EQUINE CLICKER TRAINING.....
                                                               using precision and positive reinforcement to teach horses and people
 
         

 

  

The Art and Science of Animal Training Conference (ORCA)

 

March 14, 2015

 

 

Dr. Stanley Weiss – Differential Reinforcement and the stimulus control of behavior

 

Bob Bailey - Merging Behavior Analysis and Animal Training:  Improving your animal training, from aardvarks to zebras, using academic publications, self-help literature, and formal and informal classes.


Steve Aibel and Al Kordowski –The Least Reinforcing Scenario:   Its history, definition and usage in Animal Training.

 

Ken Ramirez – Teaching Conceptual Thinking:  It’s not asking too much of your dog!

 

Alexandra Kurland – Equine Simulators and Science as Metaphor

 

Steve White – Training Resilience

 

Phung Luu-  Birds in Flight

 

Dr. Stanley Weiss – Differential Reinforcement and the stimulus control of behavior

 

Dr. Weiss was the keynote speaker at the conference. His talk was rather technical … I have detailed notes from his talk and am going to share them, but it may be more information than some of you want.  He described a lot of experiments that he had done where he was looking at behavior under different stimulus conditions and trying to identify what was really controlling the behavior. If nothing else, listening to him talk made me realize how hard it is to design an experiment in the lab where there really is only one variable and if it’s hard there, then it’s even harder in real life. 


For those of you who don’t want to wade through my notes, I have put a summary at the end which includes my thoughts on practical applications for this information. You can just skip to it. It is titled “SUMMARY.”

 

DR. WEISS’S PRESENTATION:

 

His particular interest is in looking at the relationship between stimulus control, generalization and discrimination.  He has done a lot of experiments on various aspects of stimulus control, in particular on what happens when there are multiple variables or stimuli so that it is not clear what is actually controlling the behavior.  He has also looked at the interactions between multiple stimuli in cases where there is more than one S delta.   

 

Overview:


·         Discriminative stimulus – sets the occasion where an operant behavior can be reinforced.

·         Differential reinforcement is necessary to create stimulus control

·         However often comparable appearing behavior can be the product of different underlying stimulus control

The relationship between behavior, discriminating stimuli, S (delta) and reinforcers can be described as a three term contingency:

 

S (delta) -> Response -> S (reinforcer)

 

This basic phrase presents a simple way of looking at how behavior is controlled by cues (S delta) and reinforcers, but in real life it is not always so easy to see exactly what is controlling behavior. There can be multiple discriminative stimuli and reinforcers present and the animal’s response can also be affected by the level of generalization, previous associations or combinations of variables.  In his talk, Dr. Weiss shared how he has studied factors affecting stimulus control and what we can learn about behavior by looking at stimulus control and vice versa.

 

He talked about a few specific experiments involving:


·         Discrimination training

·         Motivation

·         Compound stimulus control – overshadowing

·         Compound stimulus control – selective associations

·         Compound stimulus control – blocking

·         The incentive-motive process and loss of self-control

STIMULUS CONTROL AND DISCRIMINATION TRAINING:

 

"Attention can be focused on specific aspects of complex stimuli through relevant discrimination training.”

He started with an example of discrimination training.  Teaching an animal to select an item based on color is an example of discrimination training in which the animal’s behavior is shaped through differential reinforcement.  He had a graph showing the progress of a pigeon being trained to select green when offered a choice of green or red. Over time, the number of pecks on green increased and the number of pecks on red decreased.

 

If you wanted to refine your color discrimination to ensure the pigeon was selecting based on color (wavelength) and not intensity or hue, you would need to vary the saturation and intensity to prove that those were not controlling the pigeon’s choice.

 

The red vs. green choice is what he called interdimensional training because there’s a clear choice between red and green.  If you graph responses, the line showing the response before training is flat (all choices are pecked equally) whereas the line showing responses after training peaks at the wavelength that corresponds to green.   In this graph, the magnitude of response is inversely related between the test stimuli and the training of S delta.

 

You can also do discrimination that he calls intradimensional training, which in this case would be teaching the pigeon to discriminate between shades of green.  This would mean varying intensity and hue to so that the pigeon had to learn to pick a specific shade of green. Again, this is done through differential reinforcement.  The graph for this shows a shift in the peak to correspond to the new response.  By adding another level to the discrimination (a shade of green), we can change how the pigeons see the world. They learn to differentiate between shades of green instead of just seeing all green as being the same.

 

What do we learn about stimulus control from Discrimination Experiments?

 

Differential reinforcement and Stimulus Generalization:


·     Generalization – the capacity of stimuli similar to the S (delta) to control behavior. Without discrimination training, there is complete generalization.

·     With interdimensional training (S delta on, S delta off), you get clear differences, magnitude of response is inversely related between test stimuli and training of S delta

·     What about intradimensional (shades of green?) The peak shifts away from S delta.

·    Discrimination training affects how an animal sees the world.  Once it has learned to discriminate, it can make different choices.

STIMULUS CONTROL AND MOTIVATION:


·         Increasing motivation can lead to more generalization (errors)

·         food deprivation increases errors

STIMULUS CONTROL AND COMPOUND STIMULUS CONDITIONING:


Overshadowing – one element of a compound reduces stimulus control over behavior acquired by another element of the compound.


·     Overshadowing happens when a behavior is reinforced or punished in the presence of multiple discriminative stimuli and they do not end up having equal control of the behavior. 

·      It can make it hard to determine which discriminative stimuli is controlling the behavior, or to get stimulus control with the desired discriminative stimuli if there is another one that is competing with it. 

·     For example:  You can train an animal to select a triangle on red vs. a circle on green.  You can train it to select based on the color or on the shape.  If you train for color, the color will be more relevant than the shape. If you train for shape, the shape will become more relevant than the color.   It will look the same, whether you train for color or shape, even though it’s under different stimulus control. 

·     You need to do a stimulus-element test: present each element separately to determine amount of control acquired by each element.

Factors affecting overshadowing:


·         Selective attention 

·         Salience of elements

·         Selective associations

 

Overshadowing can be a problem because it means you don’t really know what is controlling behavior, or you can find it difficult to use one type of discriminative stimulus if there is another one that overshadows it. 

 

STIMULUS CONTROL AND SELECTIVE ASSOCIATIONS

 

He described a series of experiments that showed how you can end up with similar behavior under different reinforcement contingencies and what we can learn about stimulus control by studying selective associations.

He started with a classic experiment where rats were reinforced for lever pressing under positive reinforcement and negative reinforcement conditions.

 

Initial experiment:

 

Food group – 

·         Lever press during Tone + Light = food

·         Lever press in absence of Tone and Light = no food

·         Rats are observed to lever press only in presence of Tone and Light

Shock group –

·         Lever press during Tone and Light = delay of shock

·         Lever press in absence of Tone and Light = no shock

·         Rats are observed to lever press only in presence of Tone and Light

So the question was… is the S delta for lever pressing the Tone, the Light, or both?

 

When he did a stimulus element test, he discovered that the rats working for food were being controlled by the light, and the rats working to avoid shock were being controlled by the tone.

