Introduction by Will Shor
In recent years, new methods have been introduced into training and conditioning birds for falconry. These new training methods depend on a concept called "Operant Conditioning" based on the theories of psychology professor B.F. Skinner. When tangible rewards cannot be given at the exact instant that an animal performs a desired behavior (such as moving upwind while overhead), promises of later rewards (such as a whistle or hand signal) are substituted. Introducing an intermittent reward program strengthens the behavior.
These concepts were made understandable to animal trainers by Karen Pryor in her book Don't Shoot the Dog (ref. 1), and she used them to train dolphins, dogs, parrots and eventually many other animals. They are described more simply with examples in her booklet A Dog and a Dolphin 2.0 (ref. 2). The application of operant conditioning to falconry birds was introduced by Steve Layman in his April 1994 Hawk Chalk article, "Vertical Jumping to the Fist."
Karen Pryor spoke at the 1996 NAFA Meet in Lamar, and observed falconry during the Meet. After the Meet she undertook to provide us with pecifics on how to train falconry birds using operant conditioning. We are privileged to have her contribution.
I. Establishing a Conditioned Reinforcer as a Communication Tool
When I was learning to use operant conditioning to train dolphins back in the 1960s, I knew that the most crucial tool we had was the conditioned reinforcer (CR), the signal that means "Food is coming." In the last five years, since training thousands of people, mostly dog owners, to use operant conditioning creatively, I've come to be aware that the CR is much more important and useful than I knew, and that it has many functions besides telling the animal it's earned a treat. All of these functions can apply in falconry to make your life easier, your bird happier, your control more reliable, and your hunting and other interactions more successful and enjoyable.
What is a CR? Any arbitrary signal that can easily be perceived by the animal, that's different from everything else in the environment, that is brief in duration so it doesn't "slop over" onto behaviors you don't want, and that has no intrinsic meaning to the creature. Praise, therefore, is not a good CR -- it takes too long, and (sometimes) it already means something to the animal.
A CR can be a touch, a sound, a sight, even a smell. For raptors I suspect that a CR ought to be a sound or a sight, or both. In discussing falconry, I'm going to choose a click made by a manual clicker, or made with the tongue, for a close-up audible CR and a loud whistle for distance. (You can also use your whistle for both close-up and distance if you prefer to keep your hands free). For a visual CR I will suggest a blink of a light from a hand-held blinkable flashlight (sold as key ring lights in most drugstores) for close work, and a strobe light for distance work: a battery-operated flashgun for a camera will work just fine. If you are using a blink, or a toot, instead of a click, as you read these training suggestions, just mentally substitute our CR for my click. (Oh, but you use a whistle for other things in the field. Okay, find a different whistle for this purpose; or partly clog up a whistle so it sounds different. New Zealand shepherds have in-the-mouth whistles that can make an infinite variety of sounds; a "predator call" whistle is another possibility.) [e.g., Cabela's $14.99 "Screamin' peacock", 4221761 - Editor.]
Keep your clicker with you whenever you are caring for your bird. Each time you give it food, click FIRST, then treat. Tidbits are a common event in falconry, I gather; so click, give tidbit. Click, give tidbit. When the bird begins looking for the tidbit after hearing the click, you know it's made the association Click, pause, then treat. Show the bird that the treat will come, but perhaps not instantly. Start clicking in different places, and at different intervals. Are you facing north? Turn and face south; birds know the difference.
Do this for just a little while, one or two sessions is enough. Then, choose a behavior you want, and start clicking when it happens. DON'T pick something you are trying to repair, and don't pick something tough, or on the wing. Just click for stepping up on your hand; or click for turning towards you, or for lowering the wings, or for raising one foot, something the bird is already doing anyway. Always treat after each click. The point of this exercise is not what it looks like, teaching the bird something it already knows, but to teach the bird that by its own actions, it can MAKE you click.
If you have a baby bird, you can teach it to chase a ball, or to chase the end of a target stick such as the end of your riding crop, or a dowel. This can be VERY handy later on for inducing the bird to go into a crate, onto a strange perch, or wherever. (Also, I'll bet you can use click and a long target stick, a lunge whip, for example, for in-air control and exercise. That's a shaping program for one of you to design, not me.)
