 All right everybody, welcome back. It's Veronica Howard. So this time around we're going to be talking about generalization. We're going to go through a very basic example of what generalization is. We're also going to be talking a little bit about applications of generalization. How do you actually foster some generalization? And we're going to finish up talking about those advanced clinical applications or times when generalization goes wrong. We'll also be talking a little bit about the basic research and some studies on generalization. But first, when we're trying to figure out what generalization is, we need to know what generalization isn't. So let's return to some of our previous content. Remember that discrimination training is all about narrowing the range of stimuli that a person will respond to. The formal definition of discrimination training, it's a procedure in which behavior is reinforced in the presence of one stimulus and extinguished in the presence of many other stimuli. So to translate this to English, we're teaching people where and when and with whom to emit the response and very often that behavior will only be reinforced in a very narrow range of circumstances. It will not be reinforced in others. An example of this would be, for instance, if I were to ask you, what's the stimulus on the left? What is this one over here? Now in the presence of that stimulus on the left, there's only one naming response that will contact reinforcement. That's Apple. In the presence of the stimulus on the right, if I were to say what's this and you were to say Apple, you would not contact reinforcement. So in this particular case, that one stimulus, in the presence of that stimulus, one response and one response only will be reinforced. In the presence of other stimuli, that same response will not contact reinforcement. In this case, that Apple is the SD. For the response to Apple, the orange is the S Delta. It will not be reinforced in the presence of that behavior. Apple will not be reinforced in the presence of that stimulus. Generalization, however, is exactly the opposite. The point of generalization is that the behavior is reinforced when it occurs in the presence of lots of stimuli. And often we see generalization as a phenomenon takes place when a behavior that we have trained under different circumstances occurs in a novel or new circumstance. We also have to understand what a stimulus class is. A stimulus class or stimuli that are related along some dimension, the color or the shape or some quality that they have in common. And remember a novel stimulus is a new stimulus. It's something that our learner has never seen before. So let's go back to our Apple orange example. In the presence of this stimulus, if I were to say what's this, right, this Apple is an SD for the response, Apple. And if you emit that response, you'll contact the reinforcer. But what if I present you with this new shape, this new item you've never seen before? Well, what's this? And if you emit that same response, Apple, you'll contact reinforcement. So in this case, remember, this will contact reinforcement. But what happens when we present someone with this new stimulus and they give us no response? Well, in this case, generalization has not occurred because they have not emitted the same response, Apple, in the presence of this green fruit. Let's give a call back to just general psychology. In mainstream general psychology, we would talk about this idea of a stimulus class. Now, these stimuli that I'm showing you on screen now all share a common feature. They're all part of the concept of Apple. And a concept is a mental grouping of similar objects or items. They have something in common. But remember that behavior analysts typically don't talk in terms of mental groupings because that assumes some sort of private event. We don't like to think in that language. And so rather than relying on a kind of mental construct to explain this, we talk about behavior in terms of stimulus classes. And a stimulus class is a set of stimuli that are related along some dimension that there's stimuli that have some common feature or property. So for instance, when I show you these flowers and this Pantone color and this fruit, they'll have something in common. They all share the common quality of being items that are the color orange. Or if I were to show you again the same fruit, and I were to show you this fruit, and I were to show you this fruit, they all share the common feature of being part of the citrus family. They all descended from three major basic citrus. They all share a kind of acidic quality. And the stimulus classes don't have to be simple like this. They can be a little bit more abstract. So for instance, if I were to show you this lemon, and I were to show you this aspirin, and I were to show you vinegar, they all share something in common. All of these items are acidic and all of them, because of their acidic nature, are used in cleaning or they can share some common features they could be used to dissolve certain items if you took them long enough. So stimulus classes can be topographical. They can focus on the form of an item. They can also focus on other characteristics like the function or what they do. What matters is those stimulus classes develop through experience. It's our previous history that tells us whether a stimulus falls into that class or doesn't fall into that class. Now this would ordinarily be the point in a face-to-face course where I actually talk about folks learning history with them. And you may be familiar with a couple of years ago. There was this big fervor. There was a big conversation about this song. It's outdated. It's outmoded now, but it's a song called Old Town Road. If you're not familiar with it, check out the optional reading. Essentially, the long story short of this particular song was that it became very, very popular on some radio stations. It started on TikTok and then moved to radio stations. People were requesting it. And it started climbing the charts on the Billboard Top 100 for country music. But then suddenly the song disappeared. And it was unclear why the song was removed from the country Billboard charts. Some folks said, well, it had to do with the qualities of the song itself. It had to do with the fact that there's a kind of underlying rap-esque or urban quality to it. And others looked at the fact that the song was removed from the country music charts as being evidence of a kind of a systemic