 

This led to the Appetitive Aversive Interaction Theory of Motivation which says that there are Selective Associations:


·         Food  = positive

·         Shock = aversive

·         Absence of food = aversive

·         Absence of shock = positive

  • Therefore, in traditional experiments producing selective associations, type of reinforcer and conditioned hedonic state covary. (what does covary mean?  It means that they are related so that a change in one affects the other in either an equal or opposite way).

Can this “confounding” be eliminated?  Can you set up an experiment so that the same conditions are created (lever pressing during Tone and Light) but without using shock? Would this affect whether the rats were responding to the Tone vs. the Light?

 

Second experiment:

 

First group (food group) –


·         Lever press during Tone + Light = food

·         Lever press in absence of Tone and Light = no food

·         Rats are observed to lever press only in presence of Tone and Light

Second group (chain group) –


·      Lever press during Tone and Light -> absence of Tone and Light and delivery of food (the rats were lever pressing to turn OFF the tone and light)


·      Lever press during Tone and Light = no food (if they started lever pressing during Tone and Light, they did not get reinforced)


·      Rats are observed to lever press only in presence of Tone and Light


When he did a stimulus element test with the second group of rats, he found that they were responding to the tone, not the light, even though they were reinforced with food. This was the same as the results he got from the shock group in the first experiment. By changing the reinforcement contingencies, he was able to change which discriminative stimulus was more relevant.  

 

These results showed that in compound conditioning, there are some biological predispositions that affect what kinds of stimuli are stronger for different species and types of behaviors.  But these can be changed by altering stimulus conditions. He said “biological predispositions influence what is learned in compound conditioning.” We tend to start with some selective associations and these influence our learning unless we are specifically shaped to respond otherwise.

 

STIMULUS CONTROL AND BLOCKING

 

In compound stimulus conditioning, there is another thing that can happen. This is called blocking and it’s when there is prior conditioning history of one or more of the elements and this prevents the animal from making selective associations.

 

Kamin’s Blocking Experiment:

 

Initial preparation:  All animals are trained to bar press for food. (I think they were rats)

 

Experiment 1:  

·         Control group:  Noise + Light = shock

·         Experimental group (blocking):  Noise = shock, then Noise + Light = shock

After this had been done, both groups showed a suppression of bar pressing under the Noise + Light conditions. The Noise and the Light became associated with shock, so when they were presented the animals responded as if expecting a shock. 

 

The next step was to test and see if bar pressing was suppressed by Light alone.  He had a chart showing the results of this experiment.  In the control group, there was significant suppression (almost 100%) of the bar pressing behavior.

 

Remember that these animals were conditioned to both the Light and Noise at the same time.

 

In the experimental group, there was much less suppression (about 20%).  Because the animals in the experimental group first experienced the shock when paired with the noise alone, the light did not become as strongly associated with the coming shock.   The “pre-conditioning” of the noise in the experimental group blocked the animals from making an association between the light and the shock.

 

Blocking is interesting because it shows that you can set up an animal to pay more attention to one discriminative stimulus than another, just by exposing it to them at different times.  This is useful in cases where you want the animal to use one discriminative stimulus when it might be naturally inclined to use another. 

 

The new experiment was a variation on the Tone and Light experiment (described earlier) where the animals learned to bar press in the presence of the Tone and Light.   If you remember, the rats that were working to avoid shock responded more to the tone, and the rats that were reinforced with food responded to the light.  

 

They set it up to see if they could reverse which discriminative stimulus was blocked by doing some “pre-training.”  Prior to the experiment where they would occur together, they exposed the rats in the shock group to the Light alone first, and then Light and Tone. This set up conditions where the Light was more relevant.  They were able to do the same thing with the food group by exposing them to Tone first, and then Light and Tone.  At the end of the experiment, the rats who were shocked were using the Light as the S delta and the rats that were working for food were using the Tone. He could actually go back and forth between the two experiments and condition them to either.

 

PROPERTIES OF THE DISCRIMINATIVE STIMULUS

 

Going back to the relationship between the S delta, Response and Reinforcement, we can now look more closely at two different aspects of it.

 

The phrase S (delta) – Response – S (reinforcement)

 

Describes two processes:


·         The incentive-motive process is a product of the embedded S (delta) – S (reinforcer) contingency.

·         The discrimination process is generated because the operant is only effective in the presence of particular stimuli (differential reinforcement).

This creates a challenge: 


·         Relative rate of response and relative magnitude of reinforcement covary in most multiple schedules.

·         Therefore, the contribution of each related process to resulting stimulus control is indeterminate. 

·         Or… in other words, you don’t know if the strength of the response is due to the strength/weakness of the S (delta) or the strength/weakness of the S (reinforcer)

THE INCENTIVE-MOTIVE PROCESS AND LOSS OF SELF-CONTROL

 

Here’s an experiment that shows the effect of the discriminative stimulus on the intensity of the behavior.   This is similar to one of his earlier experiments, but slightly different in that rats were trained to lever press in the presence of Tone OR Light.  His groups looked like this:

 

Food group –

·         Lever press in Tone OR Light = food

·         Lever press in absence of Tone OR Light = no food

·         Rats lever press in presence of Tone or Light

“Chain” group –

·         Lever press in Tone OR Light leads to absence of Tone or Light = food

·         Lever press in absence of Tone or Light = no food

·         Rats lever press in presence of Tone or Light

The behavior looked the same with both groups, but when he tested the rats in the presence of BOTH Tone and Light, he got a three fold increase in lever pressing for the food group.   The chain group continued to lever press at the same rate as they did with either Tone or Light. The food group rats were reinforced under the conditions of Tone or Light. The chain group rats were reinforced in the absence of Tone or Light.

 

One difference between the groups was that even though both groups looked like they were doing the same behavior with the same discriminative stimuli, only the food group was exposed to the Tone and Light and reinforcement (food) at the same time.  The chain group got their reinforcement (food) in the absence of Tone and Light.

 

To understand this, it is helpful to look at the incentive-motive process (S delta – S reinforcer) vs. the discrimination process (S delta – Response). The incentive-motive process is classically conditioned because it is about the association between the S delta and the S reinforcer.  In the food group, the Tone or Light happened at the same time as the food so the S delta was associated with an increase in reinforcement which produces an “excitatory, energizing, incentive-motive state.” 

 

In the chain group, the Tone or Light were associated with a decrease in reinforcement because the rat had to work to turn them off before it got reinforcement.  So for that group, the S delta was associated with a decrease in reinforcement and produced an “inhibitory, suppressing incentive-motive state.”   When Tone and Light were combined, it magnified the response for the food group because that group was already in an energized state.  The chain group was not, so the combination of Tone and Light had no effect.

 

The 3:1 increase in behavior that he observed in this experiment was also found in a series of other experiments.   He commented that nature has rules and that any time you find a repeating pattern or symmetry, it’s worth looking at it more closely.  He did similar stimulus compounding with different reinforcers (water, shock avoidance, cocaine, heroin) and found an increase of between 2.5 and 3 for the Tone and Light conditions in all of them.

 

So if an animal responds more strongly when stimuli are combined, what does this mean? There are some interesting implications because it means that you can use a series of stimulus conditions to make an animal more likely to lose “self-control.”