You can work on two or three little behaviors in the same session: turn your head, duck, raise foot, spread wings, come to my hand -- each one can be shaped at its own rate. When a given behavior is improving, stop clicking every time; click every second time, or every third time, then twice in a row, then wait for four times. You will see the behavior intensify. Varying the effort required to make the click happen tends to increase the effort.
You have now taught the bird two or three ways to make you click, so that the bird understands the game. This shouldn't take more than a dozen two or three-minute training sessions. Now you can go right to using your click (or your blink) to set up Steve Layman's excellent program for bringing a mewed hawk into flying condition (See Hawk Chalk, April 1994 & April 1995). Also, you can use your click to start getting rid of undesirable behaviors, by substituting behaviors you want.
For example: let's say the bird mantles over its food. Step back from it until it relaxes a little; click. Toss it a tidbit. Repeat. Even if the bird is standing on food, the click and treat will reinforce the behavior of calming down, until you can walk in and pick up the bird without getting it upset. (The first time you do this, of course, you should "jackpot", or give an unusually generous reward. Some hawkers give the bird that has just made a kill the liver and heart, on the fist, while in the field: a nice jackpot.)
What about screaming? That's a behavior that infuriates the owners of lots of birds, not just raptors. It's hard to get rid of because, to the bird, screaming brings food. Every time the bird perceives a sign that food is coming -- the door opening, your shadow, your footstep, the car engine, etc., -- and the bird happens to be screaming at that time, bingo! the screaming is reinforced. If these signs do not ALWAYS mean food, but only sometimes, guess what: the screaming is variably reinforced, and the behavior intensifies because variable reinforcement has a stronger effect than reinforcement every time. Some behaviors "extinguish" if they are not reinforced, but just trying to wait the bird out, so the screaming stops by itself, is not very efficient because of all the accidentally conditioned cues (same thing with dogs barking in the kennels; it's often an accidentally but very strongly conditioned behavior.) So use your new CR. Take a moment when the bird is not screaming -when it has just eaten, perhaps, or when you are cleaning the cage -- and click. Treat. Click. Treat. Build on this. (The bird has to breathe; if you are quick, you can click between screams.) Your aim is to teach the bird that all those OTHER cues are not meaningful. The meaningful information is the click, and it can make you click by silence. Don't worry if it screams right after you click, and before it eats; that will stop as it discovers that clicks, not yells, mean food, and that silence makes clicks happen.
Meanwhile, pair the click (or blink) with ANYTHING you have available that is pleasurable to the bird; broaden the meaning of your CR, so that you increase its power to inform the bird about silence. Building a real CR can erase the accidental attachment to all those superstitious CRs -- and help you quiet your bird.
Now is a good time, too, to use your click to teach the bird things it will need to know in the field. Want the bird to use its feet? Click for footing. To fly to the glove? Click for hops, then longer flights. Maybe you can add a signal, waving your arm over your head: wave, hop to glove, click, treat... you are building a very useful sequence.
Chasing the lure is a great behavior to click for. It is wonderful to see how quickly a clicker-trained bird can learn what is expected, even if it does not instantly find the food on the lure. You can perhaps see now that ANY kind of lure can become desirable to the bird that gets clicked and then gets a tossed treat for landing on it. You might not need to escalate the temptation, later, by swinging dead birds (though a jackpot of a dead bird is a fine surprise) because the chance to earn a click is extremely motivating. The swing of the lure becomes a signal, or cue, that the chance is available.
II. Taking Your CR into the Field
You have now trained your bird to recognize a conditioned reinforcer -a click, a blink -- close-up, in the hand or in the mews. The bird has learned to offer behavior in order to make you click. It may have learned, too, a few signals from you: a call, to come to the glove, or a motion to follow a target or a lure. What could all this close work do for us, however, at the tremendous distances of the hunting field?
First, you need a long-distance CR, an arbitrary signal that can be perceived from far away. For birds, suggest a predator call whistle or a strobe light (such as a battery-powered flashgun). At the Lamar Meet when I addressed NAFA in 1996, Nick Fox explained to me that sounds carry very well into the air. In a glider, on a calm day, one can clearly hear conversations from below while hundreds of feet up, much farther away than on the ground with all its sound-deflecting bushes and bumps.
