Diary of a Chicken Trainer, Part 2
Day 1, Friday, October 27, 2000
This class is going to be much different from the beginning class. This class is as much about learning to make our own training plans as it is about learning training theory and techniques. This is terrific for the people in the class, but it may not be so good for the people reading these notes. It will be hard for me to detail here exactly how much planning went into determining each step.
Everyone worked with two birds. Not only is every individual in the class making his own "plan of attack," but each person is making a separate plan for each bird. I will try to explain in as much detail as possible my own plan for each bird, why I made the choices I did, and the results of my choices - good and bad. I don't have the notes - or the stamina - to detail the training of my training partner's birds, much less those of other people in the class.
My training partner, by the way, is Melissa (Mel) Bussey, whom I know from the mailing lists, but hadn't met in person before. She's an absolute delight, and I liked her right away. She missed the morning periods, but because she took the introductory course just a couple of months ago, she caught up to us very quickly.
In the morning, we all followed the same steps to get reacquainted with the process. The steps were:
Some notes from Bob during these periods:
In the afternoon, we were ready to begin training our first behavior. We'll be using a laser pen as a cue. Our task is to illuminate a piece of white paper, and the bird will peck a spot on the paper. When we stop shining the light, they will stop pecking. This is a conditional discrimination. They will have to go a certain period of time of doing nothing with the target (as well as a certain period of pecking with the cue light).
Eventually the target will be mounted vertically on a wooden apparatus. When the red light appears, the bird should begin pecking the black dot within half a second. The bird should continue to peck until the light goes off. When the light goes off, the bird shouldn't peck at all. In the final test, the bird will have to refrain from pecking for at least 20 seconds.
The light must be shone approximately three inches away from the dot - it's not the target. Some of the birds have been taught to target it. Those must extinct that behavior.
We have to get the pecking at the spot and then get it under stimulus control. This is what we want. This session is to determine what our bird "has" - what we have to work with. Last step will be to determine what we have to do to get from what we have to what we want. We have to find out what our birds' response is to the white paper and to the laser light.
We will have a pecking response. We have to compare it to how it pecked the targets before.
We put the target on the paper, black spot down, so there was only a white side showing. We shone the light on the paper and did three 15-second sessions with each chicken to see rate of pecking at the laser. My na´ve bird pecked twice, once, then zero times during the three trials, respectively. My experienced bird pecked 17, 16, and 20 times, respectively. That bird has a very strong targeting response, which will have to be changed.
Bob then showed us how to set up the apparatus with the strip and target in place. He warned that some chickens consider pecking at the spot when it's mounted on the wall a different behavior from pecking the spot when it's flat on the table.
After this training period, we went back to the lecture room for some brainstorming. Some notes from that period:
We went back to the training room for the last session of the day. The first order of business was to formulate the beginnings of a training plan for each chicken. I brought with me some log sheets I use with my dogs. When I work with my dogs, I work in series of repetitions, not in blocks of time, so the sheets aren't exactly what I need to capture info for this training. But, I made due.
Mel and I alternated, and we developed a system for alternating work with each of our birds. I won't bore you with the details. I'll try to sum up the process/progress for each bird below.
Bird #1 is my na´ve bird. It had no desire to peck the red dot. So we did one 15-second session to judge how strong her pecking-at-the-black-dot response was. She pecked it 15 times, so we determined that she had a strong peck-the-black-dot response.
Next I wanted to present both stimuli - the black dot and the red light - simultaneously, so we could reinforce pecking the black dot. Because she had shown no desire to peck the red light, I was willing to shine the light close to the black dot. My reason for doing so was so she would notice it - so it would be salient. For these sessions the red light was on the entire time.
I did one 15-second session at this criterion, recording the number of pecks at the black dot. (Each peck was reinforced.) Her response rate was low - only three pecks in 15 seconds. So I decided to remain at this criterion until her response rate was stronger. I did two more 15-second sessions. She pecked five and seven times, respectively. Then I increased the time to 30 seconds, and she pecked 10 times. I decided I was satisfied wit the rate because she was steady, if not fast.
