Symbolic Understanding in Dogs

The use of symbols in communication has long been considered to be solely the domain of humans, great apes, cetaceans and a few bird species.  Human children typically start to understand basic vocal and gestural symbols at about a year of age.  By the time that most children are 2 to 3 years old, they are also able to understand the concepts of images and replicas.  As reported in Science News, a recent study demonstrated that dogs appear to be capable of a similar level of symbolic understanding.

Border collies quickly realize that their owners want them to fetch a toy from another room when shown a full-size or miniature replica of the desired item and given a command to “bring it here,” say biological psychologist Juliane Kaminski of the Max Planck Institute for Evolutionary Anthropology in Leipzig, Germany, and her colleagues. Even a photograph of a toy works with some dogs as a signal to fetch that toy from an unseen location, the researchers report in an upcoming issue of Developmental Science.

Three dogs already trained to fetch objects succeeded on both replica tasks right away. Two untrained dogs got the hang of replica requests after a bit of practice.

“The most reasonable interpretation of dogs’ success in the replica tasks is that they understood that by showing a replica, a human was trying to communicate something to them,” Kaminski says. Dogs evolved a feel for how people communicate as a result of living in human settlements for thousands of years, she proposes.

Earlier studies have found that chimps, dolphins and other nonhuman animals have great difficulty retrieving objects after being shown replicas of those objects, even after many trials.

Dogs appear to have an innate ability to understand human gestures and body language.  Chimps (like the famous Washoe) and African Grey Parrots (like Alex) have demonstrated the ability to be taught sign or spoken language. Now it appears that Dr. Bonnie Bergin’s idea that dogs can learn to recognize and respond appropriately to written word and symbol cues may have been ahead of its time.  Her “teaching dogs to read“ appears to be just a short step away from their now demonstrated ability to grasp the idea of images and replicas.

Coincidentally, I have been working with young Audie for a couple of months on these kins of exercises.  I’ve successfully taught him to search for, find and then bring me an item that matches one I’ve shown him.  We’ve done this with a wide range of objects (shoes, boots, metal bowls, pens, bumpers, spoons, business cards and more).  Sometimes the object I send him to find is in a different room.  Quite often I don’t know where it is (hence the need to have him find it).  Still, he gets it right nearly every time.

The question that arises from this is; are dogs innately capable of these kinds of symbolic learning skills or does the ability only arise from a certain degree of training?  I suspect it’s a combination of both.  SAR work (like explosive and narcotics detection work) takes advantage of a dog’s ability to use scent as a symbol for the object he’s seeking.  In Learning to Smell, Donald Wilson and Richard Stevenson propose that smells “are outcomes of highly synthetic, memory-dependent processing that is further modulated by expectation, context, and internal state”  and they compare the process of learning how to smell to that of learning how to read.

Perhaps the combination of being a neotenized, highly social, scent-reading species that co-evolved with humans and has an innate skill to observe and interpret our behavior makes dogs uniquely well-equipped for certain types of symbolic learning. How much are they able understand?  Well, since they’re poorly equipped to talk or sign back to us we’ll have to wait a while longer to find out.

I’ll try to shoot and upload some video of Audie and I working on the “match game” this week.

Great Expectations?

Charles Dickens with one of his mastiffs

Today I came across a fascinating article in Seed Magazine on dopamine, artificial intelligence and social learning.  There’s so much good stuff in there that it may inspire a few posts.  Here’s the first one.

Dopamine is great stuff. It doesn’t just let us take pleasure in our world — it’s also key in helping us understand it.  Cambridge University’s Wolfram Schultz studied dopamine’s role in triggering Parkinson’s disease.   Schultz recorded dopamine levels in monkey’s brains to study how dopamine neurons died in the part of the brain that controls movement.  As he collected data for the study he noticed that dopamine neurons fired just before the monkeys were rewarded for moving.

