Thursday, March 6, 2008

Odds and Ends of Learning Theory Vocabulary

Premack's Theory of Reinforcement – something good can reinforce something not as good. If you do a certain thing you have the chance to do something else that you really like. This has the possibility of increasing the likelihood of the desired behavior without having to use a standard food reward, although food can be used.

My insane black Lab, Beamish, had the hardest time learning “stay”. I was very frustrated, both with the lack of progress and with him bowling me over as we would head up stairs. One day I told him to “stay” at the bottom of the stairs and he actually did it while I went up about 4 stairs. Then I released him and he tore up the stairs. I started asking him for longer and longer stays with running up the stairs as a reward. He wound up with a fabulous stay.

This evening I brought out a plastic recorder for Bug to play. She was being pushy and grabby so I told her to sit on a pillow and wait for me to call her over for it. When she finally managed to sit in one place, I told her to come over and get the recorder.


Extinction – when no reinforcement is given for a response, the response is no longer offered. Be aware of extinction bursts or periods of response, often very intense.

See http://dogtrainersbaby.blogspot.com/2007_04_22_archive.html for examples of extinction and extinction bursts.


Capturing behavior – waiting until the student does the behavior and reinforce it right away. (Works well with very active students who offer a lot of behaviors.)

I wanted to put Beamish's jumping up on cue. I stuck a clicker and treats in my pocket and walked into the house. Beamish jumped on me and I reinforced him for it.

My mom realized that Bug hadn't pooped all morning. After lunch she put Bug on the potty and low and behold, there was poop. She was heavily praised and got extra stories. (And probably candy, but she didn't mention that part.)


Shaping by successive approximations – the student is reinforced for small steps leading up to the finished behavior. (Great for fearful or aggressive dogs – or kids I guess.)

In heeling Havana was first rewarded for sitting at my left and looking at me. Then she was reinforced only if she took a step when I did. Then reinforced for taking two steps, then three, etc. until she was able to walk a straight line across the room in heel position.

Bug started off building towers by lining up blocks. Seeing how impressed we were was reinforcing, so she started putting one block on top of the other and oohs and aahs ensued. Then she figured out the towers make a great crash if they're taller and she's gradually developed her skill to be able to build towers 15-20 blocks high.


Prompting or luring – using physical placement, manipulating the environment, or a treat to encourage movement toward the desired behavior. (A quick way to get the behavior.)

To teach Havana to finish to the right (sit in front of me and then move around to my right, swing behind me and get into heel position on my left) I put a treat in front of her face and let her follow it around behind me until she was in heel position. Then I clicked her and let her have the treat once she was in position.

I'm not proud of this, but sometimes a Mom's gotta do what a Mom's gotta do. Sometimes when Bug is being very difficult about leaving a room – like her Daddy's office when he's trying to have a conference call with work – I take her beloved giraffe and walk out of the room with it. As much as she loves Daddy, she has to have her giraffe at all times so she follows me. As soon as she's out of the office and the door is closed she gets her giraffe.


Back Chaining – breaking a behavior down into its individual parts and then the last part of the behavior is trained first so that it becomes rewarding in and of itself. (Uses Premack.) Then the second to last behavior is trained, third to last, etc.

We have a flyball tournament this week so I've got those examples in my mind. The very first thing I did when I got Havana was teach her to tug, largely because in flyball I want her to come back to the tug. Then we set her at the box and had her jump from the box to me. From there we taught her how to operate the box and then to go from the start line over the jumps to the box. Each old behavior was a reward for the newly learned behavior.

Potty training is also very much at the forefront of my mind. Bug loves playing in the water in the sink. From there I let her “help” flush the toilet, then put toilet paper in to watch it swirl when she flushes. All of this before she ever even sat on the potty. We're now at the stage where she doesn't get to flush or wash her hands unless she uses the potty (as opposed to just sitting there and asking me to read her stories over and over).

Unlearning - good and bad!

Habituation – when one stops reacting to a meaningless stimuli. One is able to sense the stimulus and respond, but has learned not to. There is short term and long term habituation.


Short Term Habituation – stimulus eliciting the orienting response (or thing that makes you react) happens over and over again intensely in short intervals. The dog stops reacting relatively quickly within this period, but if the orienting response happens again after a break, there will likely be spontaneous recovery and there will again be a response.


