These are my notes from the lectures and labs I attended. I try to share the information as it was presented by the speakers, but there are places where I have added extra comments or filled in a few details. If you have any questions about the material prsented here, please feel free to contact me. The notes are all on one page so you can just start reading, or you can use the links below to get to each one directly. If you’d like more information on Clicker Expo, you can go to www.clickerexpo.com and read all about it.
- Kathy Sdao: It’s a Good Fit! Operant and Classical Conditioning
- Susan Friedman “Reducing Behavior with Reinforcement – Oxymorons and other jumbo shrimp”
- Karen Pryor “On Extinction”
- Michele Pouliot, “Pace, Place and More” – lab
- Kay Laurence, “The Art of Practice”
- Jesús Rosales-Ruiz, “Using Resurgence to Your advantage.”
- Hannah Branigan, “Performance Prep, Foundations for Success.”
- Laura Sharkey, “Give Pups a Chance: working with aggressive puppies.”
- Kay Laurence, “Freedom to Learn: training in the absence of formal competition.”
- Kay Laurence, “Repurposing Default Behaviors” – lab
- Kay Laurence, “The Craft of Fine Slicing.” – lab.”
Kathy Sdao:It’s a Good Fit!Operant and Classical Conditioning
I’ve heard Kathy talk on the subjects of operant conditioning and classical conditioning before, but I attended this talk because I was interested in what she had to say about how they related to each other. When I first learned about operant conditioning (OC) and classical conditioning (CC), they were presented as two separate modes of training. You were either using one or the other. It turns out that this is a bit of an oversimplification because the reality is that they are both happening at the same time. As trainers we can choose to emphasize one over the other, but we really need to be aware that both types of learning are happening. In this talk Kathy emphasized that there is also a progression from CC to OC.
Kathy started off by defining OC and CC, because you do have to start by understanding each separately. As clicker trainers, we are all familiar with the idea that behavior is driven by its consequences. The standard formula in operant (or Skinnerian) conditioning is A -> B -> C, where A = antecedent, B=behavior, C=consequence. In operant conditioning, the likelihood of B being repeated is determined by the consequences, which can either reinforce (increase) or punish (decrease) the behavior. As clicker trainers, we add something the animal wants after the behavior happens, and that strengthens the behavior so it is more likely to happen in the future.
In classical conditioning, the trainer is using a procedure that changes the animal’s response (and associated emotions) from one stimulus to another. It is usually written out as CS + US -> B (reflex). The CS (conditioned stimulus) has no intrinsic meaning. The US (unconditioned stimulus) has intrinsic meaning. When you pair them together by presenting the CS and then the US, you can condition the CS so that it triggers the same emotional response as the US. If this is making your head spin, just think about the clicker itself. The first time you click, the click has no meaning to the horse. After you click and treat a few times, the horse now associates the click with food and gets excited when it hears the click. That is classical conditioning at work.
Kathy pointed out that while classical conditioning is very strong, it is not a long term solution. In many cases she starts with CC and then switches over to OC because most trainers are interested in training specific behaviors and this is done more effectively with OC. At the same time, she emphasized that people often underestimate just how powerful CC can be. This is partly because CC doesn’t usually have the same “epiphany” moment where the animal “gets it.” CC is more of a gradual process.
So those are the differences. This talk was about integrating them together so she pointed out how we use CC when we are working primarily with OC. Her list included “charging” the clicker, pausing between click and treat, creating new reinforcers, adding cues to behaviors and modifying emotional responses. I mentioned that the clicker takes on meaning through CC, and here she emphasized why it is click THEN treat, not click/treat at the same time. In order for CC to work, the conditioned stimulus must come before the unconditioned stimulus. This is why it’s so important to keep the click and food delivery separate.
Her first example of using CC as a foundation for OC was her dog Effie who was afraid of buzzing insects. When Effie heard a buzzing insect, she trembled, panted, cowered and fled. These are all behaviors that happened automatically when she heard the bee. She had no choice. To help Effie get over her fear of bees, Kathy needed to change her emotional response to bees so that she no longer needed to flee. She can do this with a CC procedure called desensitization.
Desensitization is a process of “repeatedly presenting a fear-evoking stimulus (trigger) at low intensity until the animal reacts without fear, then slightly increasing intensity and repeating the process until the animal is comfortable with full-strength trigger.” This is different than habituation which is “decreasing the intensity or probability of a reflex resulting from repeated exposure to an eliciting stimulus.”
The key to desensitization work is working at the appropriate intensity where the animal can learn a new response to an existing trigger. There are 5 ways to change the intensity of something. They are loudness, duration, distance, frequency (how often), and distractions. In Effie’s case, she could use buzz-like sounds and change the intensity in different ways. By pairing the buzzing sounds with something Effie liked, she could slowly change Effie’s “automatic” reaction from one of fear to one of happy anticipation (something good is going to happen).
Once Effie is happily anticipating something good will happen, then Kathy can switch to OC to shape exactly the behavior she wants. In the long run, this means that when Effie hears buzzing, she offers a specific behavior to Kathy, who can then reinforce it. This is an example of what is called classical counter-conditioning because Kathy started with a stimulus that had an unpleasant association, as opposed to being neutral.
She also had example of using classical conditioning with a neutral stimulus. Her dog Nick didn’t have any interest in Frisbees until she started showing him the Frisbee before letting him out to eat apples. Over time the sight of the Frisbee elicited the same emotional response as being released to go eat apples. This brings up an important point which is that many people stop using CC at the point at which the animal is comfortable with the stimulus, but really a better goal would be to keep going until the animal is excited and happy about seeing the stimulus. That is the point at which you can definitively show that you have changed how the animal feels about the stimulus.
Why don’t more people do this? Perhaps it’s because they don’t realize the power of CC, or perhaps it’s because they don’t know how to do CC well. Classical conditioning sounds simple, but there are some important details. I mentioned earlier that the order and timing matters. The order is conditioned stimulus -> PAUSE -> unconditioned stimulus. If you do both at once, it won’t work. If you do them in the wrong order, it either won’t work, or you can “poison” your unconditioned stimulus by associating it with something aversive. Here are some other tips:
- The interval between the CS and US matters. The CS should precede the US by 1-2 seconds
- longer intervals sometimes work, but not shorter
- the conditioning rate is strongest when the contingency is near 100% – not being consistent is a BIG DEAL BREAKER. Don’t expose your animal to the trigger if you can’t be there to train.
- The first pairings are the most important. Get your momentum going.
- Longer inter-trial intervals are important. You don’t want the animal to be expecting the stimulus, so this also means…
- Avoid rhythmic trials. People have a tendency to fall into patterns so if you are doing a number of trials in a set period of time, plan the timing so the stimulus comes at random intervals.
- Avoid competing stimuli or using weak unconditional stimuli (use something the dog really likes)
- Not transitioning to operant conditioning when the dog is ready.
She finished with a little discussion on how operant techniques like DRA, DRI, DRO, stimulus control and using an LRS fit into the picture. Basically these are all ways to start to sort through and shape the behavior you want once the animal is offering something other than its initial response to the trigger.
I want to make some comments on this material and how to apply it to horses because I found it triggered a series of related thoughts about using classical conditioning with my own horses. Rather than mix them in with her material, I thought I would share them at the end.
I do want to start by repeating something from yesterday. In her talk Kathy’s focus was on the progression from classical conditioning to operant conditioning as part of a long term training plan. So she was talking about using them sequentially, one and then the other. It’s one way to think about combining the two types of learning. But it’s also important to recognize that in most cases the two are happening at the same time. As Bob Bailey would say “Pavlov is always on your shoulder.” So there are two ways to think about using both classical conditioning and operant conditioning as part of your training plan.
If you want to learn more about the overlap between classical conditioning and operant conditioning, I wrote about it last year after attending Clicker Expo 2013 (see my notes on my website). This subject was discussed by Jesús Rosales-Ruiz who looked at the differences between classical and operant conditioning and argued that it’s hard to separate them. In many of the examples of classical conditioning that he showed, there was an operant element too. He had a video of using counter-conditioning to change a small dog’s reaction to a previously frightening stimulus. The dog was fed using a standard counter-conditioning protocol and its response to the trigger changed.
But it didn’t change in random ways. As the session progressed, it was clear that some behaviors were now happening more than others. Even though the intent was not to select out or shape a particular behavior (using operant conditioning), there was operant learning going on during a classical counter-conditioning session. He had a few other examples, all showing that when you are working on operant conditioning, classical conditioning is also happening and vice versa.
I see this with my own horses when I try to feed without clicking. I think that once an animal knows about a marker signal and how the “clicker training game” works, then they are very quick to pick up on any information that tells them what behavior is being reinforced. So if I really want to avoid reinforcing a particular behavior, I have to be very careful about my body language, food delivery, and monitor the horse’s behavior.
This doesn’t mean it is “wrong” if a particular behavior is being reinforced more, just that it’s helpful to be aware of it. And there are advantages. It means that I get some of the benefits of operant conditioning when I am working on classical conditioning and vice versa. This is something I already knew, even though I didn’t think about it quite that way. Have you ever noticed that the objects that are used in clicker training exercises become important in their own way? I think most of us have experienced a horse that brightens up or gets excited at seeing a target stick, mat, or some other object that is associated with a specific behavior. Those objects have taken on value through classical conditioning, which was happening while we were training behaviors with operant conditioning. If you’ve ever had a horse eagerly take you to a mat, then you know the power of classical conditioning.
Since some classical conditioning is already happening, is that enough? Or are there benefits to focusing more on classical conditioning? One of the things that Kathy emphasized was that we don’t want to use classical conditioning just to get the animal to tolerate or accept a previously aversive stimulus. We want to use classical conditioning until the animal is excited about seeing it. I love this idea and now I am wondering why I haven’t used classical conditioning more. Thinking about it, I think there were a few sticking points.
One of the sticking points for me with using classical conditioning was that I was pretty wedded to the “don’t treat without clicking” mantra. This is partly because that “rule” was one of the things that convinced me to give clicker training a try. I liked the idea that the horse would not expect food without being clicked, especially since I had small kids and possibly nibbly ponies. I think this rule is really important to avoid muggy horses and for a long time I really stuck to it (meals excluded). But it did mean that training protocols that used food without clicking were ones I tended to shy away from.