Stunned, Schultz realized that he had just discovered the brain’s reward mechanism — and he’d done it by accident.  According to Seed:

His experiments observed a simple protocol: He played a loud tone, waited for a few seconds, and then squirted a few drops of apple juice into the mouth of a monkey. While the experiment was unfolding, Schultz was probing the dopamine-rich areas of the monkey brain with a needle that monitored the electrical activity inside individual cells. At first the dopamine neurons didn’t fire until the juice was delivered; they were responding to the actual reward. However, once the animal learned that the tone preceded the arrival of juice — this requires only a few trials — the same neurons began firing at the sound of the tone instead of the sweet reward. And then eventually, if the tone kept on predicting the juice, the cells went silent. They stopped firing altogether.

The data were fascinating and utterly perplexing.  Schultz knew that the dopamine neurons were a key part of the learning process, he just didn’t know how they were doing it.

Meanwhile, over at the Salk Insitute, computer scientists Read Montague and Peter Dayan were working on the temporal difference reinforcement learning (TDRL) model of artificial intelligence.  Their goal was to create a “neuronlike” program that could learn simple rules to perform goal-oriented behaviors.  The Seed article notes: 

The basic premise is straightforward: The software makes predictions about what will happen — about how a checkers game will unfold for example — and then compares these predictions with what actually happens. If the prediction is right, that series of predictions gets reinforced. However, if the prediction is wrong, the software reevaluates its representation of the game.

Montague and Dayan were holding the key to Schultz’s dopamine release as learning mechanism conundrum.  And Schultz had collected neurological data that supported Sutton and Dayan’s TRDL model.  Dayan discovered the link in 1991.  Seed quotes him as saying:

“The only reason we could see it so clearly,” Montague says, “is because we came at it from this theoretical angle. If you were an experimentalist seeing this data, it would have been extremely confusing. What the hell are these cells doing? Why aren’t they just responding to the juice?”

The truly fascinating thing about the model is that it’s based on expectations.  Predictability, correlation and contrast are the basis of learning — not reward and punishment.  I may just be a dumb dog trainer, but this isn’t news to me.

A predictable world makes sense.  Predictability builds confidence.  When you work with a shy or insecure animal, maintaining a high degree of predictability can build the confidence.  Correlation is the foundation of predictability.  Contrast tells us what things don’t correlate, it teaches us what things don’t work (and what works better than we expected).  It also works neatly in conjunction with correlation to teach animals how to generalize.  Add a little contrast to a lot of correlation and you’ve created the perfect proofing exercise.

These three processes work together to help an animal make sense out of the geopbytesof information that inundate it in day to day life.  And much of the beauty of the system lies in the fact that it’s a dynamic one.  The first time that we notice that A correlates to B we are surprised.  The neurochemistry of surprise makes novel events memorable because bursts of dopamine are emitted in the wake of unexpected rewards. After a few repetitions of the pattern, we store that bit (or byte, as the case may be) of correlative information away.  If, at a later time, the correlation fails, we’re surprised by the failure in our expectations and our neurons readjust to the new situation as dopamine production declines when we expect a reward that we don’t receive.  Sometimes those rewards are explicit ones like food treats or payroll bonuses – but they can also be implicit rewards like the excitement you feel when your team wins the big game or the thrill your dog feels when he finds his bumper in that tuft of grass where he expected it to be. 

But rewards alone won’t create learning.  We need the contrast that comes from errors too.  Again from Seed (bold mine):

“The accuracy comes from the mismatch,” Montague says. “You learn how the world works by focusing on the prediction errors, on the events that you didn’t expect.” Our knowledge, in other words, emerges from our cellular mistakes. The brain learns how to be right by focusing on what it got wrong.

And there you have it.  It’s impossible to learn unless you make mistakes.  The freedom to make mistakes is vital to intellectual growth.  This is why raising a dog in an over-managed way does irreparable harm to his intellect.  A dog can only learn to think for himself  if he’s given the opportunity to do so.  Making mistakes and coping with their consequences is a key part of learning.  It helps the dog evaluate the contrast he needs to generalize his learning to new situations — and to reject previous learning when it doesn’t fit a novel situation.

I think that it’s also interesting to note that the model appears to imply that the implicit dopamine rewards of making correct, correlative predictions are stronger and more long-lasting than explicit rewards like food.  This may help explain why the associations created by negative reinforcment (a concept most trainers don’t really understand) are very strong and long-lasting.