Now I was ready to begin teaching her that she should peck when the light is on and not peck when the light is off. My plan was to have the light on and reinforce pecks for approximately 10 seconds (of a 30-second session). Then I would turn off the light and wait for the rate of pecking to diminish. When there was a change in rate, I would turn on the light again for the duration of the 30 seconds.
The first attempt at this criterion was disorganized. I didn't decide until it was over that I should be counting the number of pecks that occurred when the light was off before the rate change in order to track improvement. Also, I wasn't prepared for her to get frustrated and start scratching. I got discombobulated and reinforced the wrong thing. My next try at this criterion was a little better. She pecked six times, then five times, respectively.
Bird #2 is my experienced bird. This bird, whom Mel named Malachi, had a strong peck-the-red-light response. Her average number of pecks at the red light in the past three 15-second sessions was 18. So we did one 15-second session with just the black dot to determine how her peck-the-black-dot response compared. In a word: miserably. She pecked the black dot just eight times.
So I decided to spend some time strengthening the peck-the-black-dot-in-the-absence-of-the-red-light response, even though in the final behavior this response wouldn't be reinforced. I did three 15-second sessions. Her pecking rate was 6, 8, and 11 pecks, respectively.
That was all the work I did with that bird today. Tomorrow I will begin differential reinforcement, reinforcing for pecks at the black dot but not at the red light. To make it easier, I will position the red light further from the black dot - approximately 3 inches - than I did with my first bird. I will also feed the bird near the black dot to encourage her to stay at that end of the target.
Finally, we went back to the lecture hall for a wrap up. Some final notes for the day:
We worked until after six, so everyone was ready for dinner. I ordered room service, got my notes finalized, and headed to bed early. I think my mind is going to explode, my back is aching, and I think keeping these notes is going to kill me. But otherwise, it was a great day!
Day 2, Saturday, October 28, 2000
The morning was very frustrating. Very, very frustrating. The chickens are fine. My mechanical skills suck worms. There's a ton to coordinate in this scenario. I have the clicker and food in one hand, and the laser in the other. The chicken pecks once, and I have to C/T and turn the light off. I'm watching for a change of state. If the chicken pecks, doesn't scratch, and hesitates, I turn the light back on. If the chicken either pecks the light or scratches, I have to turn it back off and wait for the opportunity to turn it back on. There are just a lot of contingencies, and I have the coordination of two-year-old on roller skates pointed downhill.
I had birds whom I was trying to teach a discrimination. Judicious timing of the red light and the clicker would make that task much easier - basically the chickens would simply have to experiment until they figured out what worked. However, because my timing of both the light and the clicks were not only off but flat out WRONG sometimes, I confused the issue.
We were also confused about how to keep data, because we had too much info coming at us in each session. Some ideas from Mel:
Two ways to extinct pecking the red dot:
At the lunch break, I wasn't sure that my chickens were learning anything. In fact, I was worried that they were developing bad habits due to my incompetence. After lunch my mechanical skills improved a little, thus my chickens' behavior improved somewhat.
Bird #1 is a scratcher. She gets frustrated very easily. I'm trying to get the light off and back on before the scratching behavior starts - trying to capture just a moment of hesitation. But everything happens very fast, and it's hard to do.
Bird #2 has horrible latency, but she really looks back and forth between the red and black and intentionally chooses the black. That's good! She's also doing less pecking at the black dot when the red light isn't present.
Still, by mid-afternoon, I could see an improvement over the morning.
We haven't been keeping data today. We've just been doing session after session of differential reinforcement. At one point we added the blue board under the target and put the wooden apparatus on the table. We started with the board flat on the table and have been angling it up slowly. We've been lucky - none of our birds have been phased by the addition of the board or apparatus. The real difficulty with this is that Mel is right-handed and I'm left-handed, and the apparatus is set up to accommodate one or the other. It's really a pain in the ass, and I finally gave up and changed my plan. I'm not the least bit happy with it, but such is life.
