PAM WOLFE: Behaviorism-- you've probably heard that term. It refers to the philosophy of the science of behavior. So there's a couple things there. We have a philosophy, but it's backed by science. So we're really saying that science includes behavior, and that is pretty revolutionary. For a long time, the hard sciences, physics and chemistry and those, had tangibles to look at. Skinner, along with others, really said that behavior is a science. And we can apply a philosophy or a theory to it, and make sure we're looking at it scientifically. So the study of behavior is a natural science.

My father was a genetics professor, and he used to really just truly marvel at how can you apply science? What do you do for your research? How can look at people and behavior as a science? But this philosophy says that behavior is a natural science. And critically to this, behavior is controlled by environmental influences. What goes on in the environment, particularly consequences, really control how we react to things.

A little bit about the terminology that you'll hear-- the term "behavior modification" is a bit of a past term that you might have heard of, and comes, generally, from the field of psychology. So behaviorism came from the field of psychology, and a term that was often used was "behavior modification." That has a bit of a negative connotation, because it sounded like that we're robots, that we're putting people in rooms and shocking them, that we modified their behavior.

This is an early term. The term "applied behavior analysis" is the more preferred term now, because this was established as a science later. If you choose to get the Miltenburger text, you're going to notice that they use these terms "behavior modification" and "applied behavior analysis" pretty interchangeably. But I want you to understand that the term behavior modification is not often used. It has a lot of negative connotations that go with that.

And when we're looking at the science of behavior, we're looking at, generally, two types of behavior. The first is what's called respondent-- often called reflexes-- and the other is called operant-- so respondent and operant behavior or learning.

Respondent behavior-- this is our first that looks at reflexes. Behavior that it's elicited by a stimulus-- the presentation of the stimulus evokes a behavior. So you see on our slide here a person sneezing or a person blinking. These are reflexes that occur. Those are called respondent behavior. If our physician uses a little rubber hammer and hits our knee, we have a reflex. You'll notice that young babies have what they call a rooting reflex. If they see a nipple, they'll turn their head toward it. That's because they're hoping to get fed.

So those are all respondent behaviors. They happen organically. They happen as reflexes, and very often-- most often-- cannot be controlled.

Now, respondent learning really came from the background of Pavlov. It's often called classical conditioning or Pavlovian conditioning. And this is the work of Ivan Pavlov. And really, what we've got here is a respondent reflex. So that reflex is elicited by a previously neutral stimulus. Once we do that, we have learning. So if we look at Pavlov's work, and this was often called stimulus response, it showed that you could pair two stimuli together and bring a response under the control of a new stimulus.

So looking at some of his work, you've probably heard of the salivating dog. He paired a bell with the presentation of food to a dog. And he showed that the dogs would, after he's paired that bell and the food together, salivate in response to the bell. The bell took on a conditioned response for the animals. They only had to hear the bell, and they would begin to salivate. It didn't take the presentation of the food. They knew that the food was often paired with that. So you really saw a stimulus and a response. Very classic and very important to our learning or understanding of behavior.

So we have respondent behavior-- now we have operant behavior, our second type of behavior. Operant refers to behavior that operates-- that operates on the environment and is changed by the environment. So this means that we have environmental influences that change our behavior. And this analysis was pioneered by BF Skinner, and you'll hear a lot about him. He really revolutionize the science of behavior.

What operant learning says is that learning new behaviors is the result of the relationship between the behavior and consequences. So what happens after a behavior makes a big difference in shaping or changing or maintaining a behavior. We look at the consequences-- behavior and what happens after.

And if we're looking at the area of education, almost all of the behavior of interest in education is that of operant behavior. We don't simply want to look at reflexes. We want to be able to bring behavior and change it. And so we look at operant behavior. Because we're able to change the consequences, thereby change behavior.

If we're looking at operant behavior, we really have what's called a three term contingency-- new behavior that results from the relationship between the antecedents, the behaviors, and the consequences. So you see there a diagram-- A, antecedents, B, behavior, C, consequences. If we're looking at antecedents, we're looking at what happens before-- what came before the behavior, the behavior itself, and the consequences.

So if I harken back to my days of teaching, long ago, I had a student who really hated fire drills. The antecedent would have been hearing the bell. The behavior was grabbing teacher hair. And the consequence for that was that he was put in a restraint and removed as quickly as he could for a fire drill. So we have an antecedent, a behavior, and a consequence. I knew directly that once he heard that bell, I would see that behavior.