However, in recent years I have played around with just using food (no click) and I have not seen any problem with doing it. I found that my horses were so polite around food, from all the years of clicking before treating, that just using food was not an issue. I do have to be careful about my mechanics (if I get sloppy, they get sloppy) and there are some situations where I chose to place food in a bucket or on the ground instead of hand feeding. But overall I like being able to deliver food with or without a click as it gives me more flexibility and I think it also made my horses even more polite about food.
The other sticking point was that I think I was confusing using food to create positive associations with conditioning a new stimulus. When I first started using classical conditioning, I would use food as a way to change how an animal felt about something that was happening. So if I wanted to work on trailer loading, I would load the horse and feed it while it was in the trailer. Or I might feed it while the vet was there so that having the vet there was a good thing. I think there is value in this, but I don’t think it is quite the same thing as more focused classical conditioning where the trainer identifies the conditional stimulus and deliberately transfers the associations from the unconditioned stimulus. Doing the latter is a much more focused approach and requires a lot of pairings (with the correct order and a pause between), but also has a much greater likelihood of really changing how an animal feels when the stimulus is presented.
Recently I came across an old handout from my early days of giving presentations on clicker training. As part of it I have a little chart titled “Four Ways to Use food.” They are listed as operant conditioning, classical conditioning, bribery, and indiscriminate hand feeding. I used to present this information so that people could start to recognize the different ways they use food. It was a nice way to get people to expand their thinking a little if they were in the “all food treats are bad” camp, or if they had never thought about when and why they fed treats. At that time I spent all my presentation on the operant conditioning part, but next time I think I’ll take a few moments and explain more about classical conditioning now that I know more about how the two work together.
Susan Friedman “Reducing Behavior with Reinforcement – Oxymorons and other jumbo shrimp”
The first part of this talk was about terminology, and particularly about the four quadrants (positive reinforcement, negative reinforcement, positive punishment, negative punishment) This information is available readily elsewhere (on my website) so I am not going to include it here, but I will make a few comments.
One is that Susan says she does teach the quadrants with their scientific names, even though there is some confusion over terminology because it conflicts with the way the words are commonly used. She thinks it’s important that we all use the same words so we can understand each other. Later she said there are really only about 20 words you need to learn to understand the technical jargon. That was meant to be encouraging.
When she is analyzing behavior, she decides first if a behavior is being reinforced or punished, and second whether something is being taken away or added. She likes using the words negative and positive with reinforcement because they make her pay attention to whether the animal is trying to get something or get away from something.
When she looks at behavior with a view toward analyzing the ABCs (antecedent, behavior, consequence), she asks three questions:
1. What is the focal behavior (yes, you have to pick ONE behavior to analyze)
2. Do you predict the animal will do the behavior more (reinforcement) or less (punishment) in the future?
3. Was something added as a result of (contingent on) the behavior (+ ) or removed (-)
A few comments on the ABCs:
In addition to having the antecedent that immediately precedes the behavior, there can be distant antecedents. These can come from an animal’s history and past learning. One way to think about antecedents is that they “set the occasion for the behavior.”
She mentioned the importance of predicting behavior. When we can predict what will happen, then we can plan for it. One way to get a better understanding of the quadrants is to watch a video clip of an interaction and pick out specific behaviors to observe. As you watch the video, can you see which behaviors are being reinforced and which behaviors are being punished? I thought this was a great suggestion because we all tend to make assumptions about how behavior is affected by different consequences. We might find we are more accurate if we spend more time observing first, or at least video our training sessions so we can observe what really happened. It’s so easy to think you know what is going on and then watch video and see that something entirely different is happening.
In a typical ABC analysis, the emphasis is on observable behavior. This may make it sound like there is no place for considering the emotional response of the animal, but this is not so. The emotions associated with a behavior will show up as an observable part of the behavior. We just have to learn to look for them. She used the word “valence” to describe the emotional component of a behavior.
I looked it up and got this definition. “Valence, as used in psychology, especially in discussing emotions, means the intrinsic attractiveness (positive valence) or aversiveness (negative valence) of an event, object, or situation.” Valence is not a word I had used before, but also listed was “ambivalence” which I think we all recognize. I thought it was interesting that the word ambivalence (mixed feelings) is commonly used but the word valance is not.
Susan asked “Does Valence Matter?” YES!
- we behave for an effect
- effect makes us happy
- WE want/like/desire to get appetitive outcomes
- we want/like/desire to get away from aversive outcomes
Valence is important, but it’s important to be accurate about how we evaluate it. There is a temptation to try and read the minds of our animals and this is not valid or possible. We can only evaluate valence by observing behavior. If you want to learn more about emotions in animals and how to understand them from a scientific point of view, look up the work of Dr. Jaak Panksepp. I have some notes on his talk at ORCA on my website which will get you started, and then there are more resources available on the internet. Susan’s point was that yes, valance matters, but we need to evaluate it objectively.
Do you remember that the title of this talk was about oxymorons? An oxymoron is a word or phrase that seems to contradict itself, like jumbo shrimp or negative reinforcement. Or… using reinforcement to reduce behavior. This was one of the key points of her talk. She showed several ways that we can use reinforcement to reduce unwanted behavior. This included noncontingent reinforcement and differential reinforcement.
Noncontingent reinforcement is a slightly controversial term because reinforcement is contingent by definition, but it has been adopted to mean a training protocol where the reinforcement is delivered independent of the behavior. The idea is to break the connection between an unwanted behavior and any possible reinforcement for it by providing the reinforcer on a regular basis, but for other behaviors. This takes away the functionality of the unwanted behavior.
The reinforcement is often provided at regular intervals. She had an example of a puppy on a mat. The trainer wanted the puppy to lie down on the mat so she reinforced her for being on the mat at regular intervals, regardless of her position on the mat. There were some criteria (mostly just being on the mat) but they were very broad and the reinforcer was delivered at regular intervals regardless of what the puppy was doing (as long as she was on the mat). The food was presented to encourage a down, but it was not contingent upon the puppy being down.
She listed the following steps for using noncontingent reinforcement:
- Identify the reinforcer that maintains the problem behavior.
- Deliver the reinforcer before the problem behavior occurs and continue to deliver the reinforcer on whatever fixed time schedule makes the behavior less likely to occur.
- Withhold reinforcement if the problem behavior does occur.
- Gradually thin schedule of reinforcement once problem behavior is at an acceptable low.
Another way to reduce behavior with reinforcement is to use differential reinforcement. Differential reinforcement refers to a general procedure and then there are some specific “types “of differential reinforcement protocols that are usually written out as initials. DRO is differential reinforcement of OTHER behavior. DRI is differential reinforcement of INCOMPATIBLE behavior. DRA is differential reinforcement of ALTERNATIVE behavior.
Differential reinforcement is two procedures working in tandem. One behavior is being strengthened (reinforced) and the other behavior is getting weaker (extinction). All behavior is the result of differential selection of behavior by consequences. She called differential reinforcement “the ‘other’ kind of natural selection.” We are biologically designed to change our behavior in response to consequences.
Differential reinforcement is part of shaping and stimulus control. When we stop reinforcing certain approximations because we have changed the criteria, we are using differential reinforcement to select out those approximations that do meet criteria. When we only reinforce behaviors that happen on cue, then we are using differential reinforcement to select out cued responses and extinguish uncued responses.
As part of the discussion on getting rid of unwanted behavior, she did talk a little bit about other ways to reduce behavior (LRS, punishment, and extinction). LRS (least reinforcing scenario) is a way of responding to an incorrect response to a cue. The trainer pauses for a moment (very brief), just enough to break the flow of behavior and then immediately recues a well-known behavior to get reinforcement going again. A correctly done LRS makes it less likely that an incorrect response will be inadvertently reinforced (this can happen if you just continue on) and reduces extinction induced aggression.
The use of positive punishment has been well studied and it has been shown that it is not a good way to reduce unwanted behavior. It is generally the most intrusive and has side effects such as:
- generalized aversion
- doesn’t teach the learner what to do
- doesn’t teach the teacher what to teach
- damages the relationship between learner and teacher
- requires the biggest guns
- reinforces the agent
Extinction is another option for reducing behavior. It works because in extinction you decrease a behavior by permanently withholding the maintaining reinforcer, but it also has unwanted side effects and can be hard to do (resurgence and recovery are problems). You have to be able to withhold all reinforcement and sometimes the animal finds its own reinforcement. She calls this “bootlegging reinforcers.” Extinction is more intrusive than reinforcement, causes an initial burst of intensity (extinction burst) and can be hard to do with behaviors that are reinforced intermittently. It is most effective if you can ignore the behavior the first time it happens so that you never give it a function. I will add, from personal experience, that it can be really hard to control all the reinforcers for a behavior when working in an animal’s normal environment.
When you can’t control all the reinforcers, then the “Matching Law” comes into effect. The matching law says that “Given a choice between two behaviors, the relative proportion of responses matches the relative proportion of reinforcers earned by that alternative. “ Animals do what pays the best. If an animal walks away, it means we are not reinforcing enough.
So with all this information about ways to reduce unwanted behavior, how do we proceed?
Step 1: operationalize the behavior
what does it look like?
what frequency, duration, intensity?
step 2: identify antecedent influences
under what conditions does it occur?
step 3: hypothesize (or test) the function of the problem behavior
what is the maintaining reinforcer?
steps 2 and 3 usually require more careful observation, data collection and ABC assessment. As part of this you may need to do a functional analysis to find out what is maintaining the behavior.
step 4: select DR (differential reinforcement) procedure
the target behavior should serve the same function as the problem behavior
The target behavior must give the animal at least the same amount of reinforcement
reinforce at a high rate, consistently
the replacement behavior should be the one the animal already does fluently
generalize to different conditions
She had some examples of this with lions and goats and showed how to write up a behavior change model chart listing antecedents, possible behaviors and consequences. The chart starts with the existing ABC sequence but then you can add alternative paths to different consequences. Then she showed a video clip of a boy working with a parrot that had a history of biting him. He started by just feeding the parrot while it was in its cage and by the end of the training (over a period of weeks), he was holding the parrot and asking for behaviors.
- Here are a few other quotes from the talk that I wanted to share.
- “Repetition builds confidence”
- “No animal behaves for no reason. It’s a functional behavior for that animal.”
- “Don’t extinguish communication in any form.”
- “… in a position where you can hear the door of punishment creeping open.” – I think this started with “you may find yourself” but I’m not sure. I just loved the image of being careful because the door of punishment is creeping open.
