All Whacked Out

This blog post is going to consist of a (very long) rebuttal to a recent blog post by Gary Wilkes, What is Real Clicker Training?  There is so much therein, which is counter to my entire paradigm about dogs and dog training. 

Wilkes believes it is a necessity to have "unpleasant consequences” for unwanted behavior or for failure to obey.  On the surface this sounds reasonable.  Missing in his analysis is a discussion of what is necessary or even appropriate "unpleasant consequences".

I maintain it is unacceptable to allow a guest to bait a dog into unwanted behavior and then to punish the same unwanted behavior (as shown in the video) with bonking.  The dog did not jump on the guest until invited up by the hands spreading wide.  To me this is patently unfair when you could just as easily set up a training session and teach the dog how to behave correctly and reward those successes.  If you are certain you are going to have a large number of stupid behaving guests who will make inviting gestures (spreading the arms toward the dog), then simply spend some time teaching this gesture as a new signal to sit.  Why not teach the dog an opening-door is another cue for sit?

So what is the problem with the approach shown?  For one thing, this is presented as only taking three tries to get from untrained to perfect behavior, still its clearly an edited video - so we don't really know how many times the dog got bonked.  In the first clip, the dog has a full body tail wag, a sign of a dog who is relaxed and comfortable.  By the end of the treatment, the tail wag, while still wide, is now lower and circular, showing a nervous quality.  The dog is now looking at the owner in a way which suggests anxiety.  Remember, just because the tail is wagging does not mean the dog is happy - in many dogs it is a simply a sign the dog is having an emotion, also dogs wag their tails to pacify others.

I will not say the bonking method mixed with marker training doesn't work, but remember, just because you can do a thing doesn’t mean you should.  This method has potential side affects.  One of the main side affects is a dog who now associates the bonk-giver with the bonk.  I don't know about you, but I want my dog to trust me and know I will keep her safe.  I don't think my dog should ever have mixed emotions about what to expect from me.  I don’t want her to watch my hands and ever wonder if the thing I’m carrying will be used to hit her.  I can provide relevant consequences without hitting her.  I don't have to use some tool and give it a cutesy name to cover up the aversive nature of the strike.

The other big side affect is when the dog associates an aversive with what they are looking at when it happens.  This dog could just as easily believe the approaching man is the reason he got bonked (i.e. not the jumping up behavior).  We don’t know how this dog will react over the long term.  It could be the behavior is only temporarily suppressed and may resurface.  Any question of potential negative side affects could have been avoided entirely with strictly reward-based training to teach better door manners.  I am particularly fond of how Victoria Stillwell (It’s Me or The Dog) approaches this problem.

Wilkes’ fault-finding of Karen Pryor’s book, Don’t Shoot the Dog, isn’t particularly accurate.  It said you can teach an animal to do anything they are physically and emotionally capable of achieving. I found her book made a point of a saying a dog could change/control their own behavior every bit as reliably (and perhaps more so) to earn rewards as to avoid punishment.  Punishment often gives a false impression of improved behavior due to the overall suppression of behavior, which is not always permanent.  Aversives do have potential for more side affects than reward based training.  Karen Pryor is not the only person to say these things.  Many authors with more actual dog training experience are saying the same things.

And speaking of Pryor’s experience in marine mammal and other not-dog training – why would we assume her training expertise is invalid because its not dogs.  Dogs learn by the same processes as every other species.  If zoo trainers can teach an orangutan to take a small urine sample cup, pee into the cup and then hand it back without spilling it – well I just think it shouldn’t be too hard to teach the family dog not to jump on people.  If you really want to understand the complexities of training other animals, check out Kathy Sdao’s blog about E.T. the Walrus.

Weighing in at two tons, hormone-laden E.T. could easily have killed any of his handlers and had killed other exhibit marine mammals.  Using reward based training and some environmental enrichment, E.T. was transformed from a dangerous bully to a talented boy.  One of his behaviors was to voluntarily permit a blood draw.  Kathy is fond of saying you learn what is possible in reward based training when you can’t punish the animal because it might kill you.

Wilkes says the four assumptions are “neither scientific nor true”, yet provides no science to prove them wrong.  He says the assumptions of “all positive” training are not applicable to predatory beasts like dogs.   So I guess the other predatory zoo animals trained with reward based training are some kind of anomaly.  You know the ones – lions, tigers and bears, oh my.

