Monday, August 5, 2013

TSSSST: MY CHALLENGE TO MAKE ALL TRAINERS MORE EFFECTIVE TRAINERS





This past week I have been inundated with contacts who are imitating the Cesar Milan hissing noise in an attempt to control behaviour they don’t like in their dogs.  Most of the time, all they are doing is annoying the people around them with their hissing noises, and the dog isn’t stopping the unwanted behaviour.  It helps to understand what is actually going on when Cesar Milan, the Dog Whisperer, hisses at a dog and why it works.  To begin with, you might wonder how I know what is happening and here you go.  I use what I know of Applied Behaviour Analysis to observe, record and analyze what is happening, and honestly there is no magic to it.  When you understand the science of behaviourism you too can observe, record and analyze and figure out what anyone is doing when they train.  This will make you a better trainer.

Applied Behaviour Analysis is the field of study related to determining what is happening in a training situation.  It is reliant on observations of actual behaviour and training events, collection of data and analysis of the events that occur before and after training, or what might be termed treatment.  Studying this field will allow you to analyze any trainer's work and determine what is really happening.  Every trainer should be aware of the basics in Applied Behaviour Analysis.



Here is how I do it.  I select a clip to watch, and turn the sound off on my computer.  I watch the video and look for the behaviours that the dog is doing and I make some notes.  At this point I am only looking at what the dog is doing.  I look for a likely target behaviour, usually a behaviour that is undesired, and then I watch to see if the frequency is going up or down.  I can usually guess what behaviour is the target by two things; how frequently it is shown and if it increases or decreases.  In a five minute clip of the Dog Whisperer, I will usually see a dramatic decrease in one or more behaviours. 


It helps that National Geographic has great editors who often give me the target behaviour in slow motion.  Once I have determined what I think is the target behaviour, I turn on the sound and listen to the first part of the video to determine if I am right or not.  I am usually right.  The Dog Whisperer isn’t interested in teaching dogs to sit up and look pretty, or to fetch a Frisbee; he is interested in the drama of fixing big things like aggression.  Most of the time, we are watching dog to human or dog to dog aggression, with a smattering of phobias thrown in to make us think that the show is demonstrating a broad repertoire of behaviour solutions.  As an aside, I would love to see someone do a statistical analysis of the show to see how many different problems he solves and what frequency each of the problems is shown on the show.  Never the less, once I have the target behaviour confirmed by the audio for the clip, I turn the sound off and watch again, and record what time each incident of the behaviour occurs.  Now I have some data; I know both how often the behaviour occurs on the video clip and I know when the behaviour occurs on the video clip.


This time, I watch for what might predict the behaviour coming on.  If I am watching a dog who is aggressive to people, I watch what happens just before the incidents I have noted on the time counter I was recording before.  Let’s just say for example that if a stranger reaches for the dog, the dog growls and snaps and lunges at the stranger.  And let’s say that in the first 90 seconds of the film we see this happen eleven times.  We know that this tape is edited, so we really don’t have great baseline data, but National Geographic would like us to think that they captured the dog exhibiting the behaviour in a linear way; they want us to think that they are showing us what happens every time a stranger reaches in to touch the dog. 


At about the 90 second mark we see the Dog Whisperer himself reach for the dog and predictably the dog lunges for him.  This time, the DW strikes back, and using either a clawed hand, or his first two fingers together, he “touches the dog”, and the dog shrinks back.  We see this happen several times.  The dog lunges, the DW “touches the dog” and the dog shrinks back.  Magic!  Or is it?  It is important to notice where the dog is being touched.  Often, he is touched just behind the ear, on the jaw line.  Sometimes he is touched in the area of the rib cage.  All of these areas have sensitive nerves close to the surface.  If you take two fingers and press them below your ear, next to your jaw, you will feel how little force it takes to cause pain. 