I begged my principal to let me know in advance so that I could make sure I modified the environment for my student, but she said, no, no, it has to be authentic. So I remember, we had aides that would try to go down in the hall in the office and say, I think there's a fire drill coming, so let's get ready. So we tried to get around the system. So we knew the antecedent.

Now, operant behavior and applied behavior analysis goes under the broad umbrella of experimental analysis of behavior. And this is the science of studying behavior systematically. So the key term in there is systematically. We're going to make sure we control things so we can see what happens. It involves a controlled experimentation in laboratory settings, where those variables are strictly monitored and manipulated. So it involves experimentation, and typically this occurs in laboratory settings.

So we can watch people in settings that are a highly contrived and see what they do. There are classic experiments in the field of psychology where individuals-- one might favorites, what I really fascinating, was someone was being subjected to pain. And the person who was in the research room got to decide whether more and more pain was delivered to the individual.

They heard in the other room somebody screaming. It wasn't actually occurring, but would the person continue to press the lever when they were told, when they knew that someone was screaming in the next room? How much pain would they inflict before they were unable to do that anymore? Amazingly, many people did apply a lot of pain, because they were given the command to do so-- very interesting. But that was done in a laboratory, so it could really control exactly what occurred with the individual-- experimental analysis of behavior.

If we look at applied behavior analysis, or ABA, as we call it, with this, we have the practical, real-world application. So that distinguishes this, ABA, from that of experimental analysis. We're looking at practical, real-world application, and using those behavioral principles to produce change in behaviors that we want to change that maybe affect people.

Now, a little caveat here-- there may not only be problematic behaviors. We may want behaviors that increase or decrease. But the key I want you to remember with this it is that it's carried out in natural settings. So this is where we have the real world-- the rubber hits the road. This is where we are in classrooms, seeing if we can change behavior. And this is where, really, education is highly effective.

Because we need to know if things happen in the real world, in the natural setting, when we're in the preschool, when we're in the elementary school. Things change if we're in a laboratory and we can control it. Things are different in the real world. A fire drill may go off, and we don't expect it. There may be kids absent that day, and it might affect the behaviors of others. So applied behavior analysis is really setting it in the setting where it will be used, and seeing if we can change things to affect behavior.

Let's go through some of the dimensions of applied behavior analysis, things that make it a science. And we'll go through each of these, and then we'll go through some explanations. The first is it is applied, analytical, behavioral, technological, conceptually systematic, it's effective, and it has some generality. So let's go through some of these dimensions

The first is that it's applied. What this means-- those outcomes that we get-- sometimes called dependent variables-- have practical significance. They mean something. They make someone's life better as a result of this intervention. So we're not going to do an intervention for intervention's sake. We're going to make sure that the outcome have some significance. Yes, the child learned this skill. Yes, the behavior decreased. He was not tantruming or acting out aggressively as much. So the outcome has practical significance.

If it's not practical, if it doesn't make sense, it really isn't an effective applied behavior analysis technique. If I'm trying to get a student to speed up to a certain level and they're not able to, there's no practical significance if they're not able to participate in the activity.

Behavioral, our second-- behavioral is the dependent variable of interest. So when we're looking at an independent and dependent variables that you'll often hear in science, behavior is the dependent variable. That's what we want to measure and see if it changes. We need observable behavior. We need that observable behavior so we can make sure it accurately and reliably.

If I say, yeah, he's tantruming, I need to know exactly what a tantrum looks like. I've got to be able to say it's when the student throws themselves on the floor, when there is screaming, when there is crying. I may be even more specific, and we'll get into that a little later. But I have to be able to do that to count it.

So if I'm saying a child is banging their head, you may think of it as actually banging on the table. I may think of it as this. We've got to decide which one are we going to count, or are we going to count both? What constitutes head-hanging? So you've got to be able to measure it, and it's got to be reliable. You and I both say, yeah, that was head-banging. So one must be able to see and objectively define that behavior of interest-- so it's behavioral.

Third dimension-- analytical. Here, we have the demonstrated control of the variables so that we have what we call a functional relationship. So we can say that because of this intervention, the student's behavior decreased-- a functional relationship. They go together. If I took out the intervention, the behavior came back. I've really established I've got a functional relationship. When the intervention's there, the behavior goes down. When the invention's out, it does not occur.

So I've really got what we call control of my experiment. I've been able to show a change. This process removes what we call threats to internal validity. If we're looking at science-- and this is a little beyond our scope here, but-- if we're really wanting to see if our intervention is analytical, if it's effective, if it really did have a functional relationship, we've got to remove the threats to it.