Wilkes unfairly describes the thinking of reward-based trainers in very all or nothing terms.  He also alludes to his reality where the “advanced clicker trainers” have only trained a few dogs.   I have used aversives, but only when I ran out of reward-based options and only to the minimum extent necessary to create an opening for reward based solutions.

Bob Bailey talks about people who say they had to use punishment because “nothing else worked”.  When pressed for what they had tried, they only tried 2 or 3 options before resorting to punishment.  I can think of ten different ways to teach a dog not to rush the door or jump on guests.  Doesn’t it behoove us to try most of them before bonking the dog?

Wilkes’ blanket statement, “only aversive control is capable of stopping a dog from being a dog when it feels like it” is not true.  If teaching a terrier not to chase squirrels is the gold standard, sorry Wilkes - its been done.  In fact the opportunity to bark at squirrels became the reward one owner used in agility training her terrier on her way to a MACH.  The chief difficulty with this kind of training is controlling the squirrel resource. 

I don’t know about the behavior of Karen Pryor’s dog, but found Wilkes’ mind-reading abilities astounding, the way he read her mind and knew so exactly her motivation and thinking for the things she said about training a squirrel chasing terrier.  I also do know the old saying, “the cobbler’s kids got no shoes”.  Many people admit their own dogs’ training suffers when they got famous and were so often gone on the lecture circuit, so its easy for me to assume Karen’s dog is not an exception.  Some famous trainers make it work, Denise Fenzi springs to mind, but I think its because they have an internal motivation and commitment to train their dogs and they have competitive goals to meet.  For someone like Karen, who doesn’t compete and only enjoys her family dog, I think spending the time tweaking every little behavior just doesn’t have the same driving force.

I do find it interesting he never defines “aversive control”.  He never discusses what the limits are for its use.  Certainly the example he showed in the video didn’t justify the breaking of trust, in my opinion.

Wilkes repeats his claim about there being no “real world examples of dependable performance with dogs”, which is simply false.  Service dogs of all stripes are trained using all reward training.   Many highly successful agility dogs are trained without aversives.  All the nosework and barn hunt dogs I know are trained without aversives.  I don’t know of any freestyle dogs who are bonked down the road to success.  Denise Fenzi’s dogs are trained without aversives…  so much so - her dogs find losing an opportunity to train/perform their biggest aversive.  Even if there were no examples of really reliable high quality training done with all positive training – the real fly in the ointment is how many thousands and thousands of dogs trained with aversives and competing at high levels are unable to demonstrate truly reliable behaviors.  So clearly “aversive control” is not all its cracked up to be – its not the panacea for all dog training problems everywhere.

Wilkes tries to claim scientific training is flawed and can’t be used reliably in the “real world”.  He completely side-steps the fact of how his own brand of bonk and mark training fits so neatly into the operant conditioning model.  While he states how “traditional dog training methodology has worked for more than 15,000 years in hundreds of different applications with hundreds of millions of dogs”, he fails to address the abuse and near torture which has been heaped upon many of those hundreds of millions in the name of traditional training.  In fact, at one point in human history, it was scientifically fashionable to say animals could not reason and had no feelings, essentially they were no more than meat-machines.  This view helped to justify the cruelty used to train and work animals for centuries.

Personally I’ve been to at least 300 agility competitions and have yet to see even ONE dog leave the ring to attack another dog.  Of the dogs with start line stay issues, as many were aversive trained as were positive reward trained.  One of the worst obedience stay problems belongs to a Weimaraner who is trained traditionally and jerked around or "accidentally" has its foot stepped on when it fails to stay.  The dogs I’ve seen with table issues were dealing with anxiety issues and I do not believe adding more anxiety in the form of punishment would suddenly cure the problem.

The reality is – any form of training, any training philosophy is only as good as the person doing it.  As Wilkes says, traditional aversive training has been around for 15,000 years, so it stands to reason there are many more people who are really good at punishing dogs.  Compared to the thirty years of reward based dog training.  Reward based training has come a long way and its getting better all the time.  To be sure, there are still many people who are ineffective all positive trainers.  My guess is they’d be just as ineffective with punishment.

