four of them are yellow, three of them are gray. So that's kind of perfect back and forth. But in this particular, I'm showing you on screen a different set of groups. In the left group, the experimental group, six of the smiley faces are yellow and one is gray. And on the right, I'm showing you the control group. And in that group, six of them are gray and one of them is yellow. Does this mean that there was no random assignment? No, and it's important to note that random doesn't necessarily mean equal. And so just because everyone had an equal probability of being in either group doesn't mean that I'm going to have groups that are perfectly matched to one another. Sometimes you're going to get a lot of yellow folks over here and a lot of gray folks on the other side. You just had to have had an equal probability of being in either group. And you do want to, as much as possible, make sure that you're not strategically placing certain participants in different groups because that's going to introduce a compounding variable. Now, there are also a couple of other reasons why a between groups design can be weak. So first, we have this idea of the placebo effect. And that's a pretty well known phenomenon. A lot of people know this already, but it's this idea that your expectations that the participants have in the study can influence their behavior. So we know in America, especially that American research participants are much more likely to report the phenomenon that they think is going to happen than participants in any other country. The placebo effect essentially means that your behavior changes, even though you weren't actually exposed to an independent variable. The other problem that we can have in a between groups design is this idea of an observer bias, which means that when you tell a researcher to look for a particular thing, they may be more likely to see that and they could unintentionally influence the results of the study. So it could be that their observations themselves were also biased. There are a couple of ways to get around this. So in a single blind study, for instance, we wouldn't tell participants which group that they were in. The researcher would know, but the participants themselves would not know if they were receiving the active treatment or a control treatment. In a double blind study, which is like the gold standard, a randomized clinical double blind study is the highest level, the most rigorous, the most amazing, because not only does the participant not know which experimental group they're in, but the researchers, the observers, they also don't know. So that controls for both placebo effect and observer bias. So these are just the basics of understanding very traditional mainstream psychological experiments. When we come back next time, we're going to be talking about behavior analysis specific types of studies, the single subject or time series research design. I'll see you guys next time.