Susan Friedman’s website is www.behaviorworks.org. You can find more information on her work, articles and courses on her site.
Karen Pryor “On Extinction”
This was the first of the shorter (45 min) talks that Karen gave. Extinction is something that often gets mentioned as a side note in discussions on operant conditioning and learning theory, but this was an in-depth look at what it is and whether or not we should (and are) using it.
Karen started with the statement “Extinction takes place when reinforcement stops. The term “extinction” refers to the behavior that was being reinforced: theoretically the behavior the reinforcement is supporting stops, too.” We say a behavior is going into extinction when it is no longer being reinforced and the frequency at which it occurs decreases, hopefully until it is not happening at all.
I think most of us become aware of extinction by experiencing or observing an extinction burst. This is the escalation of behavior that occurs when reinforcement first stops and the animal (or person) has an emotional outburst because something that had been working is no longer doing so. During an extinction burst the frequency of the previously reinforced behavior increases (there is more of it, often accompanied by a change in intensity) and it will be accompanied by a variety of emotions, none of them good. Karen says it is common to feel annoyance, anxiety, anger, rage and despair. This occurs across many species. She had examples of extinction bursts happening with rats, pigeons, dogs, and fish. Examples with people included the ever familiar examples of kicking the vending machine and trying to get your car to start.
In addition to the emotions that accompany the extinction burst itself, extinction can lead to the feeling of grief. This is the sense of loss that can affect future behavior. Examples of extinction that can lead to grief are the break-up of a relationship, losing your job, and leaving home. B.F. Skinner is the one who pointed out that leaving home means that a lot of extinction is happening at once because you are leaving so many familiar things behind, and that includes all the reinforcement that accompanied them. This grief can be quite long lasting. Karen had a story about a dolphin that sulked for 3 days after an accidental extinction event.
The rest of the talk focused on three questions about extinction. They are:
- Should we use extinction in training?
- Can we (and should we) build resistance to extinction?
- What are alternatives to extinction?
Should we use extinction in training?
Extinction is about decreasing (or stopping) a behavior by removing reinforcement. Right off, as a clicker trainer, this should raise a little red flag in your mind. It’s always better to focus on what YOU DO WANT than it is to train by thinking about what you don’t want. You can see this more clearly by looking at Karen’s list of ways that extinction is used by some trainers.
- As a research tool (on/off experiments)
- As punishment (taking away the reinforcement is –P)
- To end an unwanted behavior
- To get more variability when shaping (extinction bursts generate a lot of variation, especially in intensity)
You can see that most of these (potentially all) are training methods that are going to be emotionally difficult for the animal and have possible adverse effects. Not only that, but it’s not clear that they are even effective. Karen said that “the learner always remembers what used to work. The memory is always there.” You can’t replace learning. You can only cover it up with new learning, so extinction is never complete. It really only works “to all intents and purposes.”
Punishment in particular is very ineffective. Karen clearly talks about both +P and –P. She said that taking away reinforcement for a behavior is –P. Since extinction is about decreasing behavior, this makes sense and you would get a lot of the same emotional fallout that accompanies punishment.
Punishment tends to suppress all behavior or just change the environmental context or triggers for the behavior. The animal may learn not to do the behavior if someone is watching, but continue to do it in other circumstances. If the underlying reason for the behavior or reinforcement is not changed, then punishment can end up suppressing one behavior, but the animal might replace the undesired behavior with another equally undesirable behavior.
When you use punishment, emotional responses are inevitable and they include:
- A loss of trust
- hostility to the trainer/teacher/supervisor
- extinction of all responses (learned helplessness or depression)
The bottom line is that we don’t want to intentionally use extinction in our training… but there’s a little catch here. According to Jesús Rosales-Ruiz, extinction is part of the shaping process. Karen mentioned that old textbooks talk about using extinction in shaping. When you stop reinforcing a behavior because you have changed the criteria, the old behavior goes into extinction. When I first learned about shaping, it was described as “riding the extinction bursts. “ I would click for a response a few times and then withhold the click and see if the horse offered something else.
I think the thing to keep in mind here is that extinction happens all the time as we learn new things it is probably part of the learning process. The important thing is how the animal feels about what is happening. If the animal understands the shaping process and is still working toward behavior, then the temporary lack of reinforcement becomes information and is not accompanied by the emotional angst that can accompany extinction. This is yet another argument for shaping in small steps so as to avoid stressing the animal.
This leads us right into…
Can we build resistance to extinction?
If we accept the last statement that there is some extinction as part of the shaping process, and we also recognize that there will be times when we need an animal to work for longer periods of time before being reinforced, then we have to look at extinction not just as something to be avoided, but as something that is part of learning. That means that we need to train our animals so that they understand how to handle a change in the type of reinforcement, a thinner schedule of reinforcement, or the absence of reinforcement at moments in the shaping process (I could argue this one, but I’m not going to get into that here).
Karen said that this is where variable reinforcement schedules are often used. Some trainers use them so that the animal will continue to keep working even if a behavior is not reinforced each time. The animal is trained that reinforcement does not happen for every correct answer, but only for some correct answers. This leads to more persistence because when a behavior doesn’t pay, the animal will try it again, hoping that it will pay the next time. It’s the same idea as gambling or trying to get an old car to start. You keep trying because sometimes it works. The reality is that a lot of things in life are on variable reinforcement schedules. She used the examples of playing solitaire (sometimes you win), batting practice (sometimes you hit it), your boyfriend’s phone call (sometimes he calls?), and the dog’s response to cues. None of those things happen every time you try, but they happen often enough that you keep trying.
Variable reinforcement schedules are used to teach the animal that if they keep doing behaviors, reinforcement will eventually happen. And they can be very powerful. Most addictions are maintained by variable reinforcement schedules (gambling, smoking, ..). Variable reinforcement schedules can be important for some types of working dogs where reinforcement happen at the end of a long sequence of behaviors. Karen used the example of search and rescue dogs. I suspect those dogs get quite a bit of reinforcement along the way, but their main reinforcement is finding someone at the end of the search.
Just as a little note, not everyone agrees on the value of variable reinforcement schedules. Bob Bailey says he doesn’t use them. There have been various talks at past Clicker Expos about the difference between variable reinforcement schedules and reinforcement variety. You might want to read up on that material (it’s on my website) if you are curious about variable reinforcement vs. reinforcement variety.
What are alternatives to extinction?
From what I have written above, you can probably guess that intentionally putting a behavior into extinction is not often a good training choice. If you want to get rid of unwanted behavior, there are better ways. Karen lists some alternatives to extinction (and punishment) here:
- 1. Teach an incompatible behavior
- 2.Bring the unwanted behavior under stimulus control
- 3. Manage the behavior until a new behavior is under stronger reinforcement (she suggested this for infantile feeding behaviors in puppies and kittens that disappear with developmental changes)
remove the stimulus
- 4. re-associate the cue with something else (change the trigger so it means doing a different behavior)
- 5. remove the reinforcers (I know this sounds like extinction – what she means is change something in the environment so a behavior is no longer reinforcing – her example was to keep the counters clean to avoid counter-surfing). Perhaps this is more of a management solution.
At the end of the talk she pointed out that the use of extinction as part of many laboratory experiments is changing and researchers are looking for alternative ways to conduct experiments so the subjects don’t experience as much stress.
Michele Pouliot,“Pace, Place and More”
I attended the lab, but not the lecture for this topic. I do have the notes for the lecture so I am going to pull out some of the relevant information because it also came up in working with the dogs.
The subtitle for this lecture/lab was “Strategic reinforcement delivery.” It was about the importance of learning how to deliver your reinforcement. She said that every reward is a “delivery event” and the reward event greatly influences behavior. She illustrated this with two equations:
- Effective Reward Location + Effective Delivery Method = Powerful Training.
- No Reward Strategy + Animal Controls Delivery = Not so Powerful training.
Included in the “Not so Powerful” training is training where there are safety concerns. It’s important that the reward strategy is comfortable and safe for the handler and the dog.
When adding reinforcement strategy into your training, it’s important to take into account the following considerations:
- The individual animal: your relationship, the value of reinforcers available, mechanics of taking rewards
- You as a trainer: your abilities and training goals
- Each Behavior can have a different reward strategy
In addition, it’s important to note that Strategic reinforcement does NOT ask for additional behavior. After the click, the “delivery event” flows smoothly and does not “require” more behavior before the dog receives the reward. For example, if the dog is clicked for heel position but moves out of heel after the click, the handler should not require the dog to move back into heel position before giving the reward.
Reward location is the key to effective reinforcement strategy. There are several location strategies that she talked about and showed in the lab including:
- resets dog (places dog in best location for next repetition)
- rewards where dog was at the time of the click (adding value to that location)
- rewards at a specific location (moves the dog) which supports the goal behavior –
- rewards where the dog is at the time of delivery (may be different location from where the click happened).
The Speed of Reward delivery is also important.
TOO FAST: dog is startled by hand movement or doesn’t see where the food goes
TOO SLOW: dog doesn’t connect the reward to the marker behavior
A powerful reward event begins after the click and is a smooth process from start to actual delivery.
Between TOO FAST and TOO SLOW, there is room for a variety of speeds. The handler can influence the dog’s energy by moving more quickly so that her movement builds the dog’s enthusiasm. Or she can deliberately slow down so that the reward event lowers the dog’s energy. In addition to the handler moving, the reward itself can be thrown. This works better with rewards that are easy to see, stay in one piece and have a low “bounce” factor. It also helps if the handler has reasonably accurate throws (practice, practice, practice!).
More on the delivery process
She shared a number of specific food delivery strategies that she uses. I have grouped them according to her four general categories above because that made it easier for me to understand different applications of the same general strategy. It’s possible she wouldn’t group them exactly the same way I did, and it’s good to remember that a trainer can use the same food delivery technique for different effect, depending upon how the rest of the exercise is structured.
- Reward Location Goal: resets for next repetition. Initial movement toward goal behavior is more heavily reinforced. In teaching a dog to put his hind feet on platform, the dog is fed off the platform so he is set up to back on it again.
- Delivery Process: Relocating Animal via Preset Rewards. Remote controlled feeders or feeding stations can be used to reset the dog before starting the next behavior.
- Delivery Process: Location: Come n Get it! Upon click dog comes to reward at its location (commonly the handler) – this is most often used as a reset, but could be used to reinforce a location or goal, depending upon the behavior.