If I really want to see that a behavior changed, I've got to reduce the fact that the child is just getting older-- maturation. I've got to remove the threat that there might have been some outside effect that made a difference. There was a tornado that day, and that's going to change how my intervention is going-- all sorts of things that could change it. We really want to show control, that, yep, when this happens the behavior changes as a result of it.

Technological-- and this is also really important in behavioral as science. We've got to make sure that the procedures that we use our precisely described so that we can replicate. Replicate means that we can do it again and again and again with the same results.

Early on, we had Ivar Lovaas, who had some great, great results with kids with autism. And a lot of people said, man, these are fabulous. People tried to replicate that, and they could not replicate his results. They did not get the same miraculous change. He got a lot of criticism for that.

One way to make sure things are replicable is making sure that it's precisely described, so I can implement it, and you can implement it, and we're doing it the same way, right? So when I say the child received positive reinforcement, you and I both know that it was given after every third response, that the reinforcement was directly tied to the behavior-- nice job, Tim, on putting that ball back. So we make sure that the procedures are technological. We can replicate them.

Conceptually systematic-- all these procedures are derived from basic principles and they're linked in a systematic manner. This says that it all-- I like to think of it-- hangs together. So it's not some random thing where we have a theory here and a theory there. We have a theory. From that theory, we have basic principles. And from those principles, we have procedures that we can use.

So behavior analysis is not what they sometimes call a bag of tricks. Sometime you'll hear a therapist or teacher say, I've got a bag of tricks. Well, that can be a negative in some cases, because you really want to set of procedures that's linked to a theory you're not just pulling out of the air. But you know that reinforcement means one thing, and if you apply reinforcement, this happens. So you want it to be linked conceptually.

That it's effective-- we've talked about the fact that it has to have some real world significance, that the overall benefits to participants sets the standard for effectiveness. Does it really make a difference? Treatments that lack social or practical value are not effective. So you can work on something till the cows come home, but if it really doesn't make a difference to the individual or to their functioning in life or the classroom, you really haven't been effective.

And the last-- generality-- results and behavior changes that will, over time, appear in other environments, or what we call spread to other behaviors. So we teach things, and the student is able to generalize the behavior to other instances. I teach my students to tell time on a clock in my classroom. But my student can generalize that time-telling skill in other environments. They can do it in the cafeteria. They can do it on the wall clock at home. It's generalized. So we've got to make sure that a behavior isn't done only in one environment, but that it spread or changes. And generality can refer to settings, can refer to other behaviors, can refer to personnel-- lots of things.

One of the things that I noticed in my teaching case is that when I consulted with other teachers, they would say, yeah, he does this for me. He'll do this for me, but he won't for anyone else. That's a problem. We want that behavior to occur with other people and other settings. I worked in a group home, and one of my favorite guys was a guy named Richard, who had very low cognitive functioning. And he would sometimes get very oppositional. And he would show that by refusing to get off the van when he came home from his work. So he would just sit there.

And any time I went out and got him I'd say, Richard, inside now. And he'd go, OK. He'd come in. But for anyone else, he would sit out there. That was problem. We had to really work on the fact that people could get him off the van, because he couldn't sit there all evening. So we had to make sure that the behavior generalized to other people-- not just me, other people.

So those were our different components of applied behavior analysis. Let's look at the, then, some characteristics that make up what we call behaviorism. This will be a bit of a recap of what we looked at. We focus on behavior. The procedures are based on science. We have an emphasis on current environmental event. We've precisely described the procedures-- technological.

It can be implemented by people in everyday life-- doesn't have to happen in a laboratory. It measures behavior change. It de-emphasizes past events as the cost of behavior. it's not that we have some deep-seated psychological problem. We're really looking at currently what's happening with the behavior.

And along with that, we have the rejection of hypothetical underlying causes of behavior. He's doing it because of his inner child. He's never gone to his inner child. That's something we cannot measure. That's something we cannot look at. So we really reject those hypothetical constructs that often go with the underlying the cause of it. He's feeling abandonment-- eh, we can't really look at that. We can look at the events that we think might make up abandonment. But we can't look in the individual and decide that abandonment issue is what's causing a behavior. So behaviorism.

Now, we've talked a little bit about BF Skinner, and there is the man himself. And as I've mentioned to you, he really looked at behavior as a science. And in his day, he was pretty revolutionary. Individuals said, oh, man needs to have free will. We're not really changed by our consequences. We can think and reason.