 

They tested this using rats and cocaine.  Rats normally self-regulate their cocaine intake and won’t take more cocaine than is comfortable. So it’s easy to determine a baseline level for each rat.  When rats are conditioned to the Tone or Light and trained to bar press for cocaine, they still regulate their cocaine intake. But if they are conditioned to Tone or Light and then exposed to both, then they lose their “self-control” and will triple their normal cocaine intake.  They could condition the rats to have a loss of self control when placed in an environment where there were multiple stimuli for the same behavior. 

 

He could do the same thing with shock by “pre-conditioning” rats so that they accept up to three times the amount of shocks in the combined stimulus condition. He said “This self destructive ‘loss of control’ is produced by the stimulus control established to the environmental conditions confronting the subject.  It is entirely predictable by contingencies of reinforcement operating here.  Therefore, the subject has no ‘choice’!”

 

He calls this an “environmentally determined self-destructive “loss of control.”   

 

These experiments showed that multiple stimuli can act in combination to magnify the effects of either one alone if they are stimuli that reinforce the same response.  They also showed that when the stimuli are in opposition, they appear to counteract each other. This supports an algebraic combination of the two underlying processes – the discriminative-response and incentive-motivational

 

This is a lot to think about!  I have to confess that my brain reached saturation point somewhere before the end of his talk so the last part of my notes was written but not really understood at the time. I had to go back later, re-read them and do some thinking. I have tried to convey the information accurately but some of it did get very technical and he had a lot of graphs.  I apologize if there are any errors. If you want to read more about Dr. Weiss’s work, you can email me and I will share the papers that I have on his work with stimulus control.  

 

SUMMARY:

 

Discrimination Training involves both an incentive-motive process (S delta – S reinforcer) and a discrimination process (S delta – Response).  S delta is the discriminative stimulus or cue.  Discrimination is taught through differential reinforcement and leads to stimulus control.  When a response is first learned, it is generalized, meaning it has not yet been refined – either in specificity of the response or association with an S delta.  Variability in behavior means we don’t understand all the controlling stimuli. 

 

Discrimination training is about teaching an animal to respond to more specific cues and criteria.  In his example with the green vs. red color training, he showed that pigeons can learn to “see” the differences between shades of green.  So discrimination training is about teaching cues but also about changing what an animal sees, hears, finds relevant, etc…

 

We often refer to discrimination training as putting behavior on cue, and if you’ve ever tried to do that, you know that it’s not always as straight forward as it seems.  In addition to teaching the animal to respond to the discriminative stimulus that you have chosen, you also have to make sure that the animal is not using other discriminative stimuli.  Some factors that can affect the discrimination process are:


·     Overshadowing:  Overshadowing happens when a behavior is reinforced or punished in the presence of multiple discriminative stimuli and they do not end up having equal control of the behavior.  The animal will often choose one stimulus as being more “relevant” and ignore the rest.   If you have multiple stimuli, it can be hard to tell which one is really controlling the behavior.

·     Selective Associations:  Some species are predisposed to pay more attention to certain kinds of stimuli, or to certain kinds of stimuli for certain behaviors. If you chose a stimulus that your animal is biologically inclined to respond to, it will be easier to get the behavior under stimulus control.  If you want to use a stimulus that is different or that might be overshadowed or blocked by the more biologically relevant one, then you can pre-condition the animal to respond to it.

·     Blocking:  If you present multiple stimuli at the same time, they are not all perceived equally by the animal.  This can make it difficult to condition a discriminative stimulus if there is an existing stimulus that competes with it. The existing stimulus will “block” the effect of the new stimulus.

·     You can train two identical looking behaviors under different stimulus conditions. They may look the same, but they will respond differently to changes in the environment or other manipulations. 

Motivational processes clearly contribute to stimulus control. 


·     The presence of multiple stimuli for the same behavior can lead to a three fold increase in the animal’s response due to the incentive-motive process part of discrimination training.   This is a result of the classical conditioning that happens when the S delta is associated with the S reinforcer or punisher.  This means that animals can lose “self-control” when put in environmental conditions where multiple stimuli are present. 

·     This can result in an increase in behavior (increased cocaine or food consumption) or a decrease in self-control (self-preservation) as shown by rats that tolerated up to three times as many shocks. 

 

Finally, as behavioral scientists we need to be sensitive to co-varying processes that might be operating and find ways to eliminate this experimental confounding.  We have seen how profoundly informative that can turn out to be.

 

What does this mean for us as trainers?

 

I think it gives us a lot of information about how to choose and add cues, as well as information about what to do when stimulus control breaks down.  One of the challenges of getting reliable stimulus control is choosing salient cues that the animal can respond to under many conditions.  If we know about overshadowing, blocking, and selective associations, we can choose our cues wisely and recognize when there are other competing cues.  

 

The information here also tells us why it’s ineffective to give to two cues at once (a hand signal and a verbal). The animal is probably only going to use one of them and if we don’t know which one he is using, we could find ourselves in a situation where our cues are not effective.

 

I also thought the part about setting animals up to lose “self control” was interesting. It reminds me a bit of trigger stacking or putting an animal over threshold, but I think it probably also applies to those “over the top” responses we sometimes get from our horses when they are really excited about doing something. I’m going to have to think more on this one, well all of them really….

 

 

Bob Bailey - Merging Behavior Analysis and Animal Training:  Improving your animal training, from aardvarks to zebras, using academic publications, self-help literature, and formal and informal classes.


In his presentation, Bob talked about the different resources that are available to trainers who want to improve their training or teaching, and how to find the ones that are going to help you achieve your goals.   With the explosion of available information, the problem has changed from lack of information to an overabundance of information and it’s important to think carefully about who you train with and why.  

 

He had a timeline showing how educational opportunities for trainers have changed over the years. 

 

1947 Brelands started their training classes with chickens

1950s – field dog training took off

1960s – 4H

1960s- animal training for movies, etc..

1980s – beginning OC workshops (Karen Pryor)

1990s – animal models (public chicken workshops – Bob Bailey)

2000s – more schools, demos, seminars etc..

2010s – more resources are available through the internet including courses, webinars etc..

 

DO YOU WANT TO LEARN MORE?

 

Bob talked about the difference between wanting to learn more vs. wishing you could learn more.  He said that you have to WANT to do it and be willing to translate that into action.  Your success depends upon your willingness to expand resources to achieve that which is wanted.  So ask yourself:


·         Do you WANT it or do you WISH it?

Translating WANT into action:

 

Simple suggestions:


·         Make informed decisions about where you get your information, who you choose to work with, etc..

·         There is no shortage of information.  But finding accurate, valid, useful information can be time consuming.

·         Just because we disagree with information doesn’t mean it’s wrong

·         Do you go to a seminar because you agree with it or to get a new point of view?

How to learn more:


·         Direct experience – do it

·         Observation – watch others do it

·         Description – read or listen

What is best for you?

 

How did you get where you are now?  Are you willing to change your behavior?

 

In his opinion animal trainers have only slightly improved in mechanical and observational skills since he started working with people.  He sees that a lot of trainers are becoming “intellectualized.”  Thinking about doing is not the same as doing.  Imagery may help an already skilled trainer, but mechanical and observational skills are improved by “doing” – and being critiqued or observing video.  These performance skills take time and effort.  Animal training is not a spectator sport.