By the time we quit for the day - and we were both ready to quit for the day - we had the board standing straight up and attached to the apparatus. We've got to go back to keeping notes because even now, just a few minutes after class ended, I can't really remember the characteristics of the two chickens - except that I'm not particularly happy with either. Hey, if I'm going to beat myself up over it, I ought to at least have good, strong data to prove my incompetence.
I realize in hindsight that this whole exercise is about developing my skills, not about training a chicken. That doesn't mean much when I'm screwing up my chicken. Have I mentioned that I suck at training?
The group went to tour Hot Springs and out for Chinese food for dinner. I have work to do (if I can get into my work account), so I'm going to stay here.
Day 3, Sunday, October 29, 2000
Today was better than yesterday. The birds made a little progress, and my skills improved a bit. Actually, I guess my skills improved a lot because I didn't spend as much time swearing at myself.
We added a second behavior - this one designed to help us develop skill at training a non-discrete behavior. Having a second behavior to work on helped a lot. I never got too frustrated with one thing because I was regularly switching gears to train a different behavior.
The non-discrete behavior was pulling. We started by pulling a rubber band attached to a wooden block. The final behavior will be pulling a weighted pan (2 lbs.) over the length of the table in a single pull. First we had to evaluate our birds (on both pecking the wooden dot and pulling the rubber band) and decide which bird would do which behavior. Because pulling is more difficult to teach, we wanted to pick the bird that was best at pulling for that behavior, even it happened to also be the best at the pecking behavior.
Bird #2 was good at both behaviors, so I decided to use her for the rubber band pulling. In the morning, we worked with the rubber band attached to a block of wood. We put masking tape with numbered marks on it to measure progress. She had obviously done this before and had pretty strong pulls pretty quickly.
In the afternoon, we switched to the pan apparatus. It's a loaf pan with pebbles for weight. It has a piece of fishing line attached, and a plastic loop attached to the other end of the fishing line. So the bird wasn't pulling a rubber band exactly. She targeted the loop pretty well, but at first she wanted just to shake and kill it. I started watching her feet, clicking for any backward movement. Then I tried pulling on the fishing line. Bingo! I got some good pulls.
Three ways of getting the pulls:
Back-chaining is very often a great way to train because the animal is always moving toward something more familiar. However, if, in the finished process, the animal is going to be expected to do the behavior in unfamiliar territory (or not in a pattern), then back-chaining can work against you. Search and rescue, for instance.
As I said, I got some pulls. However, the behavior wasn't very strong - I didn't have a high rate of pulling, even though the pulls I got were pretty strong. More often she just pecked at the loop, or picked it up and dropped it. Apparently, I wasn't the only one in this predicament. Bob told us at the end of the day that we hadn't gotten a strong enough "grab the loop" response before going to pulling. So tomorrow we'll start there and build again more slowly.
Bird #1 stuck with pecking the dot on the board. She definitely improved some today. I got to the point where I had a second or so of hesitation where I could turn the light back on and reward her with a peck. Then Cheryl, Bob's assistant, told us to really work to extinct the pecking - not just wait for a hesitation. Sigh. Back to the drawing board. Still, things are going okay. Although I still don't have data, we're definitely getting fewer pecks (and less scratching) at the dot when the light isn't on, and we're definitely getting longer hesitations. I guess it just takes time.
We started using negative punishment. If the bird screws around, instead of waiting it out, we pick it up, removing the possibility that it can reinforce itself. For the bird that is pecking at the spot, Bob said we can put a piece of white paper over the whole area for a similar punishment.
After class a bunch of us walked around Hot Springs and ended up at a pretty good Italian restaurant. The conversation was terrific, and the walk was a nice change after being cooped up in the hotel for three days. It's 9:30 now though, so I think I'll head to bed.