Wilkes wants you to abandon an all-positive approach and return to a more traditional style of training.  Be very, very glad the Wright brothers, Cugnot and Rivaz didn’t abandon their early powered transportation attempts even though they were slower than the local mule.  If they had, you’d be driving a buggy instead of a Chevy.

The use of rats and pigeons by scientists is prevalent because of space constraints, its about economics.  Quite simply you can work with a lot more rats and pigeons than you can dogs for the same dollars.  Also, for a long time, studying dogs was considered unscientific and was largely frowned upon.  To leap from there to a conclusion saying learning processes are different is just unscientific.  I don’t know what dogs Wilkes has been around, but I have never met one who sneered at a treat when hungry because it was the wrong time of day. Seriously?  My dog can tell me down to the minute when her next meal is due.  In over four years, she has only refused treats a couple of times and it wasn’t because of the time of day. To say a dog doesn’t learn the same because its progenitors ate about once a week is just silly.

The failed examples he gives of early attempts to train dogs using Skinner-Breland style methods simply did not take into account the one thing which is truly unique about dogs – their relationship bias.  Dogs are far more sensitive and intune to human behaviors, picking up the subtle nuances of our moods and attitudes, than any other species.  I believe the early failures were because of trying to be too clinical, depriving the dogs of the relationship connection.  We’ve bred dogs to be connected and they notice the subtlest of things.  Successful trainers, like Denise Fenzi, get great results using all positive training because they use the relationship connection to enhance training.

His explanation about some carnivorous predators being more like grazers (dolphins) than others (wolves) doesn’t fly.  He states wolves eat things which are 10 – 15 times larger than they are.  The strongest studies on the diet of wolves shows they eat a huge variety of other animals, from ungulates to earthworms.  When they do eat large game, they prey on the old and infirm.  Its not because they don’t want to risk injury, its because they are conserving their own energy.

I found his argument about “instinctive drift” unconvincing at best.  While “instinctive drift” can be an issue, I saw nothing to suggest using Wilkes’ “aversive control” would resolve it.

Oddly, one of Wilkes’ arguments to refute the effectiveness of all positive training, actually proves the problem we, the ideologically myopic, have with punishment.  He said “a host of others have failed to integrate a dog’s ability to “take the hit” and continue to perform happily”.  For many dogs this is true… to the point where dogs will continue to perform despite showing calming signals and other negative symptoms of their internal stress.  This is more prevalent in some breeds than others.  Some dogs will tolerate more and more and more punishment, to where you must use excessive force to control/inhibit unwanted behaviors.  I fail to see where this is a good thing.  Again, just because you can doesn’t mean you should.

In any case, using reward based methods have greatly expanded the number of breeds which can be highly successful in performance sports.  Its no longer necessary to own one of the few breeds which are highly resilient or obsessive compulsive if you want to compete in dog sports.  From Bulldogs to Borzois, Dachshunds to Dalmatians, Huskies to Hounds – there are now breeds of every size and temperament able to compete in many different sports, because they don’t have to endure aversives to be trained.

Wilkes insists the only way to break through any dog’s special distraction is to use aversives/punishment.  Those of us who align with reward based training know we can split the elements of the distraction down to its smallest effect and teach the dog how to work through it one step at a time.  Those who work with reactive dogs (regardless of their trigger) know the most important thing we can do is to work through the individual bits of the trigger.  To quote Bob Bailey, “Be a Splitter, Not a Lumper.”

Wilkes assumes all clicker trainers have a rabid aversion to aversives.  He assumes there can be no reliable training without aversives, then demonstrates a training scenario where use of aversives was entirely unnecessary.  Any ten of my students could have solved the jumping up without hitting the dog.  Wilkes assumes aversives are a necessity – thus he jumps to their use almost immediately (just watch some of his youtube videos – I sure don’t want to be a bar setter in the ring with the agility dog he threw the bonker at).  Whereas I assume aversives are a last resort – thus I almost never need them.

To be clear, while Wilkes uses a clicker (the little plastic noisemaker) in his training, he only uses it as a marker signal.  Clearly (based on his article and the few youtube videos I watched) he has no interest in finding a less aversive way to modify dog behavior and even the simplest untrained dog behavior must and will be met with force.