 

He went into a little more detail on the advantages and disadvantages of direct experience, learning by observation, and learning through reading or listening.

 

DIRECT EXPERIENCE

 

Most people need more direct experience, which has the following advantages:


·         It combines the basics of learning, playing, and doing

·         It has the maximum requirement  for hand-eye coordination

·         There is the least rehearsal of unwanted behavior if you have proper supervision

OBSERVATION (seminars, DVDs, demonstrations, etc..)

 

If you choose to attend a seminar, you should make a point of learning some basic information about who is teaching it (background, philosophy, experience, etc…) and also think about why you are there.  You might want to ask yourself the following questions:


·         Is the topic relevant to my current or future needs?

·         What do I know about the presenter? What have I done to know the expert?

·         Why am I attending it? (going to validate what you already do is not a good reason)

·         How will I use the information?

·         Am I prepared to use the information?

·         How do I know if the expert knows what she is doing?

·         Do I know how long the expert has been doing it?

·         Will there be a demonstration?

·         Is it appropriately priced? Does it fit my budget?

You can get more out of a seminar if you:


·         Prepare 3 questions ahead of time – go after the information you want

·         Meet the speaker – you have paid for the chance to meet face to face, so take it.

·         Recognize that good trainers evolve.  It is harder for teachers to learn because we have to go back and say “oops” to our students, but good teachers learn to do that in order to become better teachers and trainers.

·         If you are teaching, you should have a philosophy – your philosophy will be your guide. 

When attending a seminar and evaluating a trainer, it’s important to notice how the trainer handles mistakes.   All trainers make mistakes and they should acknowledge them, adjust what they are doing, and move on. This is important because it gives the audience permission to make mistakes and shows the nature of training which is never about just following a pre-determined recipe.

 

DESCRIPTION  (read and listen):


·         You can ask yourself the same questions that are listed under learning by observation, but be aware that what people say they do is not always the same as what they actually do, and that video can be faked.

Most people learn using a combination of different resources, so it’s a matter of balancing time spent working directly with the animal vs. time spent observing, reading and listening. 

 

SOME THOUGHTS ON TRAINERS:

 

He had some tips for evaluating trainers. 

A good trainer should:


·         Observe

·         Apply a few rules

·         Make it simple

·         Make it worthwhile for anyone

It’s the rare trainer that can get almost any behavior quickly, so you should not necessarily expect that level of proficiency.  But a good trainer should be able to get behavior in a reasonable amount of time. He does not accept a correct behavior as good enough – he trains for fluency.   If you keep looking for excellence, you will get it. Don’t accept Good Enough (GE)!

 

All trainers:


·         Are subject to ego – having their ego involved

·         Are subject to bias – have some bias

·         Make mistakes

·         Can improve

·         Can learn new things

FINAL ADVICE:

 

Find mentors, peers, colleagues.
Make it fun for yourself

Steve Aibel and Al Kordowski –The Least Reinforcing Scenario:   Its history, definition and usage in Animal Training.

 

Steve Aibel (Sea World San Antonio) and Al Kordowski (previously at Sea World, now at Service Dogs, Inc.) talked about the Least Reinforcing Scenario.  This is a technique for dealing with errors that was first implemented at Sea World.  Steve and Al wanted to do this talk because they felt that the LRS has been modified and misunderstood as more people are using it, and that the original intention and use have been lost.  They wanted to present clear information about its history, definition and use so that any misconceptions could be cleared up.

 

The Least Reinforcing Scenario is a technique that was developed at Sea World in the 1990s.  It was originally called the Least Reinforcing Stimulus but as it became used more widely, the word “stimulus” was changed to “scenario”.

 

Steve and Al believe that understanding and using an LRS correctly is important because it provides a specific plan for what to do if your animal makes a mistake.  Dealing with mistakes in a consistent and non-emotional manner can improve your relationship with your animal because you both know what to do when something goes wrong.

History of the LRS:

 

When Al started at Sea World in the 1970s, they were still learning how to use positive reinforcement training with the animals and working out how to handle errors.  They tried different things including a delta rod (to indicate an error), ignoring it, saying “no” and giving a very small reward (1 fish).   All of these had some limitations in effectively communicating that an error had occurred and doing so without generating frustration or anger from the animal (or trainer).   So they were looking for another way to teach the animal to be calm and relaxed when it made an error.

 

In 1991, they started using a specific reduction technique, based on DRO, which they named the “Least Reinforcing Stimulus.”   In the original paper (which is in Ken Ramirez’s book on page 103) the authors say “The LRS is a 2-3 second pause, with eye contact by the trainer. It is not a fixed or pre-determined posture that the trainer assumes after an undesired response.”   It functions as a break in the flow of training.  The article also says that “The LRS provides an opportunity for the animal to learn to terminate the response it is given, think about the criteria and to try another behavior the next time.”

 

In 1999, the LRS was renamed the “Least Reinforcing Scenario” because the word “stimulus” seemed too narrow for some situations.  It also suggested that the LRS for all situations was the same, but this is not always the case.   At some point, the use of the LRS started to become more widespread as trainers working with other species and in other environments heard about it.   But along the way, the purpose and use seem to have gotten a bit blurry, so Steve and Al wanted to clear up any misconceptions.  They shared some clear information about how the LRS was used at Sea World and how you can use a LRS in your training, while staying true to the original intention of the procedure. 

 

The purpose of the LRS:


·         Proactively condition calm behavior after an incorrect response

·         Counter-condition an appropriate response to non-reinforcement

·         Teach animals to look for new direction after non-reinforcement

·         Teach a “non-punishment” attitude in people

LRS is not:


·         A time-out (you can reinforce the correct response to the LRS)

·         Punishment (the LRS is intended to teach the animal to stay calm and relaxed after an error)

·         An aversive event

·         A predetermined posture (these can become aversive if paired with lack of reinforcement so fixed postures are avoided)

LRS is:


·         A DRO procedure

·         An alternative to punishment

·         An increasing process

·         A reinforcement opportunity

If you take these three lists together, what you get is that the LRS is intended to teach the animal and trainer to calmly respond to an error in such a way that the animal returns back to the trainer or waits for more information while being relaxed and attentive.   In order to understand how this works, it helps to look at how the LRS fits into the ABC training cycle and also recognize that there are two parts to the LRS.   The LRS includes both the 2-3 second pause and what happens immediately after.

 

The LRS as part of the ABC Cycle:

 

They had a nice diagram showing how the LRS fits into the ABC loop.

 

The normal ABC cycle for a correct behavior is:

 

 A (cue) -> B (behavior) -> C  (bridge and reinforcement)

 

If an error is made, the cycle will look like this instead:

 

A (cue) ->   error (wrong behavior) -> LRS (which they drew as a box) ->

 

Why did they draw the LRS as a box?    They used a box in the diagram because they wanted to emphasize that there are different ways of implementing an LRS.  This is why it is called the Least Reinforcing Scenario.  You can have different scenarios depending upon what you think is the most appropriate way to respond to the error.  They discussed two parts, when to start it and what to do after the pause (strategies).