Feeding to promote the goal behavior:
- Delivery Process: Relocating Animal Handler Placed Reward. Reward via placement that prompts the dog to move to the reward. The trainer can place food on a platform to encourage dog to step on it.
- Reward Location Goal: Reward completion point. Final behavior position is more heavily reinforced . The trainer can feed down and back between front legs to shape a bow.
- Reward Location Promoting Goal behavior: in training a “down” – feeding low to encourage the dog to go down and stay down.
Feeding to reinforce the clicked behavior (or location):
- Delivery Process: Reward Stationing. Place or position from which most behaviors are cued. The dog can be fed on the station if the trainer wants the dog to hold the position.
- Delivery process: Pizza delivery, no need to travel – Direct Delivery – prompts dog to remain at click location.
- You can also use a reward location that encourages the dog to check in with the handler. This is helpful if training behaviors in the presence of distractions.
Feeding where the animal is (even if it moves after the click):
- Delivery Process: Protected contact. Animal remains in position behind a barrier and waits for the reward to be delivered.
Notes from the lab:
One of the activities in the lab was to teach the dog to put its head in a pop tube. A pop tube is a plastic tube that can be lengthened or shortened and curved into different shapes. When teaching a dog to put its head in the pop tube, she had the trainer feed by putting her hand through the pop tube so the treat came out of the pop tube. This encouraged the dog to orient toward the pop tube in such a way that it made the desired behavior (head in tube) more likely.
She had everyone teach one behavior doing the pop tube and then each person with a dog got to pick a new behavior to train (or one to improve) and she had suggestions for training it and how to deliver the food. Some of the ones I observed were:
- The bow: feed low and back between the dog’s front legs to encourage it to drop down more.
- backing: feed where they are when you click, make sure you click while the dog is moving or…use a platform and have the dog come forward to the handler to reset him so he backs up on to the platform.
- paw lift: one person wanted to get more height in the paw wave and she had her feed the dog in position (sitting) so the dog was ready to go again. She also added that she never withholds the click to get height. Click the rise and energy.
There were a few other behaviors where the treat delivery was more about taking advantage of the environment and dog’s natural tendencies. If the dog tends to orient toward your treat pocket, then you can use that. Use walls, corners, etc… to make the dog more likely to orient in a specific direction. If the dog is staring at you, and not interacting with the object, then feed so the dog is not facing you and then you can move out of their space so that the object is the first thing they see when they are done eating.
Additional tips and comments:
- The food reward hand should have no information.
- Remove the object so you have a fresh presentation, more like “It’s available, it’s not available”
- Accidental correct answers can mislead us into thinking the dog knows it.
- Don’t be afraid to build a lot of reps to prevent them from doing a known behavior. This is important if you are teaching a new behavior when a dog often chooses to interact with objects in a particular way. She had an example of this with her dog who likes to put her paws on things. She wanted a chin rest so she held the chair up high enough that the dog was unlikely to use her paws and didn’t lower it until she had A LOT of reps of chin on the chair.
What to do if it’s not going well….If you aren’t making progress, stop and THINK. Clicker training done well should be fast. We have a tendency to keep training and hope it will get better. Don’t lower the criteria to get things happening – stop and THINK. (I’m going to add a little note here that this comment should be taken in context. I don’t think Michele is saying all behaviors can be trained fast, but shaping a clickerwise dog to do a simple behavior should be quick).
It’s always fun to see how what I learn at these conferences affects my training. Yesterday I took out Stella, who hasn’t done anything except routine handling in about 6 months. We worked a little on targeting. I’ve done a lot of following a target work with her and her stationary targeting needed cleaning up. So I started with a new target (that would be a stationary target) and instead of feeding and waiting before I presented the target, I fed her right over the target so she couldn’t miss it. This tightened up the loop and she stopped trying to walk forward to the target and just started touching it. After I was done, I realized I thought of that change in presentation of the target because of Michele’s lab where she said to set it up so that the object is the first thing they see when they are done eating.
Kay Laurence, “The Art of Practice”
Does the idea of “practicing” make you cringe? As a mother, I know that all I have to do is suggest someone practice something and they are headed the other way. Why is that? Is it the word itself which has become associated with repeating boring drills? Or is it that we are not taught how to practice well so that practice is both productive and enjoyable? This was the subject of Kay Laurence’s lecture on practicing.
What is practice? She says “It is the purposeful and single-minded application with the intent to get better.” Not only that, but “It isn’t the thing you do once you are good. It is the thing you do that makes you good.” (Malcolm Gladwell, “The Outliers”)
In talking about practice, Kay is including many different types of behaviors. There are those skills we practice because they are important for our performance goals. There are also those skills we practice because they are life skills. Practicing is not limited to pet or performance dog training. It is about developing competence, familiarity, fluency, focus, understanding, consistency. It is also about preparation. It takes time and effort to train behaviors to meet these criteria and a big part of it is learning to see the environment as the dog sees it. She said her new slogan is going to be “Be More Dog” which she got from a commercial. I’ll put the link in a comment if you want to see it.
So why do people hate to practice? Here are some common reasons:
- Poor technique: if you practice with good technique and the science is sound, you will make progress. On the other hand, if your technique is poor, practice will be frustrating and the lack of progress will make you less likely to continue with it.
- Training for the wrong goal: If you are training under a time limit or to please someone else, then your ability to practice well will be compromised. Time is a bad reason to push something along, whether it’s because you only have 5 minutes or because you need to be ready for a competition tomorrow.
- Joyless repetitions: Some people think the dog doesn’t like repeating behaviors, but they do if they know the behavior and get paid for it. A big part of this is that the trainer learns to deliver the treat with delight. If the trainer acts bored with the repetitions, the dog will pick up on it.
- Unprepared: It’s important to take the time to prepare the environment. Make sure that all the equipment you need is ready and that your treats are prepared. Preparing also includes having a training plan.
- Selectively practice: Some people do practice, but…they tend to practice more the things that they do well and the avoid practicing the things that are more challenging. So they may have some things they like to practice and some things that they hate practicing.
Any combination of the above reasons can lead to avoidance. The trainer doesn’t put the time into practicing that is necessary. Kay said that the 10,000 hour rule also applies to practicing. You have to do 10,000 hours of practicing to achieve mastery in something.
She gave some tips to make yourself more motivated to practice, and more successful when you do it. The first part is making a plan. There should be a plan for you and a plan for the dog. The plan should identify what needs to be prepared ahead of time. This could be setting up the environment so that you have everything you need. It should also include removing potential distractions.
As part of preparation, you should rehearse mentally and physically. There are a lot of trainer skills that can be practiced before the dog comes out. She recommends practicing throwing treats until you can easily throw a treat and have it land where you intended. You should also do some mental preparation so that you start practicing with a good attitude and are be fully committed to what you are doing. She used the phrase “bring your whole-self.”
You should have a way to document the session (record data or videotape) so that you can analyze how things went. She pointed out that it’s important to be honest about what really happened (trainer errors vs. dog errors vs. other random stuff). If you are videotaping, it’s helpful to watch the video a few times, focusing on one thing each time. Then apply what you have observed to the next training session. This requires self-discipline and self-awareness. The dog has no knowledge of the outcome so the trainer needs to pay attention to the bits along the way. Plan to train in small bits and batches so you can pay attention to detail and balance.
In a session, the details are important. Your motor skills need to be fluent. Good mechanics include being able to click with good timing, deliver food appropriately, manage and manipulate targets, proper use of signals, being disciplined about correct use of cues, and balance. An example of a detail that matters is how you position the dog so that you can observe the behavior. When teaching a bow, she clicks when the dog’s elbows touch the floor. If the dog is in front of her, she can’t see it. She has to position the dog sideways relative to her so that she can time the click correctly.
She had some specific comments about clicking. Dogs pick up on uncertainty, so don’t click if you are uncertain whether or not you should click. If you do click, then deliver the reinforcement promptly. There should be no “yes, buts” following the click where you click and then ask the dog to do something else or waffle over delivering the good because of the way the dog takes the food. If there’s a problem with some element in the training that’s not related to your specific goal for that session, then make note and work on it another day.
Most of this has been about the “science” of practice. What about the “art” of practice? The art of practice includes:
- compliment the learner’s pace
- begin with warm-up behaviors
- be relaxed but alert
- be comfortable
- build organically
- the art is building a conversation. It will develop with its own natural rhythm.
- being aware of balance: doing what needs to be done and structuring the learning but filling in with creative inspiration.
“Practicing is very much about focusing on one thing at a time and working gradually toward the experience of the whole.” (Madeline Bruser)
Kay writes “The talent is being able to combine the science, good technique and the natural instinct of listening to your learner and providing what they need at this precise time. Attractive solutions are often avoiding doing the work. But…it is the work, the understanding, the practice that is the solution.”
I am reminded of a quote from Sue Ailsby that I read long ago. She was listing important steps that lead to success in training and one was very simply “Do the work.”
But…keep in mind that Kay also says “Training should be a joy.” It’s our job to reconcile these two things so that practicing is a joy for all parties involved.
If this has piqued your interest in practicing, there are plenty of good resources out there on the subject. Kay often recommends a book called “The Art of Practicing” by Madeline Bruser. I haven’t read it, but I’ll put it on my list. A few years back I read a book called “The Talent Code” by Daniel Coyle and he writes a lot about “deep practice.”
Here’s the Be More Dog commercial: https://www.youtube.com/watch?v=iMzgl0nFj3s.
Jesús Rosales-Ruiz, “Using Resurgence to Your advantage.”
Jesús started with some definitions. If some of this terminology is new to you, I would just read them, read the rest of the article, and then come back to them later. They will probably make more sense once you have seen examples of them. Also, if you haven’t read the notes from Karen Pryor’s talk on extinction, I suggest you do so. They add a few more details on extinction.
What is resurgence?
- Resurgence refers to the reappearance of previously learned behaviors when the present behavior is not capable of getting reinforcement. Usually the learner reverts to older forms of the response which were once effective.
- Also known as regression
- It’s a retreat from one’s more recently acquired behavior to that of an earlier period. (From Keller & Schoenfeld (1950)
Those were the “original” definitions. Since then, behavior analysts have made a few distinctions between resurgence and regression.
