Reinforcement « Bird Training Blog

Archive for the ‘Reinforcement’ Category

I scared my bird!

Friday, June 19th, 2009

Training is an ongoing thing. No matter how well we believe our birds are trained, no matter how hard we have worked to generalize behavior, one day in our confidence we make a mistake. That happened to me this morning …

Mijo is a yellow-naped Amazon; he is just over a year old and has lived here at Avian Ambassadors since September last year. He actually lives in the house with us and the dogs, right in the middle of everything, pretty much. He gets to go and hang out on the porch for a good time each day with a “jungle gym” built from recycled plastic tubing and other “treasures”, and in the house his cage is in the lounge. While the day can be quiet at times it also has its share of surprises; by design I want Mijo to have as many different experiences as possible and so far our strategy seems to be working. It is not unusual for one of our re-homed dogs to bark loudly while standing within a couple of feet of Mijo. Nor is it unusual for someone to suddenly appear through the hallway right next to his cage. Over time we have worked to generalize his calm behavior to anything that happens.

So, this morning I got dressed and made my way to my computer, passing Mijo on the way. Suddenly he pinned himself against the roof of his cage, wings flared, eyes pulsing rapidly. What in the world happened? I quickly stepped away from him, he calmed and returned to his favorite perch. What I believe had happened was that the t-shirt I was wearing was so different to anything I had previously had on it was scaring him. The shirt, one I have not worn in a very long time is black with several large pink logos of a sailing boat class I used to sail all over the front. Typically I wear light colored shirts.

What to do? I could go back to the closet and change the shirt, however that was really avoiding the issue and not working to fix it. So, since the back of the shirt is plain, without the logos, I turned it around. The plain black shirt appeared less scary and I was able to reinforce Mijo for calmer behavior it its presence. We worked on this for a while intermittently; I would go off about my business and come back and reinforce the calm behavior, still wearing the shirt backwards. Once he was no longer showing discomfort I turned the shirt around. At first, as expected, there was a regression towards discomfort, nowhere near as severe as the first instance but none the less still there. Gradually over the next 30 minutes I worked to reinforce his calmer behavior in the presence of the “killer” shirt. Right now I would say his behavior in the presence of the shirt is calm, he will perform cued behaviors and willing approach me, no matter how close I am to the cage.

In working with Mijo to overcome this fear I feel the most important aspect of the training was that he always had choice; he could approach me or not. Choice is a powerful thing; it imbues the subject with a degree of control and that raises their confidence. Mijo was in total control of when he chose to approach or leave the scary situation. He was never coerced into “getting over it”. I have seen removal of choice as a training strategy being promoted many times by those “internet gurus.” In fact I have a couple of other articles brewing that will focus on some of these later … for now all is calm in the world of Mijo. I just need to find some brightly colored shirts to wear so that we can continue to generalize calm behavior!

Happy training,

Sid.

Posted in Articles, Reinforcement, Sid Price, Training | No Comments »

Reinforcing and Punishing Consequences

Wednesday, March 4th, 2009

In an earlier article I wrote about the basics of Applied Behavior Analysis so I don’t propose to cover them again here. What I would like to talk about is how one identifies a consequence as either a reinforcer or a punisher.

First, remember that in order to really affect a behavior the stimulus delivered or removed must be both contingent upon and contiguous with the behavior. Contingency is the if/then relationship of behavior and consequence. If the bird does this (behavior) then this (stimulus) is delivered/removed. Contiguity describes the tight relationship in time between the behavior and the contingent consequence. The closer in time these two events the stronger the effect of the consequence (stimulus) on the future behavior.

The labeling of a stimulus as a reinforcer or punisher is contingent upon the effect that the stimulus has upon the behavior. Since to date we have no method of knowing what an animal thinks we must rely upon observing the effect of the stimulus on the behavior, then and only then do we know whether the stimulus presented (+) or removed (-) is a reinforcer (R) or punisher (P).

Therefore one can not discuss if a consequence is R+, R-, P+, or P- without FIRST defining the behavior. This is why one must operationalize (write down the ABC) the behavior, then we can identify clearly the nature of the stimulus.

This was demonstrated in the last few days with an online discussion of the use of jesses with raptors. Jesses are the leather straps attached to the legs of raptors to restrict their movement. One writer identified the strategy as punishment, the other as reinforcement. How can that be? As I wrote above these consequences do not exist in isolation from the behavior and if we operationalize the behavior we can see that BOTH writers were to some degree correct. Assuming that the hawk is standing on the gloved hand and something occurs that startles the bird.

1A. To be deterinined

1B. Hawk flies from the gloved hand

1C. The leash and jesses become taut

Probable Future Behavior (PFB) The hawk will fly less often from the gloved hand.