 

When to start the LRS (the 2-3 second pause):


·         if the animal is working at a distance, the LRS starts when he returns to you

·         if the animal is working up close, the LRS would start when the animal is attentive and gives you eye contact

·         There’s no point in starting the LRS if the animal is not paying attention to you

·         In many cases there will be a default behavior that is part of the requirement to start the LRS so the LRS is used to reinforce returning calmly to a station or default behavior after an error

 

LRS strategies for after the pause, you can:


·         ask for a high probability behavior and reinforce a correct response

·         reinforce the animal for being calm and attentive

·         ask for the same behavior again, bridge and reinforce if correct

·         terminate the session

Be mindful of using a LRS too many times in a session. You also want to monitor your animal’s response to the LRS.

 

The LRS is meant to reinforce the animal for being calm and relaxed after an error so if you are using the LRS and the animal is getting more agitated or choosing to leave the training session, then the LRS is not working as intended.

 

They had some video showing using an LRS with dolphins and with a dog.  In both cases, the animals were able to continue on with the training session and do the correct behavior. 

 

Teaching the LRS: 

 

Remember that this is a pro-active technique. If you want the animal to return to a quiet and calm position after an error, you need to teach that behavior first.


·         teach a calm and relaxed behavior  (stationing, waiting with eye contact, etc…)

·         cue that behavior after an error and reinforce it

·         reinforce the animal for offering that behavior after an error without being prompted

·         decrease the amount of reinforcement for that behavior and slowly shift from reinforcing it directly to reinforcing it by asking for another behavior or a high probability behavior

·         monitor the animal’s response to the LRS to make sure that it is not aversive

·         monitor your own behavior/evaluate your own skills because animal “errors” can often be traced back to trainer “errors.”  The LRS does not “fix” problems. If an animal is making errors on a specific behavior, you need to carefully evaluate your own training and make some changes so the animal can be successful.

They mentioned a few times that the LRS is as much for the trainer as it is for the animal being trained. Trainer frustration over errors is very quickly picked up by animals and can damage the working relationship and the quality of behaviors. 

 

The Proficient LRS user:


·         consistent approach

·         importance of reinforcing correct response

·         understands failure is ok

·         maintain strong criteria for LRS

·         counter-conditions appropriate response so it doesn’t become associated with failure

The LRS is a pro-active technique designed to help animals learn by reducing learning frustration.  Used correctly it teaches the animal what to do when an error happens and helps both the trainer and animal remain calm.


Ken Ramirez – Teaching Conceptual Thinking:  It’s not asking too much of your dog!


Ken Ramirez (www. Kenramireztraining. com) is a past Executive Vice-President of Animal Care and Training at Chicago’s Shedd Aquarium and is now the Executive Vice President and Chief Training Officer of Karen Pryor Clicker Training (KPCT).   He has been a popular presenter at Clicker Expo for many years and often talks about concept training.  In this talk, he shared some information about how to get started with concept training, as well as some of his most recent work teaching a dog to count.

 

What is concept training?


·         It takes training a step beyond the standard operant paradigm

·         Asks if an animal can think beyond simple cues to generalize larger concepts

·         Early work on concept training was done as part of cognitive studies on primates and cetaceans.

Ken was first exposed to concept training when he did guide dog work before he worked at the Shedd.  In guide dog work, the dogs start off with specific cues, but then learn to generalize in ways that are similar to concept training.    He was fascinated by the idea of intelligent disobedience, as well as how to train behaviors such as recognizing and evaluating object height to see whether or not the person would be safe.  There were also other aspects of their training where they had to learn to think beyond what they had been taught.

 

What are some of the similarities and differences between “standard” training and “concept” training?


·         Standard training teaches very specific behavior

·         Concept training teaches an idea

·         Creative games make concept training easier – allow you to transition from standard training to concept training. This includes free shaping games and more structures ones like Kay Laurence teaches in her book “Learning Games.”

·         Clear criteria is still important

·         Concept training builds on standard training and takes it to another level

How to Begin:


Before you start with concept training, you should have some important preliminary training done. This includes:


·         Solid basics for you and your animal  – (understanding and use of a marker signal, familiar with setting and sticking to clear criteria, fluent with cueing, animal is focused and eager to work)


·         Desensitizing them to new things constantly


·         Practicing generalization of behaviors and also testing your cues (will your dog sit if you are lying down when you say it?, if you are behind a door? etc…).  This is about encouraging the animal to think and figure out what you are asking for even when the cue is presented under different conditions.

Additional Key principles:

 

When you are ready to put together a plan to do some concept training, these are some important things to remember:


·         Practice without the animal (become fluent by yourself, then add in the animal)


·         Do some exploratory training - Is this training without a plan? Well, sort of.  You should have some idea what you want to do, but it can be somewhat less formal because the point of exploratory training is about doing some preliminary work so you know what to put in your plan.  It’s a chance to work out logistics, see how the animal responds to the set-up and gather information so you can make a good plan.


·         Train the rule of the concept before testing it. Concept training always starts by teaching a specific rule and then seeing if the animal can apply that rule to new situations.  You want to start with a simple behavior, train it thoroughly, test for it in non-conceptual ways (test for the rule), and then test it conceptually.   (Ken explained this more later)

What kinds of concepts can you teach?


·         Right/left

·         Up/down

·         Over/under

·         Shape/pattern/color recognition

·         (he had a longer list with more options, there are lots of choices)

Some people choose concepts that have a practical application, whereas others are just using concept training as a way to improve their own training skills or to teach their animal something new and different.    Start by determining what you need (or want) and work from there.

 

Teaching Modifier Cues:  Left and Right

 

He had some video of starting to teach left and right targets with a dog.  The initial exercise was just teaching the dog that the words left and right referred to directions.  He was sitting with the dog facing him and had two targets, one on the left and one on the right and would cue the dog to touch one of them.   The dog already knew hand signals so he started with them, then added the words “left” and “right” before giving the hand signal, and at the end he faded the hand signals out as the dog started to respond to the verbal cues.   

 

At this point the dog doesn’t know left and right as concepts, they are just another set of cues.  If Ken wants the dog to understand left and right as concepts, he will do the same exercise with several different objects or behaviors.  Then he will test to see if the dog understands left and right under new “test” conditions.  He said that it usually takes 3-4 behaviors before the dog starts to understand that left and right are not cues for specific behaviors, but provide information about how to do the cued behavior.

 

Teaching Mimicry:  training a “copy” cue

 

Modifier cues are one type of concept training. Another type of concept training that Ken has been working on is teaching dogs to mimic (copy) each other on cue.  He started this a few years ago with a group of dogs at the Shedd and showed that a dog could learn to mimic another dog’s behavior when taught a “copy” cue. 

 

This is not the first time that mimicry has been explored by animal trainers. The Navy experimented with using mimicry to teach dolphins, in the hopes that it would speed up training time. The dolphins did learn the new behavior by copying other dolphins, but some of the precision was lost, so the project was abandoned.  Claudia Fugazza has also done some work using her “Do As I Do” protocol in which dogs learn to copy what a person does.  I think the verdict is still out

on whether or not mimicry is useful, but it can be fun to play with.