- Regression – The reappearance of previously extinguished behavior during the extinction of more recently reinforced behavior. Catania (2013)
- Resurgence – The occurrence of a previously reinforced behavior (Behavior 1) when a more recently reinforced behavior (Behavior 2) is undergoing extinction. Cleland, Foster, & Temple (2000), (e.g. Epstein, 1983; Wilson and Hayes, 1996)
The significant thing to note here is that in regression the behavior that reappears is one that has been extinguished (intentionally let go to extinction), but in resurgence, the behavior that reappears is one that is still being reinforced.
Since the difference between regression and resurgence is related to extinction, he defined extinction.
Extinction: When reinforcement is no longer forthcoming, a response becomes less and less frequent in what is called “operant extinction.”
Extinction is one way to get rid of behavior, but it can come with an emotional cost. This is because the process of extinction goes through several steps. These are:
- Response bursting
- emotional responding/attack
- historically reinforced behavior re-occurs
- pauses followed by short response bursts
- reduced overall responding
Extinction bursts (the “response bursting”) are seen in all species. Pigeons will coo, flap their wings and engage in other emotional behavior. Humans will try harder, often increasing the intensity of the behavior, accompanied by emotional outbursts such as a loud voice, hitting something, etc…
Jesús compared the feelings that accompany extinction with those of Loss or Grief. If the behavior has been heavily reinforced in the past, the learner may go through the stages of grief (shock and denial, anger, depression and detachment, dialog and bargaining, and acceptance.)
Not every extinction event creates the same emotional outburst. There are a number of factors that can affect how the learner reacts when a behavior no longer leads to reinforcement. One of the most significant factors is if the learner has choices. The more other choices there are that could lead to reinforcement, the more quickly the learner will find other ways of earning reinforcement, and the more quickly she will abandon the behavior that no longer works.
Jesús also mentioned that variable reinforcement schedules can minimize the effects of extinction. If the learner is used to a pretty lean reinforcement schedule, then it is going to be less affected when reinforcement stops, compared to a learner that expected reinforcement for every correct effort. I am adding a note here that variable reinforcement schedules also make a behavior more resistant to extinction. The fact that they make the learner less emotional is probably related to the fact that extinction is actually less likely to be happening in the absence of reinforcement when the learner is on a VR schedule. So the learner may not experience the effects of extinction until there has been a very long period of no reinforcement.
So extinction is bad? Well…not exactly. There are degrees of extinction and most learning involves some level of extinction as the criteria are changed so that previously reinforced behaviors no longer get reinforcement. The reality is that extinction is going on all the time when we learn new behavior. It is also happening when we add stimulus control to a behavior because there will now be incidences of that behavior that are no longer reinforced.
Animals can learn to deal with extinction as long as other opportunities for reinforcement are available. This is one of the big differences between studying extinction in the lab vs. in real life. In the lab, the learners often have few choices. When a behavior undergoes extinction, it means that ALL reinforcement stops or that the animal has to learn something totally new to get reinforcement going again.
His mention of lab vs. real life made me remember that Karen Pryor mentioned reversal studies as examples of training that uses extinction. In a reversal study, the animal is taught to do one behavior (behavior A) to get reinforcement. This might be touching a key. Then the researcher changes so that it has to touch a different key (behavior B). Once it learns to do behavior B, then the researcher changes back to only reinforcing behavior A, and so on until the animal can go back and forth on cue (yes, there are cues). This involves extinction every time the researcher changes which key leads to reinforcement and it is very stressful on the learner. These animals often exhibit a lot of stress behavior.
Those are examples of intentionally using extinction in training, but extinction can also happen “accidentally” in training. Jesus showed a few video clips of dog training where the dog was “accidentally” put into extinction. In one of the video clips the dog that was “put into extinction” because it didn’t offer the desired behavior. During extinction the dog offered a previously reinforced behavior and then the behavior the trainer wanted. She clicked the correct response. When she repeated the exercise, the dog did both behaviors. This is an example of how behaviors that are offered during extinction bursts can “tag along” with desired behaviors.
I always thought of extinction as being about a specific behavior, but you can also have a period of extinction where no behavior is reinforced. In that case, a number of behaviors are being extinguished at the same time. Maybe this is always true, because a learner in an extinction burst is going to try a lot of different behavior. But I think there might be a difference between how the learner responds to a period of extinction vs. letting one behavior extinguish.
To study this more, he did some experiments with PORTL (portable operant research and teaching lab) using fixed rate schedules. The trainer shaped a behavior and put it on a fixed rate schedule of 5. That means every 5th response was reinforced. Then the trainer switched to a FR 30 schedule (big jump – don’t do this!). This effectively created a period of extinction when no behavior was being reinforced, but the learner tried a lot of different things. What happened was that after a few repetitions, the learner was now consistently including a lot of extra behavior that it would offer between correct repetitions of the target behavior.
Jesús didn’t use this term, but I am sure this is what is often called superstitious behavior. Superstitious behavior is behavior that becomes “attached” to the target behavior because the animal has been indirectly reinforced for doing it. It often happens when the trainer is looking for one behavior and doesn’t notice that the behavior is preceded by other behaviors in a consistent way. Instead of reinforcing the final behavior, the trainer ends up reinforcing a chain or sequence of behaviors.
The talk spent a lot of time on extinction because you have to understand about extinction before you can really understand about resurgence. In the last part of the talk he showed some video about different ways that behavior can “re-appear” during resurgence.
In the first video clip, the trainer taught the learner 4 behaviors. I can’t remember exactly what they were, but they were manipulating small objects so I am going to call them ring, button, cube, disc. The learner was taught them in that specific order so she got reinforcement for manipulating each one in a specific way until there was a trained behavior associated with each object. When she was done being shaped to do one behavior, the object as removed and they started with the net object. The only extinction that happened was during the shaping process for each behavior, not between learning different behaviors.
When all four behaviors had been learned, the trainer placed all 4 objects back out. Then she was put into a period of extinction (no reinforcement for any behavior). During extinction, she went back and repeated the behaviors in the order in which they were taught starting with the ring, then the button, the cube and the disc.
In the second video clip, the trainer taught the learner to touch objects that were placed out in a circle. The learner was reinforced for touching each object a certain number of times and then the trainer stopped reinforcing that behavior and waited for her to touch another object in the circle (the next one around). So extinction was used here to get the learner to move to another object. Once the learner had worked her way around the circle, there was a period of extinction where nothing was reinforced. In this case, the learner worked backward so she repeated the last reinforced behavior, then the one before it and so on, working backward around the circle.
The difference between these two experiments was that in the second video clip, the learner had moved from one behavior to the next after a period of extinction (the trainer just stopped reinforcing touching one object), but in the first video clip, the previous objects had just been removed (no extinction, just training a new behavior from scratch). This is actually a predictable outcome based on the way the behaviors had been trained.
I hope this makes sense without seeing the video. The point here is that in extinction, previously reinforced behaviors can come back in predictable ways depending upon their past history. This is the part that we can take and apply to our training. We can set up situations where the learner returns to a previously reinforced behavior to shape a new behavior. Clear as mud? Here’s the example Jesús showed:
- In PORTL, a toy chair (small) is placed on the table and the learner is reinforced for touching the chair. The learner touches the chair in predictable ways (touches the seat, tips it, etc…). The object itself influences how the learner chooses to interact with it (what does one do with a chair?). But the trainer wants the learner to push the chair. Nope…
- The trainer removes the chair and replaces it with a toy car. The learner is reinforced for touching the car and is shaped to push the car. This is easy because pushing a toy car is a pretty normal behavior to do with a toy car. The learner gets a lot of reinforcement for pushing the car.
- The trainer removes the car and puts the chair back. The learner goes back to touching, tipping the chair even if there is no reinforcement.
- The trainer removes the chair and puts the car back. The learner pushes the car and gets reinforced.
- The trainer removes the car and puts out a small block. I can’t remember if the learner immediately pushed the block or tried a few things first, but pretty quickly the learner figures out to push the block (this is generalizing the “push” behavior).
- The trainer removes the block and puts out the chair. The learner pushes the chair!
So what happened here? The trainer set it up so that a previously reinforced behavior (pushing) would occur in the absence of reinforcement for other behaviors. The example shows one way to get the learner to try something different and how you can “prime” the learner to try the behavior you want.
Jesús finished up with a few examples from Dr. Epstein’s work on creativity and some real life training examples that showed how you can take advantage of the reappearance of previously reinforced behavior in the absence of reinforcement. Resurgence can be used to shape new behaviors as well as to combine behaviors. The chair/car example shows how to shape the new behavior of push the chair. He had some video of Alex and Robin showing how she combined the trot and Robin added in the pose to get the new behavior of pose and trot. He also had video of Sola training passage by combining several different components so that Ember added them together to get a new behavior.
When I was watching this I was thinking that this was a brilliant way to explain to new trainers how to help learners when they get “stuck.” Occasionally when I am working with one of my horses I get into a situation where the horse is so sure it knows the right answer that it ignores any information to the contrary. I have seen this happen with people in the training game too. I have learned that continuing on without making any changes is counterproductive because they just try harder to do what they think is the right answer.
I might go work on something else or present the information in a different way, and then come back to it. Or I might do something else that gets them thinking in “the right direction,” and then come back to it. I will also do something similar for more complicated behaviors where I teach separate pieces and then let the horse put them together. That would be using resurgence to get new behaviors. With the information from this talk, I think I will be able to use it more effectively and explain what I am doing to new trainers.
Next week I will post notes on Kay Laurence’s lab on “Repurposing default behaviors.” Kay takes advantage of resurgence in her training so the report from that lab will have a few more examples of ways to use resurgence in real life.
Hannah Branigan, “Performance Prep,Foundations for Success.”
This was the first time I had heard Hannah Branigan talk. She is a KPA CTP training partner from North Carolina. I found this information on her on the clickerexpo website: “As the owner of Wonderpups, LLC, Hannah is committed to training both dogs and people with positive reinforcement methods. She has titled her dogs in conformation, obedience, schutzhund, agility, and rally.”
The subject of this talk was “Performance Prep,” or how to build a good foundation so that you can then go on to train your dog for more specific performance sports. I think that the information she presented is be helpful for any dog owner as it was about understanding canine development and creating an emotionally and physically well-rounded dog. Her talk was mostly on puppies, but she stated up front that all the material also applies to older dogs that need additional training.