Thus we have punishment (the likelihood of the hawk flying from the gloved hand is reduced) and the stimulus was added (the taut equipment placing restraint on the legs), therefore with THIS behavior we have a Positive Punishment (P+) strategy.

2A. The leash and jesses become taut

2B. Hawk returns to gloved hand

2C. The leash and jesses become slack

Probable Future Behavior (PFB) The hawk will remain on the gloved hand more often.

Thus we have reinforcement (increased staying on the gloved hand) and the stimulus was a removal (the leash and jesses become slack and restraint is taken off legs), with THIS behavior we have a Negative Reinforcement strategy (R-).

From this example we can hopefully see that before one can know what the strategy being used is one must start by identifying the behavior. Are we talking about reducing the likelihood of flying from the gloved hand (punishment) or are we talking about increasing the bird remaining on the gloved hand (reinforcement).

This example also demonstrates that in order for our negative reinforcement strategy to work an aversive (restraint) had to be present as an antecedent. This need for an aversive in this situation should lead us to seek a different strategy for training the bird to remain on the gloved hand. The preferred approach would be to heavily reinforce the hawk for standing calmly on the gloved hand and, in a controlled environment, not to be holding the restraints, but to allow the bird to move freely off the hand to a nearby perch when it so chooses. When it returns to the hand it is once again reinforced.

It is worth noting that the original discussion was not really about initial training but the continued use of jesses for restraint. When working with raptors in public situations it is a matter of public safety to have this level of control of the bird, regardless of the training level. The best a trainer can do is to build a strong history of reinforcement on the gloved hand with the bird to minimize attempts to fly away. However, again in the interests of public safety, it is necessary to have the backup restraint of the equipment available.

In summary, talking about consequences in isolation from the behavior they are contingent upon has no meaning. The important word here is “contingent” and in order to be contingent there must be a specific behavior that is being discussed. Simply by operationalizing the behavior as was done above will reveal the nature of the consequence.

Please don’t forget, if you have a training question or comment please write to TrainingBlog@avianambassadors.com and if appropriate for an article I will try to address your questions and comments.

Sid.

Posted in Reinforcement, Sid Price, Training, punishment, raptors | No Comments »

Punishment Revisted

Monday, August 11th, 2008

In my last blog article I wrote about the importance of using the terms of Operant Conditioning (OC) and Applied Behavior Analysis (ABA) correctly and consistently. The article was inspired by what I considered to be a disappointing article published by Karen Pryor, in particular her discussion of Punishment. In response to my post I received a most excellent and well thought out email from Chris Shank, a well respected parrot trainer. Chris wrote:

I was quite interested in your response on Karen Pryor’s comments on punishment from her website. I’m currently reading Sidman’s, Coercion and its Fallout.’ Fascinating reading. In chapter 2, pg. 45, he says this about punishment:

‘But we define punishment without appealing to any behavioral effect; punishment occurs whenever an action is followed either by a loss of positive or a gain of negative reinforcers. This definition says nothing about the effect of a punisher on the action that produces it.

It says neither that punishment is the oppositie [sic] of reinforcement nor that punishment reduces the future likelihood of punished actions.’

This is indeed one area where Sidman appears to be at odds with almost all other contemporary behaviorists.

After receiving Chris’ email I did some additional research just to be sure that my understanding of this term was supportable. In addition to having a long conversation with Dr Susan Friedman on this subject I also referred to my copy of “Learning and Behavior” by Paul Chance.

What we have here is an example of how science works; ideas are postulated, discussed, and tested. Science is dynamic and the definition of Punishment is a wonderful example of how science progresses and changes as new ideas are presented, challenged, and tested. If science did not operate this way then behaviorists would not be thinking of Punishment at all since B. F. Skinner himself (the “father” of behavioral science) stated that from his experiments Punishment was ineffective. What those that have followed Skinner have discovered through challenge and experiment is that indeed Punishment does work and possibly the levels of aversives being used by Skinner were too low to be effective. The Chance book cites some excellent studies on Punishment and for those who wish to dig deeper into this subject I would highly recommend reading his chapter on Punishment.