 

Teaching Counting:  how high can a dog count?

 

Ken’s newest project is teaching a dog to count.   Previously he has done counting with other species, but not in any formal way.  It was just a fun project.   Before starting this, he did some research and found out that there has been quite a bit of work showing that dogs can count to 3, and a few studies looked at counting to 5.  But there has been no real documentation and the definition of counting was pretty vague.   

 

Here are some steps that show how he is doing it:


·         Choose an indicator behavior(s):  How does the dog tell you the number of objects?   He started out with very different behaviors (sit, down, kennel) but changed to having the dog choose a different target for each number.   You can impress your friends by writing the number on the target, but the dog will learn to choose based on the target shape.

One point he made about indicator behaviors is that you have to be careful to avoid behaviors that are susceptible to the Clever Hans effect.  As the numbers got higher, Ken switched from shapes to white boards with the same number of dots as the objects. The dog would choose the appropriate whiteboard.

·         Decide how you are going to present the objects to be counted:  Ken chose dark colored objects and put them in a white tray. 

·         Choose a cue that tells the dog when to indicate the number of objects, and choose any other behaviors the dog needs to do as part of the exercise. The cue is what tells the dog when to do the indicator behavior. I think Ken asked “how many?” The indicator behavior is not the only behavior you need to train this. You also need  a way for the dog to tell you when it’s ready to go again.  Ken likes to have the dog return to a position in front of him, but any kind of default behavior will do.

·         Remember that the first part in concept training is teaching the rule.   In Ken’s set-up he had a line of whiteboards, each with different numbers of dots.  He would put ask the dog “how many?” and she would go touch the target next to the correct whiteboard.    But the first time a new number is presented, Ken shows the dog which whiteboard matches the number (he would point).   Then once she gets it right a few times, it is just added to the random list of numbers that he can put out.

·         If you want to be scientific about it, you will want to make sure that the dog is not cueing off any of your body movement, or some other variable in the environment.  Choosing good indicator behaviors can make the Clever Hans effect less likely, but it’s still a good idea to think about possible ways in which the dog could be getting information about the right answer.   Ken started out with a very simple set-up and as the numbers got higher, he added in some modifications (screens, extra people)  that made it less likely that the dog was cueing off him.

Ken is going to continue with this work and see how high Coral (the dog) can count.   He showed video of her working on numbers up to 10 and he said she is doing very well.  

 

It was fun to see the videos of Ken working with Coral. They had a nice routine and it was clear that Coral enjoyed the game.   

 

He ended with the final thought that “we limit ourselves and our animals by assuming things aren’t possible.”  So whether you want to do concept training for a practical reason or if you just want to increase your own training skills, it’s worth doing.

 

Alexandra Kurland – Equine Simulators and Science as Metaphor

 

Alexandra Kurland (www. Theclickercenter.  Com) was one of the earliest people to write about using clicker training with horses. She is the author of several books and a DVD series on the subject and is always looking for new ways to understand how horses and people learn and what leads to success.   Her presentation was a look at how excellence is a product of both mental and physical skills, which we all have the ability to learn. 

 

She started off with a comparison of two horses, one with ideal conformation (and presumably ability) to do its job and one with some physical challenges.  Then she showed a video of how the horse with less natural ability could still be trained to move beautifully.   

 

And she asked the question “What made this change possible?”  

 

It turns out that other people have been asking the same question and they have come up with some great answers about why some people achieve excellence whereas others do not.   Their work resonated with her because she has seen some of the same things.  She talked about the work of Daniel Coyle and Carol Dweck.


·    In “The Talent Code” by Daniel Coyle, he writes about travelling to places that consistently produce above average athletes to see what  make these places “hot spots” for talent. What he found was that “talent” was not something that people brought into their training programs, it was something that was carefully cultivated and nurtured through a systematic method of building skills and teaching people how to strive for excellence by using deliberate practice (more on this later).

·    Reading The Talent Code made it clear that the old idea of talent as something you are born with could be replaced by a new idea, which is that talent is a product of intensive and very deliberate practice.  

·    Daniel Coyle was looking at the development of physical skills, but excellence also comes from how we think about what we are doing (mental skills). 

 

·    Carol Dweck is a professor of psychology at Stanford University who studies achievement and success.   One of the things she talks about is the importance of “mindset.”  Mindset is how you view your own ability and your ability to change.  Carol studies how this affects your ability to learn and succeed.   

Carol talks about two kinds of mindsets, a fixed mindset (a stone) vs. a growth mindset (a seedling).  If you have a fixed mindset, then you believe that your own ability is fixed and you cannot change.  If you have a growth mindset, then you believe that you can change. This has some interesting implications for learning and teaching, because we tend to limit ourselves by our own perceptions of what we can do.  And this is a problem, not just for people who believe they are not talented, but also for those who believe they are talented.   Labeling either way creates a fixed mindset.

·    Alex shared a study Carol did with school aged children who were given a series of tests and received two different types of praise as feedback.  One group was told they had done well on the test because they were smart. The other group was told they had done well on the test because they worked hard.  The group that was told they were smart went on to choose easier tests (so they could continue to be “smart”) and were more likely to give up if the work got hard.  The group that was told that they “worked hard” chose to take on more challenging tests and showed more persistence and improvement over time.  Your own perception of what made you do well was a key factor in future success.

·         There have also been studies showing that the teacher’s mindset affects student performance.  If teachers are told they are teaching students with above average IQs, they find they have smarter students who do well.  If they are told they are teaching low IQ students, then the students don’t seem to do as well.  In one study, children were randomly assigned to the high and low IQ groups and the teacher’s mindset was the determining factor in their success.  If the teacher thought you had a high IQ you did well, if he thought you had a low IQ you did not. 

Alex likes to use metaphors in teaching.  Metaphors often come out of stories and people learn through stories which allow you to present information in a more informal way, one that often connects to people’s belief systems and is easier to relate to real life.  She likes to find metaphors that open doors, that lead people to think about possibilities.  

 

Alex said that if you apply Carol Dweck’s idea of mindset to horse training, you can see how it can undermine a person’s ability to work well with an individual horse.  If you have a fixed mindset, then you tend to be harder on yourself and on your horse, especially if you have a horse that seems to have “ability.” Your expectation of what that horse should be able to do will influence your decisions about how to work with it.  She said that these people often bounce from trainer to trainer looking for the right one because they think that if the horse is talented, then it should be easy. 

 

Is there a physical component to excellence? 

 

Rather than thinking of talent as an innate ability, can we learn something by looking at physical changes that happen when people learn to excel?  Yes, but it helps if we understand how our nervous system responds to learning and for that we have to look at some neuroscience.

 

Understanding Myelin:

 

Alex gave a short description of the process of myelination which is how nerve fibers  change from being little used and inefficient pathways to being superhighways.    Nerve fibers are insulated with myelin to prevent leakage and every time you activate a neural circuit, more myelin is added to insulate it.  As the layers of myelin build up, the pathways become more and more efficient.  Becoming more skilled at doing something is the outward sign of increased myelin around  those nerve fibers. 