The first part of her talk looked at canine development. There has not been much research on optimizing canine athlete development but there has been a lot of research on humans and some on horses. Using that information as a starting point, it is possible to draw some parallels with dogs. One thing to consider is that in children there are periods of development when strength training is more appropriate and other times when fine motor skills are easier to develop. Being aware of the stages of development can make it easier to choose when to work on different skills.
Thinking about developmental windows opens up the big question of how much variation there is between dog breeds. She had some graphs showing growth curves and time at which dogs reached maturity and there are breed differences. As expected big dogs mature slower than smaller dogs but, there are some significant differences among dogs of the same size too. This means that while breed/size is useful information, it is not enough information. You also have to look at each individual dog.
The next question was when to start teaching the dog skills specific to his sport (specializing). Are there advantages, or disadvantages to starting early? Early specialization seems like it would give you a head start, but it also has a high risk of injury, burnout and hitting training plateaus. In humans, the head start that comes from early specialization often disappears after a few years so getting an early start might be focusing on a short term goal at the expense of long term goals. She had an interesting study that showed that there’s no advantage to jumping horses at a young age. They may initially be slightly better than the ones who started later, but that advantage disappears within a year.
Late specialization does mean you can’t start competing as early, but late specialization often leads to a more diverse range of skills, talent transfer across sports, and a lower risk of injury. A well rounded individual is going to be able to pick up new sports easily because he has a solid foundation. She had a nice chart that showed the Late Specialization Model. It is a triangle with layers. Starting from the bottom, the layers are FUNdamentals, Learn to Train, Train to Train, Train to Compete, and Win. She went on to talk about each of these.
FUNdamentals: This is the foundation of all other training. The goal is to teach the dog (puppy) in a fun way so that he learns to love training. Hannah calls this building “want to.” It includes teaching basic movement skills (how to use your body), basic rules, confidence, socialization, and focus. You can invent games to play, cultivate multiple reinforcers, teach the dog about simple chains (she often uses backchaining), and teach the dog that it can control its environment through its behavior. As the teacher, you can notice what the dog likes or likes to do and start to use those behaviors or reinforcers to build motivation and stimulus control. She likes to use tracking, hunting and searching games, and puzzle games to build focus. She rarely feeds out of dog bowls.
Social interactions and focus are included in the fundamentals. Puppies should be allowed to play and engage with other dogs and puppies in well supervised ways. She had some nice video of a puppy class where the puppies were doing parallel play by playing games in the presence of other puppies and exploring the environment which was full of interesting objects and things to do. The first interactions with other dogs and people are important. Puppies can be taught early on that interaction is contingent on appropriate behavior and that checking back in with the trainer is a highly reinforcable behavior. She showed some food games she plays with puppies that teach the puppy to explore and then check back.
Within the fundamentals Hannah looks for the ABCs of Athleticism which are agility, balance, coordination, and speed. Together these all build fitness. Balance can be taught by simple exercises like raising a paw or using equipment like balance balls and rocker boards. She shapes voluntary engagement with wobble boards. It’s important that food is not used to ask the puppy to do something that might frighten it. Let the puppy choose when and how to interact with the equipment (or objects) and support its choices.
Coordination can be taught by playing games like pick-up sticks where the puppy has to step over or around objects. Hiking and walking on uneven areas is good for developing coordination. Climbing over obstacles also builds balance, coordination, and confidence. She often uses pool noodles and other soft objects so that the puppy will not hurt itself.
To build speed she plays recall and chase games. Restrained recall games where the puppy is held and then released are one way to build enthusiasm and teach a puppy to accept restraint. Multidirectional movement can be taught using games with flirt poles, backing, and through play with other dogs. The idea is to teach the puppy to be confident and aware of how to use his body so that he uses it well. These skills will actually help prevent injuries.
Learning to Train is the next layer on the Late Specialization Model (triangle diagram). In this phase, the puppy is learning about the click/treat contingency, offering behavior, stimulus control, impulse control and house manners. Even quite young puppies can start to have favorite behaviors put on cue. She had a video of her puppy and one of his favorite behaviors (jumping up) and described how she put it on cue so she could use it as a reinforcer and how this also taught him impulse control.
Training to Train comes next. This is more sport specific. It involves building an aerobic base, strength, and starting sport specific skills. Sport specific skills could include behaviors such as heeling, weaves, contacts, fronts/finishes. Many of these useful behaviors can be trained in fun and playful ways. She uses platforms and other equipment or objects in the environment to help the puppy learn to use its body in many different ways. For some people it’s helpful to think of these useful behaviors as just another kind of trick because tricks don’t come with the emotional baggage that performance behaviors do.
Training to Compete is a just adding a little more focus and conditioning to the previous skills. The trainer might start working on more sport specific skills and optimize the dog’s fitness. She does a lot of cross-training so she can condition the dog without drilling the same exercises again and again.
Training to win is the top of the triangle. In this phase you are maximizing fitness preparation with frequent breaks to prevent injury and burnout. You need to continue to maintain skills and work on proofing.
I think there are a lot of parallels here with horses. Young horses need to be encouraged to learn to use their bodies well through play and carefully structured exercises. They also need to have freedom to figure out how to move. A young horse that is started carefully, with attention paid to his physical, emotional, and mental development is going to end up having a longer and more enjoyable performance career, and more importantly, a better life.
Laura Sharkey, “Give Pups a Chance: working with aggressive puppies.”
The clickerexpo website has this description of Laura Sharkey. “Laura Sharkey, KPA CTP, CPDT-KSA, OSCT, is a pioneer in the areas of puppy training and development. With a Ph.D. in Molecular Biology from Georgetown, Laura’s strong scientific background drives her approach to dog training. Laura has more than a decade of training experience, and established WOOFS! Dog Training Center to provide owners with the resources their dogs need from puppyhood to adulthood. Her specialty is helping owners with challenging puppies overcome their problems.”
This lecture was about her specialty which is dealing with aggressive puppies. Since I do not work with dogs a lot, I tend to choose lectures that are more general, but I attended this one because we do see aggression in horses (mostly biting) and I thought there might be some interesting parallels. I also got into clicker training because of an aggressive horse so I have an interest in using positive reinforcement to change our relationships with difficult animals.
She started with some general information on aggression in puppies, based on her experience. The truth is that aggressive puppies are born every day and that nature and nurture are both in play. Some puppies do seem to have a genetic disposition that makes them become aggressive and this can either be improved through correct training or become worse through inattention or bad handling. There are also puppies that are not genetically inclined to be aggressive, but that become that way through life experiences. She makes the distinction between “snarky” puppies of which she sees a good number, and truly aggressive puppies which are much less common.
In order to talk about aggression, we have to define aggression. Most puppies exhibit some degree of mouthiness and play behaviors that can get a little overwhelming at times but she defines aggression as “a willingness to threaten and/or bite under mild duress.” What is mild duress? A person could put a puppy under mild duress by doing any of the following:
- Picking it up
- Moving it from place to place
- taking an object away from it (guarding)
- touching, approaching, seeing it
- looking at it with direct eye contact
These are all reasonable things to do with a puppy, but an aggressive puppy will react differently than a normal puppy. She showed some video of a normal puppy in a trainer’s lap and an aggressive puppy in the same situation (but using a little caution) and there was a pretty clear difference in the puppy’s body language.
Every puppy is an individual so there are variations in responses and she looks at these puppies as being at different places along an aggression scale. In her aggression scale an “angel puppy” is on the left and an “off the chart aggressive puppy” is on the right. Snarky puppies are somewhere in the middle. On the scale she also has a line for “guarding” as this is one of the indicators of puppy’s potential for aggression. Guarding can be minor but if it is not addressed, it can lead to true aggression.
In 2013 there were a lot of puppies enrolled in Puppy Kindergarten at her dog training school (300) and only a small number were truly aggressive. With these puppies she looks at the puppy’s current environment, previous history (rescue, puppy mill, breeder?) and the available resources to help the owners decide if they should try to work with the puppy, send it back to the breeder or rehome it. These are often difficult decisions and to learn more about aggressive puppies, she actually started looking for (through shelters and other trainers) aggressive puppies so she could develop an aggressive puppy protocol.
Here are the main components of “Sharkey’s Puppy Rehab:”
- Absolutely no place for aversives in dealing with aggression
- Clickers rock – they allow you to take the emotion out of an emotional situation
- Must have a crate
- no children under 12 in the home (she’ll still work with a them, but she really recommends rehoming an aggressive puppy if you have young children. It’s not worth the risk.)
Her program is built around desensitization and counter-conditioning and uses various handling exercises, recalls, exchanges, and play with toys. Some puppies are very oral and need to bite things so it’s important for them to have an acceptable way to satisfy that need.
She shared three case studies which were Joey, Mack and Grover.
Joey was an Amish farm puppy who showed intense aggression for no reason. She had some pretty scary video of Joey freezing and then biting for no obvious reason. He was dog-dog aggressive, dog-people aggressive and seemed to have very few puppy qualities. She adopted him from his family and spent a long time rehabbing him (6 months – 1 year?) with significant improvements in some areas but not in others. He was eventually PTS when it became clear he was not going to be able to cope with life without putting other people and dogs at risk.
Mack was surrendered by his owner to the shelter because he was “aggressive when disciplined.” He showed some guarding tendencies and could be aggressive to same age puppies and when held or restrained. He was successfully rehabilitated.
Grover was also successfully rehabilitated. He was a shelter puppy and showed resource guarding, especially when food was taken away. He would growl and threaten to bite or bite. She showed some video of how she taught him to trade one object for another and reinforced him for letting go of a desirable object. One comment she made was that with dogs that guard objects, it’s important that you are clear about what you are doing. You offer a different object and then let the puppy see you take the other one away. If you sneak it away, then the puppy is not learning the more important lesson which is that it’s ok if something is removed.
She finished with some final thoughts about working with aggressive puppies. These were:
- You need a deep toolbox. Don’t take on an aggressive puppy unless you are a skilled trainer with many options.
- You must be committed to +R only
- You must be willing to deliver the truth – if the puppy is not going to work out for that family, you need to tell them.
- You need to identify when a situation is unsafe
Aggressive puppies should not be in homes with children.
It’s no one’s fault –
I wasn’t sure what I would get out of this talk, but it was pretty interesting and produced some interesting physical responses. I could feel my body tense in some of the videos where she showed aggressive puppies. I really give credit to anyone who takes on one of these because it takes a lot of commitment. Early on in the talk Laura said that you really have to believe the puppy should be helped and could be helped and give it your all.