Taking the weather into consideration, acoustic signals, particularly non-fuzzy, high-pitched signals, should carry very well, but a strobe light should be good too, and there is no reason you can't establish both, and use whichever seems handy or works best for you and your bird.
Conditioned reinforcers, once established, are easily transferred. With a clicker, you have already taught the bird about the CR. Now, simply take advantage of any opportunity that presents itself, and pair the NEW CR with primary reinforcers: food or other treats. The bird will quickly register the information that the NEW CR means just what the old one meant: food is coming, plus: you are doing the right thing, plus: what you are doing made your person click, so now you get your treat. Pretty powerful information. From here on, whenever I say "click" I mean Your CR, whatever that may be.
To establish a new CR, you can start with tidbits on your fist, just as you did indoors: whistle/treat. Whistle/treat. The bird already knows the game, which is greatly to your advantage: a new kind of "Bingo" signal is good news, and quickly acquired, once the concept is in place. You can also, of course, follow the new CR with the old CR, to help tide the bird over.
Also there are many new kinds of treats available once your bird is on the wing. Just flying is one of them: I would start by pairing my long-distance CR with the release, every time. I would also pair the long-distance CR with presenting the lure, and with any feeding I did in the field. When possible, I'd activate my long-distance whistle or blink the instant the bird actually took prey. Above all, I would prepare carefully so that I could pair my long-distance CR with flushing of game.
With a CR that's noticed and understood at a distance, you have a powerful tool not only for reinforcing your bird's desirable behavior (if it gives you any) but, more important, for teaching cues. Let's take a cue everybody needs in the field: "Come back to me," which we open-ocean dolphin trainers refer to as the recall.
Pick the signal you are going to use for the recall. I saw that arm signals and yelling were popular in Lamar. That'll work, birds can see arm signals a long way away. I might add flapping cloth, or colors, or both. Just make sure you are consistent and give a clear cue that is different from everything else you do. When you want your bird to come, give the signal.
Swing a lure too, if you need to, to tempt the bird; but this should be a temporary, training measure, only. (I'd advise fading the luring after two or three successful "clicked" recalls.) Now the bird starts to come your way. CLICK and toss the lure. What if the bird then veers off? Just repeat.
When you "click" is a judgment call. Sometimes the bird will not perform the whole recall, and sometimes it won't get the treat; but that's okay. At first, you might click for coming a very short distance; then for coming a longer distance; then from a longer distance, and all the way to you; then for just starting to come from a very long distance. You are, as it were, putting money in the bank. Every time you manage to click the bird for heading your way, after you waved at it, and it then comes all the way in and gets its prize, you are strengthening the chances that it will respond to your cue in the FUTURE.
Vary the distance, vary the difficulty. ALWAYS click if the bird actually comes to the glove, and ALWAYS treat. You can and should stretch the bird's capabilities, by deciding when you'll click and what you'll treat with; but don't lie with your strobe light or whistle. A "That's right!" message needs to pay off.
I would want to fade out the swing of the lure, which is baiting of a sort, nearly immediately. As fast as I could, I would substitute my arm or flag "come here" signal, and click for the behavior of zooming toward me on that signal. Then I would click again when the bird is very close, and freely give the swung lure, with its delicious treat on it, as the primary reinforcer as the bird homes down to where I am. Like a bribe, when you constantly swing the lure you're offering food whether it's been earned or not. The faster you can switch over to just the recall signal, with the lure presented after the CR, the less likely you will be to develop a bird that sits on a phone pole and watches you swing food around. And herein lies another conditioning trap (like the accidental conditioning of screaming described in Part I): a trap ably described by Steve Martin (Hawk Chalk Aug 1996). When a bird has learned that a signal means "Do X and you will be rewarded," the signal itself becomes reinforcing. Don't you feel relieved when the light turns to green? Birds, also, get reinforced by "go" signals. S0: be careful when you give the signal. As often as possible, give your recall (or swing the lure) when the bird is heading toward you. If you fall into the habit of trying to MAKE the bird come, by waving or luring, when it, let us say, is huddled on the top of a phone pole, you may actually shapethe behavior of landing somewhere and staying there. You don't want a superb pole-sitter, but you made one. The best way to avoid this is to keep in mind the modern trainer's question: "What am I actually reinforcing?" Your aim is not to get rid of behavior you don't want, but to make sure that what you are doing is STRENGTHENING behavior you do want.