 

We can take away a few important facts from this:


·         Myelin is universal – we all have it and we can all add myelin to nerve fibers (it is not unique to “talented” people.

·         Myelin wraps, it does not unwrap

·         You can insulate good training skills and habits with myelin

·         The more times you activate a neural circuit (repeat a movement,  the more myelin will be added

This takes us to “Deep Practice.”  Deep practice is what happens when you set up your learning to promote myelination of nerve circuits.  How do you do this?

 

There are 3 tiers:


1.      Look at the task as a whole

2.      Break it down

3.      Play with space and time

Let’s look at them in more detail:


1.   Look at the task as a whole. This is about knowing where you are going, your goal if you wish.  Alex often tells people to find a look that pleases their eye.  If you have this image in your mind, you can train for it. This is partly due to mirror neurons which are activated when we watch something.   If you watch good riding, your body will want to copy it. Alex described the experience of being able to feel what someone else was doing.   At the same time, be careful what you watch or you might end up imitating work that you don’t like.

2.  Chunk the task down into small units. This is about finding the key, core component behaviors and perfecting them.  In her teaching, she has students thin slice behavior down until they can get a clean loop of behavior and then build from there.  If you allow little bobbles, little bits of “almost-good-enough-but-not-quite” to be practiced, then those errors will become insulated along with everything else.  It is better to begin with a small clean piece of behavior and develop it over time, than it is to start with a more general behavior and try to clean it up later.

3.  Play with Time and Space. 

Playing with time:  Slowing the action down can be helpful.  It exposes any bobbles in your technique which might not be visible at the MACRO level, but which are limiting your performance.  At the MICRO level she wants people to focus on all the little errors so that they can smooth them out.  You want to slow everything down so you can observe how you move.
Horses notice everything so detail matters.

While this all sounds good, the complication in working with horses is that our horses are learning from what we do while we are practicing.   I may be focusing on learning better skills with an eye to improvement, but my horse is learning from what I am actually doing. This is one reason that Alex often has people practice without the horse first.  By using equine simulators (usually other people), people can do many repetitions and build mechanical (technical) skills before they ask for the same thing from a horse. 

Playing with space:  There are different ways to play with space. One method is to set up training exercises/games where you learn to prepare for the unexpected.    In Brazil, soccer players play a game called futsal which is similar to soccer, but is played on a very small court.  Because there is less room, everything happens much faster and players spend more time in contact with the ball.  This accelerates the learning curve for handling the ball because they have more opportunities to interact with the ball, and the small court changes the dynamics of how the ball moves.   The ball is less predictable and players get more opportunities to interact with the ball in different ways.

These ideas can be applied to working with horses and Alex said they show the value of setting up exercises that are structured to allow many repetitions of the same movement, and also to play with time and space.  These are the kinds of exercises that Alex does in clinics, where people work in groups and study their own mechanical skills in great detail.  They can ask lots of different questions to deepen their understanding of what makes something work better.

 

She finished with a few thoughts on one of the most common questions about riding and clicker training, now explained through the idea of deliberate practice.  One of the questions that people often ask is how clicker trained horses make any progress with all the stopping that happens.  Alex said that you have to see the stopping as an opportunity to get in more practice.  In a traditional lesson, a horse might do 5 canter departs, so after 10 rides, it would have done 50 departs in total. But the clicker trained horse might do 20 canter departs in one session.  So after 10 rides, it would have done 200 canter departs.  Which horse is going to be better at canter departs?

 

She likes to think of the clicker as a skill accelerator.  Working in small pieces, each of which is repeated many times, is about deep practice and strengthening the individual components. This forms a really solid foundation for training related behaviors, or to which one can add new behaviors.

 

If you are interested in learning more about Daniel Coyle and Carol Dweck, I suggest you do an internet search.  I was going to post links, but there is a lot of information out there and it’s easy to find. They have both done TED talks and have books available.  If you want to read more about Alex’s thoughts on this topic, she has several blog posts on mindset and deep practice and you can find them at her blog  (www. Theclickercenter. com).

Steve White – Training Resilience

 

Steve White (www.proactivek9.com) has many years of experience training and working with police dogs and is always looking for ways to teach dogs and handlers to do their jobs better.  Part of doing the job better is making it so the dog can be good at its job without suffering from the kind of stress that tends to come along with that kind of work. 

 

In his presentation, he talked about resilience, which is performance in the face of adversity and the ability to bounce back and do it again.  Resilience could also be defined as the ability to spring back into shape and to recover quickly from difficulty.   He listed the following as components of mastering resilience:

  • Crisis preparation (have you prepared adequately?)

  • adaptive coping (can you adjust to the unexpected?)

  • psychology of happiness (understanding)

  • attitudes (building positive attitudes)

  • calm under pressure (training for)

  • skills (good handler skills)

  • managing distressing emotions (need this for both the dog and the handler)

 

Resilience is a hot topic right now in many fields.  While resilience is often measured in terms of individual performance, when those individuals are part of a larger group their resilience can have a more widespread effect because other people are counting on them. 

 

Because of this,  a lot of companies and organizations are looking at resilience training.  He did mention that resilience training is not without controversy as it has components of mindfulness and positive psychology and some people object to the spiritual aspect of it.   If you want to read more about resilience training as applied to people, two authors who write about it are Martin Seligman (Learned Optimism) and Richard Davidson (he has several books).

 

Steve’s interest is in looking at resilience as applied to dog training and he broke it down into two types:


·    Emotional resilience - “the ability to regain emotional balance after a setback rather than wallowing in anxiety, anger, depression or other emotion.”  Richard Davidson

·    Behavioral resilience – “the ability to retain both short and long term performance after a setback rather than shutting down or going off track.”  Steve White

I’m not sure that he means that these two types of resilience are independent of each other, but more that you might want to start by focusing on one which can lead to the other.   Steve talked mostly about behavioral resilience.  If a dog is very comfortable and fluent in its work, then emotional resilience should follow.   

 

So how do you build resilience?

 

There are some parallels with how resilience is taught to humans.  Some of the key factors in working with people are:


·         Calming skills

·         Rehearsal

·         Antecedent significance

·         Consequence mitigation

If we apply this list to dogs, we can come up with some essential components for creating resilience. Steve put this into a formula which is:

 

R = (resistance to ratio strain + resistance to distraction) x motivation

 

Ratio strain is a measure of the decrease in the fluency of a behavior when you go to longer variable ratio schedules.

 

Resistance to distraction is the ability to work under many different conditions where other reinforcers are available.

 

So to build resilience, what do we need to do in our training?


·    Honestly assess fluency.   Every behavior should be tested under many different conditions and assessed for accuracy, intensity, latency, duration, distance and distraction.  As you might guess, it is helpful to actually have a way to record this (fluency assessment form) so that your evaluation is complete and you can track your progress.

·    Build resistance to ratio strain and distraction.  This is part of fluency but requires that you evaluate the behavior under less than ideal conditions.  Think of all the possible scenarios in which you might need a behavior and train under those conditions. Steve had a video of practicing in water because sometimes that’s where the “bad guys” are.

·    Socialization, stress inoculation and generalization.  Are you taking these into account in your training?