Day 4, Monday, October 30, 2000
Bird #1 is improving. I'm definitely getting longer pauses when the light is off, and the displacement behavior is less vigorous. I have to be extra careful to vary the amount of time that the light is off. First of all, I don't want the chicken to think that there is automatically going to be a long wait between red lights. Second, I don't want to fall into a pattern and have the chicken "counting" the time between lights. And third, I don't want the chicken to inadvertently chain displacement behaviors in superstitious belief that the head bobs, turns, glances, etc. are part of the required behavior. Slow, but we're definitely seeing improvement.
I went back to the beginning with Bird #2. I'm clicking for good solid grasps of the loop. I'm not going to increase my criteria until I'm getting a steady rate of solid grasps without pecking or weak grasps between times. It's hard not to compare this bird's progress to Mel's bird. She's getting good solid pulls. Oh well, I'll get there.
Throughout the class, Bob calls us to various tables to point out good things and bad things. For example, one person was intentionally training an incompatible behavior to do instead of pecking when the light was on. Bob said that's an excellent way to teach an animal to do something while you're doing something else. However, he doesn't want us to use that technique for this exercise because it's easier than teaching the chicken to just wait.
Bob told us a story about using this technique with their dolphins. When no other cue was given, the dolphins were supposed to "orbit" - swim in a circle of a certain diameter (about a hundred meters). This behavior was incompatible with swimming away. This kept them from losing dolphins when something happened - sudden storm, waterspout, etc. - came up that required them to abandon their dolphins. They actually had a dolphin maintain this behavior for 36 hours for them to return. (They never lost a dolphin.)
Another example was a trainer who had gotten her cup closer and closer to the where the bird was pecking and ended up shaping a rake from the target to the cup - not a solid peck. Bob fixed that by changing the position of the cup.
A third example was a trainer who was upping criteria too fast - getting several correct responses at a certain criteria and then increasing. This causes the behavior to break down at higher levels because the behavior wasn't solid enough at the lower levels. Patience. He said dog trainers, who aren't getting the high rate of response that chickens give, are particularly prone to accept a few trials at a certain criteria as "good enough" to progress.
After lunch I asked Bob about the scratching issue with Bird #1, the bird doing the pecking behavior. He said to disregard it for the most part - though I should be careful not to train a constant "dancing" chicken as an incompatible behavior. With that edict in place, increasing the time between pecks is much easier - trying not to reinforce scratching was really holding us in one place. My bird is still handicapped by my goofs, but that only slows progress, not stops it.
I'm less happy with Bird #2. I don't have the high rate of behavior that I should. However, if I click for any less - I'm only looking for a slight weight change, the beginning of a pull - then I'm reinforcing head shaping or just standing there. It's frustrating. I *get* the behaviors I'm looking for, but I only get a few in a 30 second session.
If we finish with the pecking behavior, we're supposed to teach another behavior that is both cued and non-discrete. In this behavior, we're teaching the chicken to spin in place when the cue, a target stick, is held overhead - and to stop and stand still when the stick is removed.
Tomorrow is our last day, and I doubt very seriously that Mel or I will get to the third behavior. I talked to Bob about it, and he said it's not whether the chicken gets it, but whether we get it. He wants us to know why our chicken did or didn't get it, and what to do. I'm sure he wishes we all would get to the third behavior because that would give us the opportunity to prove that we have generalized the information about adding cues and training non-discrete behaviors.
Do I understand? I *think* I do. I believe I do. I don't know what Bob thinks. I guess the advanced class will tell, eh?
Tonight we're meeting for an informal pizza party. It's a chance to talk about whatever we want and just have a good time together. I'm looking forward to it.
Day 5, Tuesday, October 31, 2000
| Training Articles Contents || Site Home |
List and Site Owner: Melissa Alexander, mca @ clickersolutions.com