People misunderstood a lot of what Skinner said. He said, yes, we do have rational thoughts. Yes, we can. But we are all controlled, or we are all under the influence of the consequences of our behavior. So you work for a paycheck. It's very rare when somebody works for free. I work for my paycheck. I may love what I do, but I still expect my paycheck. I'm bringing that under operant control, the consequences of that.

What you're going to see is a video clip of some of the work of BF Skinner. And you've probably heard of his work that's done with pigeons and with rats and another animals. And this was a problem, because people said, I'm not the same as an animal. I think and, again, I have rational thoughts. Yes, you do. But those same principles that are applied to animals can be applied to human beings as well. There's that basic principle, or theory, or philosophy, that overrides that, that services as its umbrella.

You'll see in the video some pigeons, and you'll see operant conditioning in action. You'll see that Skinner is using shaping to change the behavior of the pigeons. You'll notice that he uses food as a reward for this. And he keeps the pigeons a little hungry, so that they're not satiated, that they'll want the reinforcer of the food, and will perform the behavior. So pretty fascinating video clip for you to look at.

We've indicated that we have, with behaviorism, or applied behavior analysis, empiricism. We have science that goes behind it. Popularity should not be mistaken for proof. Enthusiasm is not a substitute for evidence. What you have often, particularly, I find, in the field of autism, is exciting things going, but we've got the cart before the horse. Everybody's excited about something, and they put it out there, and it's before we've really looked at the proof of it. Do we really know it works? Are we so desperate for something that we'll try anything without the evidence, without empiricism?

So we really want to keep our interventions under control and looking at evidence, using empiricism. What you will hear is that applied behavior analysis has a very empirical evidence base. It has lots of evidence for it. But there are so many things, if you get out on the internet, that it's, in my mind, a little of unethical. There are so many things-- try this new injection, try this new pill, and this will work. This has not been proven. We need to step back, and we need to make sure we have empirical evidence before we move forward.

We look at empiricism. It's a practice that involves objective observation. So we're going to try to make sure that others can look at the behavior, see the same thing, that we've observed it, and we see it with our eyes. It's not something we intuitively think. Because that can really fool us. We have to see it in an objective fashion. And the data must be objective, reliable, and quantitative. So it can't be something that, yeah, I noticed an overall increase. It seems like he's getting better. I've got to be able to quantify that. No, his data jumped from three instances to seven. So I've quantified it.

And now I'm going to get on my high horse again and talk about evaluating autism treatments. There are so many treatment out there. And I think it's unethical, sometimes, for parents or others who are really desperate for things like a cure, thinking that autism can be cured, that they're willing to try anything. But there are agencies that are coming up now that are really taking time to look at different interventions and seeing whether or not they are promising or effective. So lots of projects-- common ones that are well known in New Jersey, New York. And we have some others we'll look at.

So when we look at these agencies as a whole and the research they've done and the synthesis of the information, really what we find is that effective interventions are more similar than different. They're not way out in left field. Many of the really effective interventions use applied behavior analysis and incidental teaching. They have a strong structure. They have ongoing monitoring and data collection. So they're more similar than different. But applied behavior analysis is a philosophy that goes along with many of them.

Truth be told, many teachers and parents and others use a combination of interventions and approaches when working with individuals with ASD. It's very rare that you'll have only one thing going on. But it's important that any of the interventions or procedures that are used have an empirical basis.

What are evidence based practices? Well, they're scientifically based. They've been found effective through replicated research. That's that technological-- it's been replicated. I can do it, you can do it, somebody in California can do it, and we will get similar results. It's got a substantial amount of rigorous research. And by rigor, we mean that it's well controlled. It's not just one data point taken. There are many. It may be done with multiple students. So we have amount of rigor. We've been able to use some systematic control of that.

And evidence supports that increases in skill acquisition are a result of that intervention. So we've got that control. If I've applied the intervention, the behavior will change as a result of it.

Now, there has been some legislative and research influences that have really changed the face of autism research and how we look at autism. You have Individuals with Disabilities Education Act, IDEA, that says that we really only can use empirically evidence practices. So if you have someone who's saying, I want dolphin therapy with my child, and I want a part of the child's education plan, under the schools, that does not have empirical evidence. Aside from being logistically impossible, it does not have empiricism behind it.