Dog as Inkblot?

Pareidolia, the phenomenon where a person believes he sees significance in stimuli that have none is the basis of Rorschach inkblot testing.  Psychologists who use the test believe that our interpretations of patterns in the blots are based on the patterns of our mental and perceptive processes, not the patterns of the blots themselves.  The Rorschach is a projective test where the patient is believed to project his personality traits into the inkblot as they interpret it.  Projective tests like the Rorschach are designed to encourage subjects to respond to ambiguous stimuli and thereby reveal hidden emotions or internal conflicts.

The test was developed by Swiss psychoanalyst and artist Hermann Rorschach in 1921 after he noticed that children that had different psychological problems  responded in different ways to the popular parlor game  Blotto.  Victorian Blotto players would create inkblots then compete in developing elaborate descriptions of them (note: this differs significantly from the modern version of the game).

The test quickly gained popularity and by the 40’s and 50’s, it was the test of choice among clinical psychologists. The test is still used today but it’s less popular than it was in the past because many psychologists see its interpretation as problematically subjective.  The shape and pattern of the inkblots are ambiguous, so each subject must project his own thoughts, feelings, and narrative to make sense of them and the therapist must use his judgement in scoring the results.  Because our responses to these kinds of tests are, at least to some degree, subconsciously elicited; projective tests like the Rorschach are believed to be especially useful in situations like criminal forensics and employment screening where the test subject may not be willing respond candidly.

Other projective tests like the Thematic Apperception Test, Children’s Apperception Test and Blacky test ask subjects to tell stories about pictures they are shown. These narrative tests provide insight on the content of a person’s thinking, whereas the Rorschach provides information on thought process and forms.  The Children’s Apperception Test and Blacky Pictures use drawings of animals to elicit responses and an Animal Thematic Apperception Test has been designed specifically to assess attitudes toward animals.

Why are projective tests valuable?  Well, as a species, we humans tend to greatly overestimate causality. We like need to see the world as a nice orderly place that makes sense, so we seek explanations for situations even when there are none.  This drive to make sense out of nonsense helps explain why a devout Christian sees an image of Jesus on a dog’s bum and a neurotic dog owner sees the Freudian reflection of his own unresolved bed-wetting issues when his dog refuses to pee outside.  Our dogs have become furry Rorschach blots.

Because dogs can’t talk, we dog owners often speak for them. Inwardly and out loud, we put their actions and expressions into emotional and intentional context.  Our dogs are more than just companions, we see them friends, brothers, children, protectors — even adversaries.  While dogs and men evolved to live together, no human is born speaking dog; so their behavior can be both comfortingly familiar and maddeningly perplexing to us.  To add another layer of complication to the matter, our social identities are tied to the way we perceive them our dogs too. We don’t just love our dogs for who they are, we love them for what we think they say about us.

The emotions and intentions we project onto our dogs typically take the form of complex narrative.  And while much of this interpretative narration is part of a healthy reciprocal, thoughtful relationship – sometimes it can lead to serious problems. As we project ideas and emotions onto our dogs we need to remember that they don’t think, feel and react the same ways we do.  And, like it or not, we should also keep in mind that part of the function of this narrative is to satisfy our desire to have our hopes and fears mirrored back to us.  If we project our fears onto a dog we can create a problem where one didn’t exist.  In projecting our hopes we can blind ourselves to problems that are obvious to others.  

Our minds have a very strong tendency to create images that fulfill our preconceived ideas, hopes or fears.  This can be as amusing and  harmless as seeing animal shapes in clouds. But when it’s combined with dysfunctional ideas or intense emotions, pareidolic observation can be a huge problem because it provides us a with very strong built-in validation of those feelings and beliefs. Our imaginations are really good at convincing us we’re right — especially when we’re not…

As a dog trainer I’m often confronted with dog owners who are completely and utterly convinced that they “know” what’s wrong with their dog — and who are also completely and utterly mistaken in their beliefs.  The woman who was a abused by her father and is now positive that her dog hates all men.  The man who seeks freedom from the restrictions and limitations his work and family place on him and is convinced that his dog can only be happy if he’s allowed to run free. These  people see the idealized dog they need to validate their world view; not the living, breathing one sitting right in front of them.  And they’re making their lives — and their dogs’ lives — a lot more complicated than they need to be.