How does this relate back to horses? I have to wonder how many horses are born truly aggressive. It seems to me that because horses are not predators, there would be less genetic inclination toward aggression, but I am just guessing. Here are some things I pulled out of this talk that I would put into a plan for dealing with an aggressive horse:
- Absolutely no place for aversives in dealing with aggression
- Clickers rock – they allow you to take the emotion out of an emotional situation
- Must have a protective contact.
- You must have a deep toolbox.
- You must be committed to +R only or –R applied as information
- You need to identify when a situation is unsafe
- You must believe the animal can change
When I was dealing with my horse, the biggest turn around came when I could look at what she was doing as “just behavior.” I didn’t take it personally. I didn’t see it as a reflection on who she was. I didn’t feel I had to react because she “would get away with it.” Most of us are not dealing with truly aggressive animals, but I think some of the lessons from this talk are worth remembering regardless of what or who we are training.
Kay Laurence, “Freedom to Learn: training in the absence of formal competition.”
This talk was originally titled “Freedom to Train,” but Kay changed the title. That was actually very appropriate because the message from this talk was that we need to take a closer look at the words we use, the exercises we teach, and the functions of the exercises we teach. Are we blindly following some traditions, training for a specific sport, or are we teaching behaviors that are useful to us?
Let’s think about the words we use. Most of the words we use as cues come from traditional training and sound like commands because much of early dog training was done by the military. This is not to say all traditional training is bad (think of welsh shepherds), but the training was done to prepare dogs to do certain jobs and it may be that these are not the same things we need our dogs to do now.
Here are some of the words/phrases that might be worth a second look:
Loose leash walking – what behavior is this? The loose leash is an outcome, not a behavior. If you click for a loose leash, you end up clicking for 1000 different things. Kay prefers to use the phrase “walking together,” which means walking with connection and awareness of each other.
Get down – ok, this is probably not a cue many of us teach, but it is certainly used by many dog owners. Kay’s point was that this phrase and the tone that often goes with it are not how you would speak to a family member who just wants to say hello. Greeting is important to dogs. They want to say hello to us when we come back after being away. We should take the time to train an appropriate greeting ritual that meets the needs of both parties.
Sit – Kay is not a fan of training the sit as an all purpose behavior. People tend to use it as a control mechanism instead of looking more closely at what the dog needs in any given situation. For example, in greeting a dog wants proximity to your face and social approval. Making the dog sit is something that we do because it suits our needs, not the dog’s needs. The same thing is true in many other situations where we ask for a sit. Also, some dogs don’t find sitting comfortable. A down or a stand is often a better choice.
Leave – it! What does that mean? What behavior is the dog supposed to do? It’s better to train a “walk on by” behavior that tells the dog what to do.
An obedient dog – What is an obedient dog? Most people’s view of an obedient dog is what Kay would call a depressed dog. It just lies around and does nothing. The original reason for obedience training was to demonstrate control of dogs for traditional military purposes. Is that relevant to most of us?
Precision – What does it mean to train with precision? Is the precision training done for the love of training? Or is the precision training meant to show something else? Is there a difference between precision and being rigid or picky? When we “require” a lot of precision, as opposed to teaching with precision, does that affect our mindset?
These examples are intended to get you thinking a bit about how the words we use are important. Words carry more than just their literal meanings. They come with associations.
She talked about how we need to learn to think more like dogs. Her new slogan (maybe?) is “Be More Dog.” In order to train a dog well, you need to think like a dog. What does a dog want? Are there natural behaviors that address the dog’s needs and natural inclinations that can be used as part of our training so we are working with the dog’s nature, not against it?
She started by looking at the recall. She had some video of a formal obedience recall (very stylized) and a recall that was trained with the same basic behaviors but by someone with no history of doing formal obedience work. The two dogs did the same behaviors but there was a huge difference in how they looked (energy level, attitude). Is this difference because of how the person views the exercise, which then gets transferred to the dog?
Additionally it’s helpful to think about the purpose of a recall (should we even call it that?). A recall has two functions. It can be about moving the dog away from danger. Or it can be used to ask a dog to come toward you for interaction. If this is the point of a recall, then how important is a formal recall? Not very! The idea here is to think a little more closely about the function of the behaviors we teach our dogs.
Can we replace traditional ideas about what they should look like and how they should be trained by a more dog-centric version? This is Kay’s “Be More Dog.”
One might start by asking why a dog would NOT return when recalled. One common reason is that the dog has found some more reinforcing activity and chooses to continue doing that rather than return to the handler. This can be taught as a separate skill by spending time on toy exchanges. The dog learns that giving up one item leads to getting something better. She had some video of teaching a dog to catch and bring a Frisbee back to the handler by reinforcing that behavior with another Frisbee throw with a second Frisbee. If the trainer pasy attention to the timing and direction of the second Frisbee throw, the dog will learn to bring the first one back and drop it. She also showed how you can use tug to teach a dog that disengaging from the “dead” toy will lead to the opportunity to play with another toy.
Taking this to the next step, she makes use of the fact that dogs are cooperative hunters. One way to get a dog to want to come when you call is to teach it that you have found something of interest and need help. Or you can teach the dog that if it finds something of interest, it should come to you to get help. She had a video showing training a dog to come get the handler to help get an object it wanted. In this case, the object was a dead pheasant or toy in a cage. This dog was difficult to recall in the field because he found chasing pheasants more interesting than returning to his handler.
By using an actual pheasant and teaching him that his handler would help him get the pheasant, he learned that going and getting the handler was a good thing. This can be shaped into a recall which is very strong because the motivation for the dog to come back is based on the dog’s natural behavior. It’s interesting that if this is not done very carefully, the cooperative hunting aspect means that the handler is trained to come help the dog, not the opposite. And now that I think about it, that’s often what we do when we can’t get our dogs to come. We go find them, which means they really have us pretty well trained.
Like much of Kay’s work, this talk was pretty thought provoking. I do think that some types of animal training (certainly dogs and horses) are very heavily influenced by tradition and we end up following tradition because that’s what is being taught. But one thing to consider is that when you are learning something as it is traditionally taught, you are learning a mindset along with everything else. Even very subtle things (like the words used) can end up influencing how we feel about what we are training or the animals we are teaching. I spend time with a lot of traditional horse trainers and I have learned a lot from them. But I have also learned to be very careful about evaluating what I hear and see.
Kay Laurence, “Repurposing Default Behaviors” – lab
There was no lecture that accompanied this lab so this lab was a combination of Kay presenting information and people working with their dogs.
Kay started by talking about the word default and how it can be used. Then she shared her definition which is “the option which the chooser (animal) ends up with if she does not make an active choice.”
This is a little different than how I would have defined a default behavior which is a behavior that the animal can actively offer in the absence of other cues, or a behavior that is under some kind of environmental stimulus control. My horses have some behaviors they can offer without being actively cued (head down, standing quietly) and some behaviors that are based on environmental cues and reinforcement history. If I approach and open a stall door, my horses will back up. I don’t have to cue them with a verbal “back” because the door opening is the cue. Even though there is a cue, I still consider this a default behavior because if I stand in the doorway and don’t open the door, but just wait, a lot of them will still back up. It’s the most likely behavior to get reinforced in that context.
However, over the course of the lab I could see that Kay’s definition includes both those possibilities as well as another one, which is what the animal does if it doesn’t understand, recognize, is waiting for, or can’t comply with the cue. Teaching a default behavior for this scenario is an important component of training stimulus control. The way I had been using default behaviors was actually allowing some behaviors to have looser stimulus control. The way Kay is using a default behavior is actually intended to tighten up stimulus control. It means that if the dog doesn’t know what to do, it has a default behavior it does instead of throwing behavior at you. It also means that if the dog doesn’t know what to do, it has a clear way to say “I don’t get it.”
So how do you train a default behavior?
Kay explained how to train default behaviors that are taught so that the dog has a specific behavior it can do while waiting for a cue or if it isn’t sure what to do. A common scenario is when you cue the dog to do a behavior and the animal doesn’t do it. There could be several reasons for this.
It could be that it fails to recognize the cue, it can’t remember what behavior is associated with the cue, or it feels it cannot do the behavior in that situation (conflict of interest/reinforcers). In any of those cases, what do you want the animal to do? A lot of clicker trained animals will just start throwing behavior at you. For many people, this seems preferable to having the animal ignore you or do nothing, but it can create a whole string of problems with stimulus control and the animal can get almost frantic about trying to find the right answer. Kay’s approach means that when the dog doesn’t know what to do, it waits for more information. This is different than a dog that ignores you or your cue.
One of the reasons she teaches this is that she has found that learning the cues for behaviors is one of the hardest parts of training. In her training, she only trains behaviors that are part of a dog’s natural repertoire so the behaviors themselves are not something the dog has to learn. What the dog has to learn is when to do them and this puts more focus of training on getting behavior under good stimulus control. The more behaviors you have, the more important this becomes.
She talked a little bit more about the reasons for “non-responses.” If you give a cue and the dog does not respond with the correct behavior, consider the following possibilities:
1. The behavior you want puts the dog in conflict against an instinctive response. An example would be asking a dog to lie down on an uncomfortable surface.
2. The behavior might be an environmentally inappropriate behavior. That means there is something in the environment that says it’s not a good idea to do that behavior when you asked. In many cases, the dog will be saying “not right now,” and will comply if given a little time or an adjustment is made. She described how a student was training a small dog to sit and he was stepping toward it (unintentionally) while he cued it. The dog felt threatened by the man’s forward movement so it would scoot back instead of sitting. Once the man learned to keep his feet still, the dog would sit quite willingly.
3. The behavior costs the dog because it has to give up something that it already has. If the dog is already engaged in a reinforcing behavior, it might be reluctant to stop and do something else unless it has a long history of these trades being worthwhile.
When you set up a default behavior, you want to think about one that will be functional in the situation in which you want to use it. I know this sounds obvious, but actually it takes a little thought (GIMT- a Kay phrase which stands for Give it More Thought). With dogs, sit is often used as a default behavior, but there are some behaviors that are harder to do from a sit (the dog has to get up first) and if the dog is sitting all the time, then it becomes hard to practice sitting on cue.