·    Narrow repertoire width – it is better to do a few things well than many things only so-so. I asked him about this because it would seem that having many behaviors to reinforce might make it easier for the dog to be successful, but he says his dogs are all specialists. It is much easier to keep a few behaviors strong than it is to maintain a larger number.


·    Motivation – motivation is one of those “hard to define” things where you know when you have it, but don’t always know how to build it.  Part of motivation is finding the best reinforcer for that dog.  Dogs can work for many things including food, a ball, toys, tug, praise, etc… They had a police dog that worked for a teddy bear, so guess what the officers was carrying around in his back pocket?  Steve says that finding/building motivation is about building a big enough “why.” Why does the dog do it?

Summary: 

  • We need resilience and it can be cultivated.

  • Build enough "why" - what's in it for the dog?

  • Try a mindfulness based approach, but don't forget the importance of antecedents and consequences

 

I think a lot of this can be applied directly to horses.  If we want to take our horses out and about, then we need to prepare them for all the things that can happen and make sure our behaviors are really strong.  I thought this presentation dovetailed nicely with Phung’s presentation on training for free flight because good results are the product of careful and thorough training.  Both talks are inspiring me to think more about some of my training goals and add some extra steps. 

 

 

Phung Luu-  Birds in Flight

Phung Luu’s company, Animal Behavior and Conservation Connections, is an educational organization  that offers “unique free-flight education programs featuring native and exotic birds. Our goals are to connect people with animals and contribute to conservation of wild animals via our presentations.”

 

They do this through bird shows where they have birds fly over the audience’s heads, showing natural behavior and interacting with people.  It’s a totally different experience than watching a bird from a distance or in a cage.   While the end result can seem simple (the bird flies from one person to another), there is a tremendous amount of preparation and training that goes into teaching the birds to do this.   

 

In his talk, Phung explained the process and what they look for at each step along the way.  While the steps I am going to share are written for training birds for free flight, I hope you will read them and think about how they can apply to training many different behaviors.   The steps outline a nice system to train behaviors efficiently and effectively which are two of the goals of Phung’s program.

Before he started Phung had two points. One was that birds know how to fly, if we let them have time to practice and learn. We don’t “teach” birds to fly; we allow them opportunities to learn to fly.  He also said that as with most training, what you get out is directly related to what you put in… Input = Output.  

 

He said “If you want incredible flight, you have to put out effort to get it – you can cheat the system every now and then, but not often.”

 

Step 1:  Develop trust in interactions with the trainer


·    When do you start?  There are some species where they start working with them at an early age, either by imprinting them or hand rearing them. There are other species where they prefer to have the birds be reared by their parents and start later.   It depends upon whether a high level of interaction at an early age makes the birds more confident such that it minimizes stress (good for more flighty species), or if the confidence makes them a little too bold (which can be a disadvantage with predatory birds like hawks and owls). 

·    They also take into consideration the long term affect of early interactions with the trainer.  Phung said they always have to consider if the amount of interaction is done to make the job easier for the trainer, or for the welfare of the bird.  Some birds are easier to handle if you start earlier, but too much human interaction can interfere with their ability to interact naturally with their own species.  It’s important to build a relationship with the trainer without interfering with the natural development of the bird so that it can interact with its own kind. 

Step 2:  Establish trust and associations with the environment


·    Psychological component:   It’s important to expose the birds to as many natural experiences as you can so that they are comfortable with many different types of environments and changes in the environment. They take the young birds out and let them see lots of different things and also let them walk, run, or fly in areas with different ground cover, trees, bushes, etc…


·    Physical development:  Early development of flying skills is important.  The young birds are given lots of opportunities to fly so that they can practice their flying skills, get stronger and learn to fly under different conditions.    This is about developing motor skills.

Step 3: Build a 100% positive association between the bird and the trainer (this is the target)


·    The relationship between the bird and the trainer is crucial.  If the relationship is not solid, it will show up in the bird’s willingness to return after free flight and also in all the other husbandry behaviors that are part of caring for the bird.  He had some nice video of vaccinations and talon trims with birds that were very comfortable being handled by their trainer. This comes from having good mechanical skills and spending time developing foundation behaviors.


·    Develop, build and refine purpose and attention span for the learner.  Teaching birds to have longer attention spans or more focus on the trainer comes from a good balance of food and other reinforcers.  In some cases, the motivation may be all about the food, but birds that only work for food are going to be less reliable than birds that have learned to work for other things. This is where secondary reinforcers and the trainer’s skill in creating interesting training sessions are important.


·    Apply error-reduced teaching strategies.  It’s always better to set your learner up for success so that it can learn to do behaviors with confidence.  For free-flight birds, Phung takes into account variables like weather conditions so he is not asking a bird to return into the wind, or to fly on a day when there are not good thermals.    Error-reduced teaching strategies also include choosing to set up training so that unwanted behaviors are less likely.

Step 4:  Test and establish the learning speed of an individual


·    Phung didn’t spend much time on this, but the idea is to review each bird’s progress and learning speed to see if it is ready for step 5.

Step 5:  Allow the bird to learn, experience and refine its flight skill.


·    By this point the bird has already had practice flying in a familiar environment and should be ready to fly in new environments or in the presence of novel objects.  Taking the bird to new locations can provide additional challenges and they can also vary other conditions (weather, time of day, wind conditions, etc…).


·    Safety considerations:    Part of learning to fly well is learning to avoid potential dangers in the environment.  There are some risks associated with flying and Phung takes time to teach the birds about power lines, high tensile wire and other possible structures they might encounter while flying.


·    Introducing telemetry:  Another way to minimize the risk to the birds is to use telemetry to keep track of the bird as it flies.  There are several types of tracking equipment that can be used and he often uses a pack that mounts on the bird’s back. He desensitizes the bird so it is comfortable having it attached.  Telemetry is not just useful for finding lost birds, it also provides information about whether not they need to go search for the bird. In general, they want the bird to find its way home so the tracking device allows them to locate the bird and then wait a bit to see if it will return on its own.

Step 6:  Identify target (goal behavior)


·    With birds of prey, they often demonstrate lure chasing.  Phung had a video that showed an exercise they do with the birds of prey. They teach the bird to catch a mouse that is thrown up in mid-air as the bird approaches.  This is a way to evaluate the bird’s skills, learning ability and also to see what other training might need to be done before this goal can be reached.

Step 7:  Strengthen flight through repetition and increased criteria


·    Continue to expose the bird to new situations and train for stronger behaviors.  If the bird is going to be used in a show, they might want to train additional in-flight behaviors.  This step is about increasing fluency and allowing the bird to continue to develop its flying skills.

These basic steps can be adapted to apply to training any behavior.   I can think of a lot of applications for horse training, not just for liberty work, but also for any kind of work where there is the potential for competing reinforcers or distractions.  Phung said that if the training breaks down, he always goes back to Step 3 which is building a 100% positive association between the bird and the trainer.  He had some nice video showing three Crowned Cranes going through the steps and their final flight. 


Fun Fact:  They talk to their birds while the birds are flying. Sound travels very far in a vertical direction so a bird flying 100 feet up in the air can easily hear a person on the ground.

 

Back to Articles