Perhaps therapists should explore the relationships we have with our dogs instead of using Rorschach blots to ferret out our hidden thoughts, fears and desires.  Dogs are comfortingly familiar and charmingly mysterious.  Our images of them are ambiguous and emotionally charged at the same time. We’ve turned them into perfect projectors of our inner selves — like Rorschach blots on steroids.

Just Get Over It

In a series of studies published last year in the Journal of Personality and Social Psychology, Colgate University psychologist Kevin Carlsmith and his co-authors, Timothy Wilson of the University of Virginia and Daniel Gilbert from Harvard University explored the emotional costs of imposing retaliatory harm on a social transgressor. The abstract of the article presents their rather unexpected findings:

People expect to reap hedonic rewards when they punish an offender, but in at least some instances, revenge has hedonic consequences that are precisely the opposite of what people expect. Three studies showed that (a) one reason for this is that people who punish continue to ruminate about the offender, whereas those who do not punish “move on” and think less about the offender, and (b) people fail to appreciate the different affective consequences of witnessing and instigating punishment.

So, while we anticipate that exacting revenge will give us an enormous sense of satisfaction, they found that getting even just cements that sense of dissatisfied frustration more firmly in our psyches. If instead we ‘d just pull our heads our of our hinterlands and get over it — that sense of righteous indignation will fade and we’ll end up feeling less stress and frustration.

While human beings seem to have an enormously difficult time figuring this out, our dogs have a pre-programmed predisposition to forgive and forget.  It’s one of the best lessons they can teach us.

It also gives me a chance to harp on one of my pet peeves in the world of dog training.  This would be the difference between punishment and correction. Most new age dog trainers will tell you that there is no difference between punishment and correction. They’re both pigeonholed into the P+ quadrant of their sacred, reductionist operant conditioning diagram.

Dictionary.com defines punishment as:

1. the act of punishing.
2. the fact of being punished, as for an offense or fault.
3. a penalty inflicted for an offense, fault, etc.
4. severe handling or treatment. 

It defines correction as:

1. something that is substituted or proposed for what is wrong or inaccurate; emendation
2. the act of correcting.
3. punishment intended to reform, improve, or rehabilitate; chastisement; reproof.
4. the various methods, as incarceration, parole, and probation, by which society deals with convicted offenders.
5. a quantity applied or other adjustment made in order to increase accuracy, as in the use of an instrument or the solution of a problem
6. a reversal of the trend of stock prices, esp. temporarily, as after a sharp advance or decline in the previous trading sessions

Did you happen to notice how that definition for correction was a lot longer and more detailed than the one for punishment? Did you also notice that while the definition of punishment relates entirely to retribution or harsh treatment — that the various definitions of correction very specifically relate in all but one instance (#4) to transmitting information?

In the four sacred quadrants of operant conditioning punishment is defined as an aversive stimulus, such as introducing a shock or loud noise, resulting in a decrease in an antecedent behavior. Note that in this context, all that is required of punishment is that it be aversive (i.e. unpleasant) and that its presentation should reduce the frequency at which the behavior that occurred before its application.  An operant punishment doesn’t give the punishee any information about what he did wrong or how we might prefer he behave. It just tells him that unpleasant consequences will tend to follow it.

Yet — radical ‘purely positive’ trainers will tell you that punishments and corrections are the same thing. And they will condemn you to a Skinnerian hell for using either.

And this brings us back to Carlsmith, Wilson and Gilbert.  The ‘purely positives’ are right in one way.  Retributional punishment, the kind where you get even, er – umm; apply an aversive stimulus to a two- or four-legged animal after it behaves in a way you don’t like is as unhealthy for you as it is for the target of your anger, oops… the subject.

Whether the purely positives want to admit it or not, outside of behaviorspeak, correction is not the same as punishment.  Correction can be gentle and it should always be fair. It doesn’t just reduce the frequency of behavior by applying an unpleasant consequence, it provides information on why the behavior is not desired and / or what other behaviors to engage in instead. And when it’s done without anger and with even a moderate degree of skill, it doesn’t create fear or stress in the student or the teacher.