You have the National Research Council, the NRC, that we'll be looking at here, and the Nationals Autism Standards Project. All of these agencies and legislation have said, stop, we need empiricism before things are used.

Now, the National Research Council, in 2001, said that research should be able to demonstrate what we call a causal relationship between an intervention and any changes that occur in the development, behavior, social relationships, or normative life circumstances. We have our initial behavior. When we apply our intervention, we have a change in behavior.

Now, there's something I want you look at on this slide. Research should be able to demonstrate a causal relationship. Really, if we're looking in terms of true science, it is very rare that you get a strict causal relationship. More commonly, we say that there is a correlational relationship-- the two go together. But I can't in all good conscience say that it's because of only this. So that's language that's a little stronger than we typically use in science. But we want to have a situation when our intervention's applied, we have a change, and the change is in the direction that we want it to be.

A little bit about our research designs-- and this is something that I'm not going to go into depth in our course here. But very often in applied behavior analysis you have single subject designs. These are different than what we call group designed. Group designs tend to be larger, have many, many participants, and use statistics as a way to approve or disprove an intervention.

With single subject designs, we can use within subjects or single cases to look at our interventions. So I might have three students. And I might give my students an intervention at different points in time. If my students' behavior changes in the way I'd like it to be, I have some evidence that it is effective for this student, that there may be a correlation with it.

Single subject is very different in that it does not apply statistics only. It can use some statistics, overlapping data points and such. But typically, it's looking at the graphing of the behavior to see, through visual inspection or analysis, whether or not we're seeing a change. We're not applying experimental statistical evidence with this. That's different.

Single subject is often used with individuals with autism simply because they're not large groups of individuals, and it's such a heterogeneous group, very, very different in the characteristics that are displayed by the individuals. So single subject is often used in research designs in ASD.

Now, we've talked a little bit about the evidence, and that there are many, many different kinds of therapies or interventions that are out. Simpson, in 2005, put out a variety of categories that looked at some of the interventions that are used. You see his categories of scientifically based, promising, have limited support, or not recommended.

Let's look at the scientifically based. Applied behavior analysis comes up there. Discrete trial training, often thought of as synonymous, but within applied behavior analysis. Pivotal response training-- also part of the science of behavior.

Promising practices-- they're starting to show some promise, but more studies typically need to be done. Play-oriented picture exchange systems, and AAC. I find it hard to believe that assistive technology and AAC isn't scientifically based, but I believe that it's probably due to lack of systematic control of some of the studies. It's very commonly used. PECS-- also very commonly used, and often thought of as having a strong research base. More research needs to be done in PECS to see if it moves over to the scientifically based.

We've got our limited support-- animal therapy, art therapy, music therapy. A system called power cards. Auditory integration training, where the individual listens to tapes and it's thought to be affect behavior.

Then we've got our last category of not recommended. and there are very few that Simpson uses that have this label of not recommended. These have to show actual harm to the individual. Not recommended are holding therapies. That's where, you might have heard of it in the past, kids, particularly with attachment difficulties, they would just hold the individual, holding them tight, to help the individual attach to someone. We think of kids with autism as being aloof, not attaching, not being warm or hugging or loving. So this therapy just held on tight.

Not good results with it-- it was not recommended. There was a death-- one that I know of-- associated where a kid was resisting. His head hit the bathtub and he died. So that was not recommended. It's something that has been used, and I think it's still used in some circles. But it's not recommended for ASD.

And then facilitated communication, or FC. And you may have heard of that. That's still very ongoing, very heavily used at Syracuse University. Doug Biklen pioneered that in the United States, saying that this was a way to support individuals in their communication.

And he found, in his eyes, that individuals needed to just have their communication unlocked. And if they were given an alphabet board or another way to communicate, and someone facilitated it, or help them move their hand, that they could be able to communicate great things, and that it really unlocked IQ and all of these myths that individuals with ASD had cognitive disabilities.

What they found was that this was totally rebuffed by scientific research. They did controlled studies where they were double blind. And they had on the individual looking at something and the facilitator looking at something, and they typed or held different things. They were not the same. So science rebuffed that totally. But you'll still see facilitated communication being used. And, again, I think it's a disservice to use things that are not empirically based. So facilitated communication-- pretty controversial.

Looking at characteristics of our quality researcher, our empiricism-- those are inventions that are scientifically validated through controls analyses. We've got careful selection and assessment of our participants. They are alike enough that we can make an assessment. They are selected based on certain characteristics that will allow us to test our interventions.