Our friend's dog Pepper is terrified of children. Unfortunately (for her) Bug adores her. She sees Bug for about 4 hours two to three times a month. By the end of a visit she is calmer around Bug, but reverts to her fearful self the next time they meet.


Bug does not like the sensation of goats licking her. She loves to visit goats and feed them, but as soon as they lick her she squeals and tries to get away. By the end of the visit she allows the goats to lick her, but as soon as they try it the next time she squeals again.


Long Term Habituation – the stimulus eliciting the orienting response happens less frequently and with longer intervals between. These long intervals encourages long term habituation and as a result, there is less spontaneous recovery.

When we crate trained Havana to be alone at our house, the disappearance of a human brought out her orienting response (a puppy “hey! Don't leave me alone!). We made a point of leaving her crated alone one or two times a day and sometimes skipped a day here and there. (The advantages of working from home.) Over time her response to being alone was reduced and now she has no orienting response at all when she is alone in her crate.


There is a play room at the gym I use. Bug is perfectly safe there and has a pretty good time when she settles in, but she had a fit when I leave. My disappearance creates an orienting response of screaming as if she's being dismembered. My gym attendance isn't very regular, but as we've followed this same routine, Bug's response to my leaving gets less loud and of shorter duration. Maybe by the time she's 12 she'll be fine with me leaving.


Sensitization – the opposite of habituation. This happens when the reaction to an orienting stimulus gets stronger instead of weaker. It is usually involved in strong emotional responses and is more general than habituation.


Havana used to have no opinion about squirrels. However, she has been seeing more and more of them and has become very frustrated with her inability to chase them. Now when she sees squirrels she barks like crazy. (Even crazier than usual for a collie.)


Bug was watching a movie about a lost puppy. During a sad scene a sad song was playing and the music caused an orientating response. At first Bug didn't respond, but the more she listened to the song, the more upset she got. When there was a reprise of the song at the end of the movie she began wailing and could not settle down for the next 15 minutes.


Adaptation – does not involve learning. One stops reacting because of physical fatigue.

While there are examples of my own dogs going through this, what most quickly comes to mind is Cesar Milan. He will take a dog aggressive dog and run it on a treadmill until it is exhausted and then bring in another dog. The aggressive dog will not react because he physically cannot. He hasn't learned to ignore the other dog or to do a behavior incompatible to lunging or biting, just simply can't get up the energy to be aggressive. As soon as he's rested he will most likely respond aggressively again.


Before we got on the plane to visit family in Florida, Bug and I ran up and down an empty corridor in the airport. When we got on the plane she was (relatively) quiet. She didn't learn that she was supposed to be quiet on the plane, she was just too tired to bounce and be noisy. We didn't have time to run before our connecting flight and she proved that she hadn't learned to be quiet on a plane.


Learned Irrelevance – one stops responding to a stimulus because there is no consequence in the presence of the stimulus. Unlike habituation, it is unlikely to have spontaneous recovery. Instead it actually makes it more difficult for the student to condition to the stimulus because it has been irrelevant for so long.

I introduced the word “heel” to Cousteau long before he ever understood what it meant. As a result he didn't really have a heel. When I tried to teach him heel position he had a very difficult time putting together the word “heel” and where he was supposed to walk because for so long heel just sort of meant I was going to walk away and he'd better come with me eventually.


Bug used to ask to go to bed. Now when we say “one minute to bed time” and then “bedtime” she has a fit. We've gotten into the bad habit of saying these things and then not following through with putting her to bed because it's such a pain. When we do pick her up for bed she has a fit because it's completely caught her off guard. The transitional minute and the phrase “bedtime” have no mean for her any more. (Way to go, Momma and Daddy! :P )

Wednesday, March 5, 2008

Pavlov vs. Skinner

This section really has very little to do with every day life. Feel free to skip it unless you're a big behavior nerd like me.

There are two types of conditioning or learning. One is Classical Conditioning and the other is Operant Conditioning.

Classical conditioning happens when the student makes a connection between two things. She then responds to the first event in anticipation of the second event. The response is not required for the unconditioned stimulus to appear. This type of learning is often associated with Pavlov and his drooling dogs.