Its time to drop the behaviorist’s strict reductionist / operant idea that correction and punishment are identical.  Correction is not punishment.  Correction is rooted in explanation – not in retribution.  Correction should never be done in anger.  And — because a good correction creates an immediate improvement in behavior (however small)  it should always be followed immediately by forgiveness and praise.

dog is rewarded - behavior is reinforced

Playing With a Dog

I stumbled upon a little gem in my computer today.  An insightful article titled “Playing With a Dog” that was published in The Quarterly Review of Biology in 1936.  The article is a vehicle for E.S. Russell to use observations of his fox terrier dog at play to illustrate some principles of animal behavior and cognition.

Russell wrote that before he trained his dog Gina to play with a ball, a ball held no meaning for her.  Balls were not objects she noticed or paid attention to.  But, after he trained her to play fetch, not only did a ball become significant to her but she also began to appear to categorize other objects as being ball-like.  That is, she treated objects that could be used in the same way as a ball as if they were balls (i.e. she picked them up, brought them to him and dropped them.)

Humans and animals divide objects and events into meaningful categories as one of our most basic cognitive functions.  Categories range from very basic ones like edible versus non-edible to abstract human concepts such as poetry versus prose.  Categories are bounded sharply, not transitionally.  A thing either is or is not part of a category.

The ability to categorize is an adaptive trait.  Without it every object and every event would be perceived as unique and it would be impossible for animals to generalize.  Complex behavior is based on elaborate abilities to categorize.    

Categorization is highly context-specific.  Items that on the surface seem to be utterly different (such as Frisbees and squirrels) can be viewed as highly similar if they are placed in a context (“things that are fun to chase”) that highlights an aspect they have in common (chase-ability).  The way we categorize things also depends on our life experiences and the goals we have in mind as we consider them.

The objects that Russell’s dog treated like balls didn’t look alike.  They didn’t have a common size, shape or smell.  The only qualities they shared were that they were of a size and portability such that the dog could easily pick them up and carry them. Their functional value was the basis of the dog’s categorizing them as ‘ball-like’. 

In another game, the dog was taught to bring Russell pennies to earn a bit of cheese as a reward.  Soon after she learned this trick, she began looking for penny-like objects to bring them to him to try to get cheese.  Some of the less preferred ‘ball-like’ toys were small and bright-colored or disc-shaped.  Though the dog showed only a low interest in bringing these toys to Russell when she wanted to play fetch games, she showed a stronger preference for them once she learned they might earn her cheese.  Russell stated that he thought that this indicated that an object may have different values in different contexts. 

Russell discussed the importance of the dog’s Umwelt in perception and categorization.  He noted that humans are so used to perceiving a vast number of discrete, easily discriminated objects in our own environment that we tend to assume that the world appears in a similar highly articulated and abstractly meaningful form to our dogs. In doing this, we forget that the dog’s interests are different and simpler than ours.  The dog attends to and responds only to objects or events that bear a functional importance to it.  Objects and events that only hold an abstract value (such as books, birthdays, paintings, etc.) may hold great meaning to us and be utterly unremarkable or even unknowable to our dogs. 

On the other hand, objects and events that we see as insignificant may hold great meaning to our dogs.  For example an unremarkable (to you) bit of crumpled paper on the ground might mean ‘possible bit of food’ to a dog and the soft, low rumble in the street that you tune out as meaningless may mean ‘delivery truck coming’ to your dog. 

Sadly, many dog owners aren’t aware that not only do our dogs perceive the world in a much different way than we do, but also that their system of values is poles apart from ours.  This leads to a lot of unnecessary misunderstanding and confusion for both species. 

A dog owner can’t understand why her dog would ‘thoughtlessly’ urinate on an expensive Persian rug — and her dog can’t understand why his owner is upset that he peed right in the same place where the cat did.  The rug exists in completely different realms of value and functionality in the woman’s world and in the dog’s. 