Over the next minute or so, we see the dog gradually stop lunging at the DW.  We may also observe some lip licking, blinking and ear pinning.  I would also put this in my notes.  My notes might look like this:


Time (minutes and seconds)
Observation
0:00
Credits/background of home, family and dog
0:05
Dog approaching family member, no reaction
0:07
Stranger reaches for dog on leash
0:08
Dog’s eyes go round, lips raise, tongue draws back
0:09
Dog lunges to end of leash and falls back



Time Elapses between 0:10 and 1:31 (I would fill in the chart, but this is just an example of what I might do, so I won’t bore you with eleven repeats of the same sequence)


1:32
DW reaches for dog
1:33
Dog’s eyes go round, lips raise, tongue draws back
1:34
Dog lunges, DW touches dog
1:35
Dog recoils, licks lips, blinks
1:36
Dog turns head away, pins ears
1:42
DW reaches in again
1:45
Dog’s eyes go round, lips raise, tongue draws back
1:46
Dog lunges, DW touches dog
1:47
Dog recoils, licks lips, blinks
1:48
Dog takes two steps back
1:49
Camera goes to close up of dog blinking and lip licking
1:55
DW reaches in again
1:56
Dog’s eyes go round, lips raise, tongue draws back, body freezes
1:57
Dog lunges, DW touches dog
1:58
Dog recoils, turns away, lip licking, blinking, hides behind owner holding leash
2:00
Close up of dog shaking as though he is wet
2:01
DW approaches dog, pauses, reaches out
2:02
Dog freezes
2:03
DW begins to stroke dog, dog crouches down, arches back, tucks tail, licks lips, blinks
2:04
Dog Freezes, DW freezes
2:05
DW continues to stroke dog, takes leash, dog averts gaze




I will do this sort of recording second by second for whatever length the clip runs.  I want to have data to analyze.  At this point, I would probably start doing a preliminary scan of the interactions.  All lunging and snapping is preceded by a stranger reaching for the dog, so I presume that the target behaviour that they are going to work on in this episode is to stop the dog from lunging and biting strangers who want to pet the dog.  I don’t want to talk about how I would handle the case, I only record what I am seeing.  Watching and being annoyed at what someone else is doing under the guise of training is not helpful; I want to learn what the trainer in this case is doing.  As I watch the DW touching the dog, I am seeing the dog decrease his lunging and snapping behaviour.  Pulling out my understanding of the quadrants, I can see that when a behaviour is decreasing, then punishment is at play.  I can also see that the trainer is doing something, so I know that this is positive punishment. 


I will analyze the whole video, and as a synopsis I will just give you the analysis.  For the first 90 seconds we get baseline data.  Then we get about a minute of training data with the DW.  Then we get the rest of the video showing the decrease of aggression over time with a variety of strangers.  Specifically, I see a series of strangers being introduced to the dog, the trainer touching the dog, the dog showing a variety of displacement behaviours such as lip licking, blinking and retreating, shaking off, sitting and scratching, and hiding behind the trainer and the aggressive behaviours of the dog decrease with experience.  I can see a decrease in the target behaviour of lunging and snapping.  I don’t want to look at emotional state in this situation in this case, because I am analyzing what the trainer is doing, not what the dog is feeling.  As someone who is well versed in ethology, I can tell you for certain that the dog is distressed, but I don’t want to talk about that right now; that doesn’t help us to analyze the training that is happening.


By the end of the video, we see a dog who is allowing strangers to pat him.  Finally it is time to watch the video with the sound on.  By this time I know this video fairly well, and may have watched it ten to twelve times with the sound off.  When I watch with the sound on, I start to record what is said at each point or what noises I hear and add that data to my chart.  Now the section where the DW is interacting with the dog looks like this:


1:32
DW reaches for dog (long dialogue about taking control finishes)
1:33
Dog’s eyes go round, lips raise, tongue draws back (TSSST from DW)
1:34
Dog lunges, DW touches dog
1:35
Dog recoils, licks lips, blinks (discussion of dog’s dominance and how this is unwanted)
1:36
Dog turns head away, pins ears
1:42
DW reaches in again
1:45
Dog’s eyes go round, lips raise, tongue draws back (TSSST from DW)
1:46
Dog lunges, DW touches dog
1:47
Dog recoils, licks lips, blinks(discussion of dog’s dominance and how this is unwanted)
1:48
Dog takes two steps back (further discussion of dog beginning to show calm submission)
1:49
Camera goes to close up of dog blinking and lip licking
1:55
DW reaches in again
1:56
Dog’s eyes go round, lips raise, tongue draws back, body freezes (TSSST from DW)
1:57
Dog lunges, DW touches dog
1:58
Dog recoils, turns away, lip licking, blinking, hides behind owner holding leash (discussion of dog’s dominance and how this is unwanted and how the dog is being helped by learning calm submission)
2:00
Close up of dog shaking as though he is wet (discussion of dog’s dominance and how this is unwanted, how we are starting to see calm submission and how the dog must know his place)
2:01
DW approaches dog, pauses, reaches out
2:02
Dog freezes (DW comments that the dog is showing his calm submission)
2:03
DW begins to stroke dog, dog crouches down, arches back, tucks tail, licks lips, blinks
2:04
Dog Freezes, DW freezes (TSSST from DW)
2:05
DW continues to stroke dog, takes leash, dog averts gaze