Conditioned Stimulus or CS – a stimulus or cue which has no meaning and is presented before an unconditioned stimulus.

Unconditioned Stimulus or UCS – a stimulus or cue that always causes an unconditioned response.

Unconditioned Response or UCR – something that automatically happens in the presence of the UCS. It does not have to be taught. (May be a reflex.)

Conditioned Response or CR – what is taught to the learner when she associates the CS and the UCS.


Cousteau has learned that a certain route leads to the dog park. The first time or two we went to the park nothing happened. Then he made the association between that route and the park and from that point on as soon as we approach the route to the dog park he begins to howl and sing at the top of his lungs.

CS – the route to the park

UCS – being off leash around dogs at the park

UCR – enjoyment of freedom and fun

CR – anticipation of enjoyment of freedom and fun


Bug has learned that when we shop in a certain place we will stop for Starbucks. In the beginning we could do some shopping and if I felt like a hot drink we would go into the coffee shop. That routine has been repeated enough that as soon as we are at the cash register, Bug asks me if she can have a “vanilla milk”.

CS – our presence at a cash register at a certain shopping center

UCS – vanilla milk

UCR – crave vanilla milk (or drools...)

CR – ask to go to Starbucks for vanilla milk


Operant Conditioning is when a student understands that his behavior has consequences. He must do something to make something else happen. This type of learning is associated wit B.F. Skinner, Kellar Breland, Marian Breland Bailey, and Bob Bailey.

ABCs of Operant Conditioning

A ---> B ---> C

A = Antecedent or cue (or discriminative stimulus if you want to be fancy)

B = Behavior or the student's response to the antecedent.

C = Consequence (or punishment or reward) as a result of the behavior.


In flyball Havana is told to “hit it” and must go over the 4 jumps, hit the box, catch her ball and bring it back over the jumps to me. When I have the ball, she gets her tug.

A – hearing “hit it” (or just seeing the jumps and the box)

B - bringing the ball to me

C – playing tug


Bug has learned to sit in her chair if she wants popcorn at snack time. As soon as the popcorn popper comes out she goes to her chair to wait.

A – presence of popcorn popper

B – sits in chair

C – gets a snack of popcorn (or as she says, a “nack a POPtorn”)


This is also where the concept of Thorndike's Law or the Law of Effect comes into play. That is “if a consequence is pleasant, the preceding behavior becomes more likely. If a consequence is unpleasant, the preceding behavior is less likely.” Basically – you do something and something good happens, you'll do that thing again. Do something and something bad happens, you won't do that thing again.

If I was cruising for punishment I'd get into the four possible types of consequences, but since I understand those fairly well I won't risk confusing myself. :)

Technical learning theory jargon applied to dogs and kids

Wow, it's been an embarrassingly long time since I last updated my blog. You'd think with a toddler and two active dogs that I would have tons of time to write, but somehow I don't. Go figure!

This entry will be a bit of a departure from my other entries. I'm getting ready to take the Certified Pet Dog Trainer exam (www.ccpdt.org) and I'm going to multi-task and up date my blog while I solidify some concepts in my head. I'm a visual and physical learner so typing gives me the chance to read what I've written and it helps to get it stuck in my brain better. But more on learning styles when I get to that point in an entry!

Much of this section on basic vocabulary comes from Pam Reid's book Excel-erated Learning. As she explains in the intro, this isn't a “how to book” it's a “why” book – interesting for the dog owners out there but particularly useful for dog trainers. The format of my study entries will be a definition of the term, an example of the concept learned from working with dogs and another observed by working with my daughter or other children.


Learning
At the most basic level, learning is a change in behavior based on past experience.

Cousteau has learned that when I am out of the kitchen, putting the entire front end of his body on the counter will enable him to reach food anywhere on that surface.

Bug has learned that positioning her stool in certain locations along the counter enables her to reach almost anything on that surface.

Both Cousteau and Bug experimented several times before finding a technique that was effective and this is the first technique tried at the next opportunity. Before they would try to reach an object with all feet firmly planted on the floor.


Learning/Performance Distinction
The distinction between knowing how to do something and actually doing it.