The definition of Punishment I use is the one used by the majority of respected contemporary behaviorists and animal trainers:

“Punishment is a consequence delivered after a behavior that serves to reduce the frequency or intensity with which the behavior is exhibited,” Susan Friedman – “The Facts About Punishment“

“The procedure of providing consequences for a behavior that reduce the strength of that behavior,” Paul Chance – Learning and Behavior

“The procedure of providing consequence for a response that reduces the frequency of that response” – International Marine Mammal Trainers’ Association – Glossary

“The procedure of providing consequences for a behavior that decrease the frequency of that behavior.” – University of South Florida Glossary of Behavior

The way that science tests these definitions is by challenging them with real behavioral examples; indeed this is what Sidman does in another article where he challenges the definition (The Distinction Between Positive and Negative Reinforcement: Some Additional Considerations). He presents a number of test examples that he believes show that the above definition is wrong. For example he asks “When a parent ends a child’s eating between meals by hiding the cookie jar, has the child’s cookie eating been punished?” The answer is patently no, however I see this as comparing apples and oranges because what the parent did was to change the antecedents of behavior (hide the jar) and NOT apply a contingent consequence to reduce the behavior.

In reading the above article by Sidman it appears that his main argument against the contemporary definition of Punishment is two-fold. Firstly that it is not how Skinner defined it and secondly that the contemporary definition was simply adopted by behaviorists without the proper scientific discussion, debate, and challenge. In the light of the fact that Skinner stated that Punishment was ineffective one can only assume that any conclusions he drew about Punishment after this were flawed or at least based upon shaky ground. Since we now know that indeed Punishment does work we need an update to that part of the Skinnerian hypotheses; the contemporary definitions above provide that update. Secondly, in challenging the contemporary definition of Punishment Sidman is addressing the second part of his objection to the definition and it appears that the majority of contemporary behaviorists are meeting his challenge and successfully defending the definition I gave and those above. Therefore through his challenges he is actually encouraging contemporary behaviorists to fulfill the need for challenge, debate, and test of the definition. This is a good thing since to date those challenges have been answered.

So, I stand by my original article and its definition of Punishment, at least until through its continuing journey science brings a better way of expressing the concepts that we use to describe behavior. And that after all is what these terms are all about; building a common language that practitioners of behavior change (i.e. trainers) can use to communicate clearly with both peers and students.

Chris asked in her email if in the light of the writings of Sidman on this subject it was correct to say that Karen Pryor was “wrong” in her discussion of Punishment. Having read some more and talked with others about this I stand by the point of my article; the terminology of OC and ABA can be confusing even when used correctly, to mix historical and contemporary concepts can only lead to deeper confusion; especially when these concepts are held to be correct by the majority of contemporary behaviorists. I find this especially important in arena that Karen Pryor publishes her writings, the pet community. It is vitally important that those who have respect and reputation in that community communicate in a cohesive and accurate manner the principles and terminology of the science. I still feel that Karen Pryor’s article failed to meet those criteria and yes I believe she was wrong in her definition of Punishment.

My thanks go to Chris Shank for opening this discussion in such an interesting way.

In closing I would like to refer back to my previous articles about Primary and Secondary reinforcers. I was sent a clip from an internet posting that stated that my understanding of these terms was incorrect. I do not propose to reopen that discussion since I believe I clearly stated the correct definitions of those items in the original articles and that the poster of the message continues to be mistaken. I will simply refer anyone who is confused about the terms back to my original articles.

Sid.

Posted in Articles, Dr Susan Friedman, Questions, Reinforcement, Sid Price, Training, Weight management | No Comments »

Trainers who use science are the best … maybe

Thursday, April 24th, 2008

Seeing discussion of training in online groups is excellent, not only because it means that it raises the profile of training but also because it gives me an opportunity to understand better what this blog may bring to the community. What peaked my interest this morning was a discussion about trainers who use science and trust building.

I think everyone should be pretty much aware of my approach by this point in the life of my blog however just for the new readers let me say that I place trust and relationship building ahead of everything else when it comes to training. As I have previously said there are methods that can be employed that will override the lack of trust an animal may have in the trainer; their use however depends upon the ethical position of the trainer.

I noted one comment in the thread I was reading this morning that said:

“There is no morality or ethics attached to Operant Conditioning.”

Now I think I understand what the writer was saying but that sentence kind of upset me a little, it tweaked on an important subject, the ethics of trainers and how they affect the choices those trainers make. What I believe the writer was saying was that the science itself does not imply or apply any ethical or moral judgment. When we use the scientific term “punishment” it is simply describing a consequence of a behavior that is likely to reduce the presentation or frequency of that behavior in the future. As far as the science is concerned there is no judgment about the consequence. However, when we come to the application of the science we certainly do find ourselves needing to make ethical, even moral judgments in our choice of strategy.

This is especially true when it comes to the use of weight or food management in the training process. As I have written before, motivation is a balance and one can certainly tilt the balance in favor of an animal performing a requested behavior by reducing its weight through food withholding. The ethical question is whether it is the right thing to do before all other factors, including better trust/relationship building, have been exhausted. In my opinion it is not.