I have highlighted the entry at 2:04 because something interesting happens here.  Now, the DW is using the sound, but not the touch, and here is where we find the key to understanding what is happening.  This isn’t news to those of us who study behaviour but this is a major mystery to most viewers because they don’t have the tools to analyze what is actually happening.  Without understanding the whole picture, it LOOKS like the TSST sound stops the behaviour, somehow or another magically.  If you look at the whole analysis you can see that the first two times that the DW makes his hissing sound, that is immediately followed by him touching the dog.  Remember what it felt like to put two fingers behind your ear, near your jawline?  It hurts.  It doesn’t take long for any animal to learn that a particular noise or signal will predict pain.  You don’t have to beat a dog to hurt him; as with many martial arts, it isn’t strength but rather a good knowledge of how the body works that allows you to control an opponent.  Often by the third or fourth repetition the dogs in the hands of Cesar Milan figure out that the hissing sound is going to predict pain.  The dog wants to avoid pain, so when he hears the noise, he stops his behaviour. 


So what is the hissing noise?  It starts out life as a conditioned positive punisher.  We use the term conditioned to mean learned in this context.  He could have used any short, unique sound; he could have clucked his tongue, said “beep”, used a handheld electronic gadget or he could have sounded a little bell.  In the final Harry Potter movie, we see the goblin, Griphook, use a large bell in the same way to control a dragon.  He rattles the bell and tells us that the dragon has been trained to expect pain when he hears the bell.  Tssst is nothing more nor less than a conditioned positive punisher.  Over time though, and you can see this in the chart at 2:04, the conditioned positive punisher morphs into a delta signal.  A delta signal tells the animal to change his behaviour and avoid the unpleasant consequence. 


This is the true power of the hissing noise.  If it is connected with pain, then the dog can learn to avoid the pain by changing his behaviour.  If however, the noise has not been properly conditioned, or if the dog doesn’t experience the pain when he doesn’t change his behaviour, then the hissing just becomes what it is in most of the cases I have seen; a noise that annoys the humans and does nothing to change the behaviour of the dog.  When the DW does this, he is absolutely consistent.  He teaches the dog methodically that the hissing will produce pain.  Then he sets the dog up to make a mistake over and over and over again so that the dog can learn that if he engages in a particular behaviour, he will get a warning and if he doesn’t change his behaviour, he will experience pain.  When you chart it, analyze it, and understand the process behind it, there is no magic any longer.  It is simply a conditioned positive punisher that morphs into a delta signal and the dog learns to avoid the pain by changing his behaviour.


I am not going to say one way or another if I think that Cesar Milan is a good trainer, or a bad trainer, or a good person or a bad person; I will leave that rhetoric for someone else.  What I want to show is that when you do a proper analysis on his work, you can see what he is really doing.  As a behaviour consultant, when I watch interactions between trainers and dogs, my job isn’t to say that someone is right or wrong; my job is to pick apart what is happening and then figure out what to change to get the results we want.  Most of the time I can do a jackleg job of this in real time as I watch people train.  Some of the time, I cannot see exactly what is happening without going through the whole process and if I want to communicate with a fellow practitioner, I might make a video with my analysis  written out in long form.  Most of the time, trainers don’t need to do the long process.  When you understand the theory behind what is happening, the descriptions by the trainer cannot fool you and you can pick apart what is really happening.  Being able to pick apart what is happening is part of what makes me a better trainer than many of the trainers I meet.  When training goes right or goes wrong, I can use the process outlined above to figure out what I can change in order to achieve the change in the target behaviours I want.  My wish is that every trainer could do this and would do this.

No comments:

Post a Comment