Cousteau can execute a perfect heel pattern in a lesson. He has learned how to do it. However, when we are in a trial setting he does not perform a perfect heel pattern. Just because he does not do a heel pattern in a trial does not mean that he never learned how to do a heel pattern. (It just looks like it!)

Bug loves to sing. She has learned the words to “Twinkle, Twinkle Little Star” (or Tinkle, Tinkle, Widdle Stah). When asked to sing the song for guests she does not know well, she will not sing. The absence of her singing does not negate the fact that she has learned the words to the song.


Latent Learning
The act of learning how to do something but not having the ability to display that knowledge until later.

Cousteau has known for years how to get something off the counter, but he did not display that knowledge until Bug became old enough to leave food on the counter and walk away. (I swear I'm going to stop picking on the big yeller dog soon!)

Toddlers are great demonstrators of latent learning. Bug learned how and when to say one of my Dad's favorite phrases “It's a tough job, but someone's got to do it.” She did not display this knowledge until at least a day later when I offered to carry something heavy for her and she replied “It's a tough job, but someone's gotta dew IT.”


Four Stages of Learning
acquisition (acquiring) – first phase where new knowledge is acquired

fluency (automatic) – second phase – response is fluid and automatic and speed of response can be improved

generalization (application) – third phase – learns that new knowledge is relevant in a variety of places and circumstances. (Rarely automatic – must be emphasized during training)

maintenance (always) – new knowledge is incorporated into the learner's “tool box” of behaviors. Needs review to stay at this level.

Cousteau took about 2 years to go from the acquisition to the maintenance of flyball skills.

Acquisition: teaching him all of the little parts – first going over one jump and then two, three and four. Then triggering the box, passing other dogs, and dealing with distractions.

Fluency: going down four jumps, getting the ball, coming back over four jumps, and getting his tug without breaking it down or frequent cueing, so basically, the sequence became automatic.

Generalization: applying the knowledge of doing the entire sequence in practice and also in tournaments both home and away.

Maintenance: we continue to practice despite the fact Cousteau has the basic steps. He will (almost) always complete the flyball sequence no matter what – other dogs in the lane, bad passes, jumps knocked over, commotion on the sidelines, different handlers, etc.


Bug LOVES her tricycle. It was hands down her favorite Christmas gift and she went from acquisition to maintenance very quickly.

Acquisition: watching how the pedals move to push the bike forward and that moving the handlebars changes the trike's direction. Figuring out where to put her hands and feet.

Fluency: pushing the pedals first for any movement and then smoother and faster movement. Holding handlebars straight and then turning them to change direction.

Generalization: being able to ride the trike not just in the kitchen, but the living room and rec room and at Nana's house. (It's too darn cold and snowy to test generalization outside yet.)

Maintenance: riding the tricycle all over the place without thinking precisely about how she's moving her hands and feet.


Principle of Parsimony
The simplest explanation is usually the correct one, unless there is evidence to the contrary. In other words, if you're sitting at a horse ranch and hear hoof beats, don't assume there's a zebra coming.

Cousteau started drooling for no apparent reason. The first time it happened I checked in his mouth to see if there was anything caught there. This is a logical and simple explanation of the behavior. I didn't see anything, but noticed a broken tooth. He stopped drooling for awhile and then started it up again. Even though the tooth didn't appear to bother him during a couple of exams by some dog savvy friends, my vet, and myself, I decided that must be the problem and scheduled a removal. This is logical, but not a simple explanation. Three weeks after the removal he drooled again. This time I realized each previous episode was when he was around a female in heat and drooling is a common event for males when a female is in heat. A simple and logical explanation for the behavior and ultimate the correct one since I've now noticed he only drools in the presence of a female in heat.

Bug has been very cranky lately. I had begun to wonder if she had some sort of illness that didn't physically manifest itself as a fever, rash, or intestinal distress, but caused a change in her behavior. This is not a simple, nor especially logical explanation given the facts I have. She was just horrible at flyball last night – refusing to sit down or cooperate with me even in the simplest requests. I commented on the behavior change to a teammate who is also a nurse working up to asking her if she'd ever heard of such a disease. She told me her son went through the same thing at that age. Then we got home and discovered Bug had horrible diaper rash. Age-appropriate behavior and obvious physical discomfort are two simple and logical explanations of the behavior. (Diaper rash is cleared up today and she's much happier.)