Another behavior falconers need in the field is the behavior of "waiting on", or hovering overhead waiting for game to be flushed. Speaking as an animal behaviorist and trainer, one of the biggest thrills of my working lifetime, just about as exciting as working trained dolphins in the open sea, was seeing a peregrine, a gyrfalcon, and a prairie falcon each "waiting on", watching their owners in clear expectation of what might happen next, while hunting several sorts of game at Lamar at the NAFA Meet in November 1996. The training problem in "waiting on" is that one needs the bird to be high enough so that it has a decent hunting chance when the game flushes. The simplest step, which only takes self-control, is to try not to flush the game until the bird is high enough. With a new or young bird, one might make the first flush a reinforcement for waiting on at all, then the second flush for waiting on higher, and thus help the bird, in three or four experiences, to learn to climb.
The CR, however, can easily be inserted here. Get set up, with game out there. When the bird climbs, flash the light, THEN kick the ducks up or send the dog in. Like all predators, (and this I learned in Lamar) these birds are experts at learning something from a single event. The CR just intensifies this learning experience. "Oh HO!" says the bird, "I had something to do with that."
Several people have remarked to me that it would be impossible to "click" the bird for waiting on, because it would then have to come down for its treat, which negates being at altitude. I hope you can now see there are two ways around this: one, what the bird was doing when it was clicked is what it will repeat again: not what it was doing when it ate. So, occasionally, you might want to click for altitude and then bring the bird down. More importantly, in flushing the game you have a wonderful primary reinforcer, an event the bird is hard-wired to want, which you can present as the pay-off for clicked behavior no matter how far up the bird is.
Why not just use the circumstances, the flushed game, to train the behavior, rather than all this folderol of artificial signals? Well, you can. Lots of people have been doing so for centuries, and it works. The "click" just makes it work faster and better. These birds are hard-wired to learn fast. Rather than waiting for random successes and failures to teach the bird to climb higher, you can use your CR and the flush to notify the bird that climbing is good. Then when it starts climbing on purpose quite soon after being released (which it will) you can start using your CR to tell the bird what altitude you like. You can also use other kinds of situations to communicate to the bird what's wanted. So, for example, if you find no game, but your bird is up there, click it (strobe, whistle, whatever) for its altitude, give your recall, click it for coming to the glove, and give a good treat. So, you got skunked today: the bird had a good learning experience, anyway.
There are some even fancier tools in the operant-conditioning kit: for example, the "keep-going" signal, which is tied to NO primary reinforcer, and which nevertheless is extremely useful in maintaining strenuous, long-duration behavior that an animal must do on its own. Before getting into all that, however, I'm hoping
- that some of these suggestions will indeed prove useful to falconers,
- that there will be aspects of raptor behavior and learning, and of falconry itself, that I am not familiar with, that you will be able to tell me about,
- that you will build on the operant methods described here, and share your creative insights and approaches with the rest of us clicker trainers out there. Because this is a bit of a revolution, and not just about birds.
I was truly thrilled, in Lamar, to discover that the falconry community, after millenia of being governed by very wise but circumscribed tradition, is springing into a new world of breeding, handling, and biology and field work. And, behaviorally, you are joining the rest of us "clicker trainers" in what is turning out to be a grassroots revolution in behavioral science. Welcome!
(1) Don't Shoot the Dog!, Revised Edition, Karen Pryor, 1999. Bantam Books, ISBN 0-553-38039-7
(2) A Dog & a Dolphin 2.0. an Introduction to Clicker Training, expanded version reissued as
Getting Started: Clicker Training for Dogs, by Karen Pryor. ISBN 1-890948-11-X
Sunshine Books, 49 River Street, Waltham, MA 02453.
For more information, links to useful sites and lists, and clicker training books, videos, and equipment: www.clickertraining.com.
Post new comment