And we thoroughly assess them. We know their cognitive abilities. We know their communication abilities. So we don't have an apple and orange analogy, where we've really got a disparate group. We've got objective, accurate, and reliable measurements of the effects of that intervention.

And lastly, that they're published in peer-reviewed scientific journals. Peer reviewed simply means that I may get an article to review, and I don't get to see the author's name. That means that I'm looking at it as I would look at everyone else's, very objectively and fairly. I'm not saying, oh, this was done by so and so. She or he's a very big name in the field. I'm going to give this one a go.

Peer reviewed means it's blind. I don't know who it is. So I'm supposedly getting, or I believe I'm getting, a more of objective assessment of my results. So we want them to be peer-reviewed and published in scientific journals.

Now, people say, oh, you know, it doesn't hurt to do something. Just give it a try. But really, if you use interventions that have limited support, you are undergoing some risk. You have wasted money, wasted time. You've wasted energy. And this is really, really difficult. Time is so important. You can't get it back. If you spend five months on an intervention that has no empirical basis, you've missed that amount of time that you could be doing something that has empirical basis.

You have exploitation of vulnerable people. A big group of this is that you have families that are very desperate-- I'll do anything for my child. Physical and emotional injuries that come along with it. And then lastly, they lost opportunities to make a real advances. We've just lost the time. We could have been using that more effectively. It's wasted. So it's not enough to say, eh, it won't hurt to try it. Use empiricism to choose your interventions and practices very, very carefully, because it does make a difference.

Looking applied behavior analysis, again, there are so many procedures and many principles and interventions that are used. But ABA, applied behavior analysis, is the only approach that's approved by the Surgeon General of the United States. It has the strongest empirical evidence at this time. And that's why we're giving you such a good, heavy dose of this.

Now you have some time to complete a practice quiz that's located below your video.

INTERVIEWER: Can pigeons read? This one gives every indication, because he's been taught to distinguish between two words, and to behave appropriately. He's learned his different response to each sign by being rewarded with food. So the bird isn't acting independently. Its behavior is shaped by controlling its environment.

The first task was to isolate an individual piece of behavior, and see how that could be changed. Skinner did this by keeping individual pigeons at about 3/4 of their normal weight, so that the birds were always hungry, and food could be used as an automatic reward.

The pigeon was studied in a uniform box, one it quickly grew used to. One piece of behavior, pecking a colored disk, was measured on a graph.

The pigeon learned that pecking the disc produced a reward. Then, the behavior of pecking could be studied in relation to how often that reward was offered. Or in Skinner's terms, what was the "schedule of reinforcement."

SKINNER: The main thing is what we call "schedules of reinforcement." Reinforcement is what the layman calls "reward," and you can schedule it so that a reward occurs every now and then when a pigeon does something.

We usually use a response with a pigeon pecking a little disk, a little spot in the wall, and you can reinforce with food. But you don't reinforce every time, or perhaps every tenth time, or perhaps only once every minute or something like that. There are a very large number of schedules, and they have their special effects.

There is a good example of how you can move from the pigeon to the human case, because one of the schedules, which is very effective with rats or pigeons, is what we call the variable ratio schedule. And that is at the heart of all gambling devices, and has the same effect. The pigeon can become a pathological gambler, just as a person can.

Now, the fact that we found that out with pigeons, and could prove it by removing or changing the schedule, makes it easy for us to interpret the case with the human subject.

We don't say that the human subject gambles to punish himself, as the Freudians might say, or gambles because he feels excited when he does so. Nothing of the sort. People gamble because of the schedule of the reinforcement that follows. And this is true of all gambling systems. They all have variable ratios built into them.

So what we've learned from the pigeon has made it possible to interpret this vast field very effectively.

INTERVIEWER: Where does that leave free will? Because we all think we have a choice whether to do things or not to do things.

SKINNER: Yes. Well, you see we leave it in the position of a fiction. We have assumed, somehow or other, that these internal states, feelings, and so on, have initiated something. They have started something. They have created. We have done something in a voluntary way. We have will to act.

If you now look at the actual history, we find that there are external reasons why this has happened. In other words, by discovering the causes of behavior, we can dispose of the imagined internal cause. We dispose of free will as an American divines the 18th-century. Jonathan Edwards did.

He said, "We believe in free will, because we know about our behavior, but not about its causes."

And of course, it's the object of a science of behavior to discover causes. And once you have found those causes, there is less need to attribute to an internal act of will. And eventually, I think would need to attribute nothing to it.