racism. The idea that the person who had created this particular song was a black artist. And people thought, well, the reason it was removed from the charts was because you can't have a black artist on the country music charts. The question that we're asking here is, does this song as a stimulus deserve to be in the same stimulus class as other kinds of country music songs? So I might invite you here to take a moment, stop the video, and define what are the qualities that make country music country? What is it that country music has to have that it embodies that concept or that stimulus class of country? Is it a particular sound? Is it a particular type of artist? Is it a particular sort of format? Are there certain instruments that have to be present or must be absent? Does it have to have a particular type of narrative? For instance, do they have to talk about particular themes or tropes? What is it that makes country country? And does the song Old Town Road, remember there's a link to it in the description and in your study guide, does that song fit into the stimulus class of country? Why are we not? What's interesting about the scenario is it kind of embodies for us this idea of a generalization gradient. Now, when we're talking about generalization, we're talking about whether or not we respond in the same way to a novel stimuli as we do to ones we've been trained to in the past. And you can actually visually chart our probability of responding on this thing called a generalization gradient. The generalization gradient is a graph. It's a visual demonstration of the probability that we will respond to a novel stimulus based on the features that it shares with the stimulus that we have been trained to. So in this case, we're talking about probability indicated on the y-axis and along the x-axis. It's showing you how similar is the stimulus that you're being presented with to the stimulus you've been trained to. So this is kind of like a match to sample. How similar is this to the thing that you were trained to that's in the center? And as that stimulus changes slightly, in this case, we're talking about wavelengths of a noise. As that noise gets a little bit higher in pitch or a little bit lower in pitch, we may be less likely to select it. And then the more different it is, the less likely we are to select those stimulus features. I can give you another example. Let's imagine that we have a scenario here where our discriminative stimulus is an apple and we present different fruit. We might say, where are we likely to present? What are we likely to say? Shares those stimulus properties with the apple. And in this case, maybe you're trained on the red apple. And I present you with a green apple and you have maybe an 80% likelihood of selecting. Yeah, that's the same. I'll select that. And then I present you with an orange and you're 50-50 on that. Then I show you a banana and you're really unlikely to select that. And all the way down here on the right, I show you a chicken and ain't nobody going to say that a chicken is the same as an apple. So this is a visual demonstration of the probability that we're going to select a stimulus that's similar to the one that we've been trained to. It kind of shows us whether we're going to respond to a novel stimulus in the same way. Another way of looking at a generalization gradient could be, for instance, this. So if you've been trained to this kind of teal color in the middle, this shows you graphically the probability that you would endorse that the colors, these other colors, are that same color. So if I train you to the teal, you're very likely to select this kind of off-green color just to the right of it, but much less likely to say that orange is the same color as teal. Now generalization is a very basic feature of all animal behavior. For the most part, we need to have generalization. We need to have this in our repertoire. Like could you imagine a world in which we had to learn new skills in every single new environment that we went to? We would not be very adaptable creatures. So on some level, we generalize because this works for us because we are able to take our behavioral repertoires into new vistas, into new places, and actually contact reinforcement. We might see that our ancestors who had a repertoire that was able to generalize were those that were more successful in branching out from their homes and finding new places to live and contacting food and resources. So on some level, generalization allows us to work in settings or with people that we've never experienced before. It allows us to be very robust and adaptable, but that doesn't mean that generalization is always good. We can in fact over generalize. So for instance, if we're talking about the consumption of certain beverages, there are certain liquids that we would potentially consume that have a kind of light yellow color that wouldn't necessarily be bad. But there are also other liquids that have a kind of light yellow color that if we consume them, we would not enjoy the flavor. We might even get sick from it. So we see that discrimination, being able to tell the difference between the liquid on the left and the liquid on the right, that's a valuable repertoire as well. So remember that generalization is not always beneficial. You can have a phenomenon known as over generalization. And over generalization is where the behavior occurs in the presence of a novel stimulus where it will not be reinforced or potentially even punished. Okay, so remember that we talked about discrimination training. It's where we're trying to narrow the range of stimuli that a learner will respond to. We want them to respond in the presence of only a few stimuli, whereas generalization is exactly the opposite. We want to widen the range of stimuli people will respond to. We want to increase the range of stimuli where people are going to respond and give us that response so we can reinforce it. So it's just a question of, do you want to teach people to tell the difference between, or do you want to increase the likelihood that people are going to emit that response in the presence of new stimuli? Then you select your training procedure based on your goals. In discrimination training, we reinforce only in the presence of a narrow range of stimuli. And in generalization training, we reinforce in the presence of lots of different stimuli. So we're going to come next time and talk about generalization training. I'll see you guys next time.