We can never inhabit the same perceptual and contextual worlds that our dogs do – but as big-brained humans we can maintain an awareness that that difference exists.  And we can use that awareness to be more patient, creative and mindful in finding ways to bridge the gap when misunderstandings occur. 

Go play with your dog.  Do it with an open mind and an open heart and you just might learn something new.

Pat Smith plays with Fly.  This is a great game and they’re both obviously enjoying it, but I’m quite sure it has very different meanings and values to both of them.

Temple Grandin in Translation

In her best-selling book “Animals in Translation” Dr. Temple Grandin argues that animals and autistic humans share some important cognitive abilities.  Dr. Grandin, who is autistic herself, uses her unique insight into autism to explain animal behavior. Some of her theories include:

• categorizing autism as an intermediate condition between animal and human consciousness;
• attributing hyper-specificity as a basic characteristic of animals (i.e. they are not capable of seeing the forest, only of seeing many trees); and
• arguing that the worst thing you can do to any animal is cause it fear; and comparing animals to autistic savants. 

According to Orli Van Mourik over at www.neurontic.com

‘For those who’ve read up on Autism, Grandin’s ability to “relate” to animals may come as a surprise. Autism is marked by an inability to empathize. Autistics find it next to impossible to grasp the inner workings of someone else’s mind. They lack what psychologists call “a theory of mind.” For a normally functioning person, this is a difficult concept to grasp. Our ability to infer another’s emotions is so instinctive.’ 

Van Mourik goes on to explain that empathy arises from projection, from assuming that other peoples’ (or other animals’) internal experiences and reactions to stimuli will be similar ours.  The ability to project is a skill that most autistics lack.

Despite this, Grandin writes that early in life she sensed that, like her, animals tended to focus on details in the environment and that, also like her, they seemed to understand the world based on sensory experiences instead of narrative. 

Her intuition on how animals, especially prey animals, perceive and understand their world has led her not only to a best-selling book, but also to a successful career in studying livestock behavior, designing stock handling facilities and consulting on the humane handling and slaughter of meat animals.  Her brilliant insights on livestock were highlighted in the 2006 BBC documentary “The Woman Who Thinks Like a Cow.” 

As intriguing (and as popular) as the idea that animals think like autistic savants is, new research by Giorgio Vallortigara et al of the University of Trento, Italy casts doubt on the hypothesis.  Vallortigara is quoted as saying “Autism is a pathological condition.  The extraordinary feats of remembering thousands of caches or sounds shown by some animal species are exhibited by healthy animals.”  The exceptional skills of savants, on the other hand, arise despite (or because of) the loss of other cognitive skills.

The Italian researchers assert that, “the left hemisphere sets up rules based on experience and the right hemisphere avoids rules in order to detect details and unique features that allow it to decide what is familiar and what is novel. This is true for human and nonhuman animals, likely reflecting ancient evolutionary origins of the underlying brain mechanisms.” Vallortigara stated that while Grandin’s book “shows extraordinary insight into both autism and animal welfare,” the question of equivalent cognitive abilities between savants and animals “deserves scrutiny from scientists working in animal cognition and comparative neuroscience.” 

For some time I have questioned Grandin’s assertion that fear is the primary emotion experienced by animals. In her book “Animals in Translation” Grandin wrote that if animals, like autistic humans, are prone to sensory overload, and their emotions would also likely be largely governed by fear.  She goes on to relate the propensity for sensory overload (and therefore also to be strongly affected by fear) to improper function of the frontal lobes.  The article published by Vallortigara et al may cast doubt on this theory, as they based much of their findings on the fact that animals have healthy frontal lobes.

Temple Grandin’s brilliant insights on livestock handling have not only made her surprisingly successful, more importantly, they have also lessened the suffering of the animals that go into our meals every day.  While a few of her theories may not, ultimately, stand up to academic scrutiny, she undeniably has a unique voice and a perspective in understanding animals (and autistic humans) and her contributions to the fields of humane livestock handling and animal cognition should not be belittled.In a future post I hope to go into detail on why I think that her book “The Unwritten Rules of Social Relationships” may provide deeper insights on how to understand and live with animals that the bestselling “Animals in Translation.”