She had some examples of scenarios and default behaviors from her own life:
- If she’s home and busy – go play by yourself
- home and visitor comes – go to the garden
- in class and listening – parking
- formal training – station in front
Don’t let the dog choose the default. You choose the default and teach it. The conditions (environmental or context cues) provide information to the dog about what to do. It wasn’t clear to me if that means she never cues default behaviors, or just that at some point the dog learns to expect to do a certain behavior in each situation so the active cueing (on her part) is minimal.
When we got to the working part of the lab, she focused on how to train the station in front and parking.
Teaching Station (stand) in front:
1. Throw food so dog goes out away from you. Pay attention to where you are throwing the food because you want to encourage the dog to re-orient to you after it finds it. She likes to throw the food into the corner of the room so the dog naturally turns. If you are not good at throwing food so the dog can track it, or have bad aim, then practice.
2. When the dog comes back, click as it approaches you, but while it is still moving. This is especially important with dogs that have been taught to return and sit (or do some behavior other than stand). If in doubt, do not click. You can always repeat the food throw to try again.
3. Make sure your body language is not asking for anything specific. A lot of dogs have been trained to react to the trainer’s posture when returning, so be casual and vary your stance/position/posture a bit so the dog does not attach the stand behavior to any particular body language cue.
4. It can be helpful at certain points to slow down the food delivery. This prolongs the “waiting in anticipation” which is the behavior you actually want.
5. Once the dog is reliably returning and stationing in front of you, you can start to cue behaviors. She doesn’t teach eye contact as a specific behavior. Instead she waits until the dog connects with her and shows it is ready. Then she cues a behavior. This builds a natural response of orienting toward the trainer when the dog is ready.
6. It’s important to mix in some nonsense words with the cues. You want to make sure the dog is really listening to the words and not guessing. Even the inhale before you speak can become a cue so you can practice changes in breathing. She noted that for many dogs “silence is a killer.” If you are quiet, they are going to start trying to do things. Working on nonsense cues is one way to teach the dog to wait longer between actual cues, but without the frustration that can happen if you try to build duration by just standing there and waiting.
7. Keep practicing sending the dog out (by throwing food) and returning to the station in front position even after you start working on cues. This will break things up a bit and also give you some information about if the dog is getting tired and needs a break.
8. You are going to reinforce both the default position (stand in front) as well as correct responses to cues. In the beginning, you may have to reinforce the default position several times between cues.
9. You can add a “wait” cue to the stand in front if you want. Default behaviors can have specific cues as well as environmental or context cues.
1. Parking is a useful behavior for greeting people. If you’ve never heard of “parking,” Kay has written/talked about it before. There is some information in my Clicker Expo notes from last year on Connected Walking and it’s also in her book “Every Dog, Every day.”
2. In parking, the handler holds the dog’s collar and steps on the leash so that the dog is limited in how much it can move. The leash does not pull the dog down, it just provides a boundary.
3. When you park your dog, you should relax too. You can rest one hip and relax your arms as cues to the dog that nothing is happening soon.
4. In general, she uses park as a “we are not training now” position so she does not click and reinforce specific behaviors, but she will provide some reinforcement when the dog is learning. She said this was a good time to wear a jacket with holes in the pockets. Just let a few treats fall out every now and then. Or crumble some and let the dog hunt on the floor. She even suggested grinding them into the carpet!
In the course of the lab, she shared some other useful tidbits:
- Don’t have clicker trainer’s elbow. Learn to stand with your arms relaxed.
- The social distance for dogs is farther away than we think. We often ask them to be too close to each other and that causes problems that could be avoided just by giving each dog a little more room.
- Waiting for a dog to be ready for a cue builds up trust. The dog learns that you will not ask for something until they are ready.
- Our training history can get us in trouble when we try to move forward. This was said in the discussion on dogs that sit all the time. Teaching a default sit seems like a good idea, but if the dog wants to sit all the time, it starts to interfere with training new behaviors.
- Cues are always associated with context (unless you take a lot of time to generalize and proof them). Dogs are always looking for relevant information that tells them when reinforcement is available. We can try to make some information more relevant, but that doesn’t mean the other information is not out there.
- With older dogs, she adds a distinct hand movement (a flick) when she clicks. The dog will learn to associate the wrist flick with the sound of the clicker and it becomes a visual click.
- Working on default behaviors will help you identify if your dog discriminates between cues well or not. If the dog keeps returning to the same behavior, then it is probably not discriminating well. If this happens, you can try changing something about the set-up (environment) so that the behavior that is being repeated is less likely to be triggered.
- If you train a behavior by shaping it and then adding the cue, that cue is the only piece of information the dog has about what behavior you want. This means that if the dog doesn’t respond to the cue, you have to go back and re-shape the behavior again. There is no “old cue” that you can use to jog the animal’s memory. One way to get around this is to associate any new behavior with an object during shaping. This object doesn’t have to be part of the shaping process; it just has to be present.
I think there’s lots of food for thought here. It made me think about how I use default behaviors and the connection between default behaviors and base position. Kay didn’t talk specifically about base position in this talk, but the idea of base position is another way of describing what you want the animal to do between repetitions or when waiting for a cue. I like that these are ways for the animal to communicate with you that it’s ready and waiting for a cue.
Kay Laurence, “The Craft of Fine Slicing.” – lab
This was a shaping lab, but we started off with an interesting conversation about dealing with dogs that are anxious in their environment. For many dogs, the set-up of the learning labs is quite different than how they are trained at home or at their usual training locations. It’s important to give the dog time to check out its environment and feel comfortable so it can pay attention to the training.
There was one dog that reacted when the sound system was turned on and Kay showed how she uses deep stroking to release tension. She finds that adding food is not helpful in these situations as it doesn’t address the cause of the tension and/or help the dog learn to relax. She called the stroking “sausage dog,” because she presses quite hard as she goes down the length of the dog’s body.
There was a little discussion on default behaviors, base position and resurgence. Kay uses all of these in her training to build and combine behaviors, teach the dog to wait, and improve stimulus control. You can read about these in my previous reports and one thing I always fine helpful is to look for connections between the material presented by different speakers. This helps me see the bigger picture of how many trainers use the same tools but with slight variations. Once I started to see the connections, similarities and differences, it was easier to come up with my own applications.
Kay had a nice image of shaping behavior. She thinks of starting at the base of tree. You build a good foundation (strong trunk), but then you can choose to go in many different directions. You have to decide which way to go and that will depend upon the a number of factors including the dog’s preference or needs, the long term use of the behavior, your own skills, and the resources available.
The focus of the lab was on shaping behavior in small increments so that the dog was meeting criteria 95% of the time. Behaviors should be shaped with care and precision. Here are Kay’s instructions for and comments on some of the behaviors that people chose to train:
STAND ON A MAT:
The base behavior is walking.
- Start by throwing food out, observing to see where dog stops when it returns. Most dogs will choose to stop the same distance away every time. You can repeat this step a few times to determine the dog’s stopping distance.
- Place the mat so that it is in the same location where the dog has been stopping.
- Throw food again and observe the dog’s behavior as it returns. If the dog looks at the mat, they are more likely to stand on it. If they don’t look at the mat, then they are less likely to stand on it.
- If the dog avoids the mat, then it means they are conscious it is there (which is good).
- If the dog continues to avoid the mat on its return, you can make the whole behavior smaller by placing food instead of throwing it. Instead of going many steps out and back, the movement can be reduced to one or two steps.
- Place the food so the dog has to step back to get it. The dog will come forward after eating and you can place a mat in the dog’s path. The goal is to have to dog step back to get the food and come forward and step on the mat. When training this, be aware that touch, tap and stand are all different behaviors. The more precise you can be about reinforcing only the one you want, the cleaner the behavior will be.
- She doesn’t use food to lure the dog on to the mat. You want the dog to step on the mat with confidence.
She teaches a pivot using a tray. Start with a large tray and then make it smaller. If you want the dog to be able to pivot in both directions, then train both ways at once.
She teaches dogs to move sideways using targeting. She teaches the hind end movement (rotating) and then adds a target in front to get the whole dog moving sideways.
This is the “under the chair method.” It is combined with teaching the dog to back to a mat. She teaches backing by having the dog back to a mat. This makes it easy to add distance and it fixes the problem of trying to time the click to reinforce backward movement, not stopping.
- Start by sitting in a chair with the dog in front of you.
- Place food under the chair so that the dog has to come forward under the chair to get it. The dog will get the food and then back up to look at you.
- Place a mat so that when the dog backs up to look at you, it puts its front feet on the mat. You can use back feet if you prefer, but it’s easier to see and you can time the click for when the second front foot steps on the mat.
(this is brief. I wrote detailed instructions on teaching this after the last expo so that’s in the 2013 ClickerExpo article on my website).
- With the dog lying down relaxed on one hip, place food so that the dog moves one front foot to get it. The food placement should encourage the dog to take one leg away from its body and then return the leg back to its previous position after the food is eaten.
- Once the dog is moving one leg out and back, you can place a target in the “return to center” location. You may have to experiment to find where to place the food. There are significant differences between breeds with some having very loose shoulders and others being tighter.
- Click the dog for touching the target. Over time you can move the target so the dog has to cross one leg over the other to touch the target.
You can teach your dog to pull a cloth to open a drawer or cupboard by putting the cloth around a toy and reinforcing the dog for holding and pulling the toy. Then take the “skin” off and put it on the new object and reinforce the same behavior.
Pulling and tugging are related. Teaching a good tug requires some thought. She recommends teaching toy exchanges so the dog learns both to pull and to let go within the same session. She often buys two of the same toy so the dog’s preference for one over the other is not a factor.
When you are tugging, make sure you and the dog are at an appropriate distance so he’s not in your body space and you’re not on top of him.
Make use of their natural learning environment which is play. We want to use that to teach them.
A few other comments:
Make sure your shaping plan includes the position of you and the dog and that this allows you to see any relevant details. If you are teaching the bow and want to click for elbows touching the floor, the dog must be positioned so you can see when that happens.
In some behaviors, you want to click the moment of stillness after the dog is in position. This will make the behavior more solid and avoid the problem of the dog fidgeting because it’s being clicked for slight variations on the same basic position.
We don’t “fix” dogs. We add more behaviors to their repertoires.
In behaviors that are trained by placing food, the placement of the food is part of the cue.
If you have any questions about these topics or about Clicker Expo in general, feel free to email me with questions or comments. I’ll try to answer them or point you toward more resources.
Katie Bartlett, 2014 – please do not copy or distribute without my permission