Also, the subject of the discussion, “Trainers who use science and trust building” I think missed a huge and important point. Even the strategies that are thought of as “bad” or inappropriate are using that same science. The use of aversives and punishers is also included in the science. Therefore even the trainer who towels a bird to “break” it, a horrible strategy that hopefully is now way behind us, is using the science (flooding). One simply can not claim that a trainer who uses Operant Conditioning and Applied Behavior Analysis is doing it the right way. It is the ethical choice of strategy made by that trainer that should define them.

One more point pops into my mind too. I keep reading people who say “we train only with positive reinforcement” like it somehow validates them and their strategy. Let us not forget about the ethical choices before we place these folks on a pedestal. For example, think about someone who makes this claim who uses weight management as their primary strategy, they have not built a strong trusting relationship they have simply built a food dependence. They can rightly claim to use positive reinforcement, that’s what they are doing, reward correct behavior with something that increases the likelihood of the behavior being repeated. However, consider this; what if that same bird was capable of performing to the same level with only the smallest reduction in their diet and therefore their weight and that this level was achieved by the trainer taking the time to build trust, confidence, and a good relationship with the bird. By gradually presenting new environments to the bird so that its confidence grew. Which of these trainers would you think is the better trainer?

This same thread brought a couple of other points to mind that I am hoping to expand on in future articles. Right now with spring in the air it is time to go and work some birds.

Sid.

Posted in Food management, Reinforcement, Sid Price, Training, Weight management | No Comments »

The Primary/Secondary Reinforcement Dichotomy

Thursday, March 27th, 2008

A couple of weeks ago I got an email about a statement I made about secondary reinforcers.

“Once the subject understands the training process it may be possible to introduce secondary or conditioned reinforcers such as attention, verbal praise, or access to toys.”

The writer questioned if I really thought that my examples were secondary reinforcers. This question lead me into some research because while I really did think they were secondary, otherwise I would not have written it, a good trainer should not blindly hold on to what they currently believe, but investigate and question those beliefs. This is one of the tenets of good science too. I try to keep in mind the following quotation:

“A belief is not merely an idea the mind possesses; it is an idea that possesses the mind.” – Robert Oxton Bolt.

Before digging into this subject I would like to thank Dr Susan Friedman for her help in clarifying my thoughts on the subject of reinforcers and also for sending me into some interesting thought loops as we discussed this subject over the past week or so.

In my discussions of this with Dr. Friedman she pointed out that primary/secondary is a man made concept and in nature it is unlikely that we find such a clear cut division. The division is used by behavioral scientists and trainers to convey the concept that many neutral stimuli can become reinforcers through close, repeated pairing with a primary reinforcer (secondary reinforcers).

From a behaviorists standpoint there are three groupings of reinforcers; those that are from the evolutionary history of the animal, the so called “hard-wired” reinforcers, those that are based upon the past history of the individual, and those operating in the immediate environment. Of these three the first one groups together the primary reinforcers. To quote Dr. Friedman, “… primary reinforcers are a very short list — when primary reinforcers are understood to mean automatic, without prior experience.” She also added “Our (behavior analysts) adage is, when in doubt call it a secondary.”

So are “attention, verbal praise, or access to toys” primary, i.e. “automatic, without prior experience”? Personally I don’t think they are and to again quote Dr. Friedman “… it is likely that for social species, affiliation behaviors in a broad sense is [sic] a primary reinforcer but the behaviors used to get that outcome are largely learned.”

This last point about affiliation is worth expanding a little. Affiliation describes the interaction of social species, e.g. the mutual preening of a bonded pair of birds. It is often cited as support for head-scratching of companion birds as a primary reinforcer. However, as Dr. Friedman also points out, there are no hands in the wild! A bird needs to learn that a hand approaching its head will deliver a potentially enjoyable scratch. So, if it needs to learn this by definition it is a secondary reinforcer. Similarly being in close proximity to a social group member or mate in the wild can not be used as justification for similar close contact with a human. Once again the bird needs to learn that a human may bring reinforcement, therefore it is secondary!

Plus, there is another point that excludes reinforcers such as attention from being primary and that is that it may not be reinforcing across all members of a species. True primary reinforcers are primary reinforcement across all individuals of a species.

This is a subject that has been endlessly discussed by behavioral scientists for almost as long as the science has been practiced. It is worth restating that the primary/secondary dichotomy is manmade and nature is rarely so clean in its distinctions. The best way to think of primaries is as a very short list with the distinct characteristic of being automatic and without prior experience, i.e. instinctual, not learned, and from the evolutionary history of the animal. Everything else is secondary.

Keep the questions coming by email.

Sid.

Posted in Dr Susan Friedman, Questions, Reinforcement, Sid Price, Training | No Comments »