Very smart dogs can learn to comprehend words, but a new study finds that the way they learn is substantially different from the way humans learn to connect words and objects.

In other words, if a dog manages to learn that the thing he should not chew on is called "shoe," he may be able to apply that knowledge to a shoe-size chew toy, but not, say, a pair of boots.


Scientists have recently confirmed that dogs can learn to understand human language, to some degree. One border collie named Rico, who can supposedly understand 200 words and fetch objects when asked to get something specific, has a memory and learning process comparable to a three-year-old child. Another border collie, Chaser, was reported to know more than 1,000 words.


But there is still much mystery surrounding a dog's comprehension. A trio of researchers from the University of Lincoln in the U.K. decided to investigate one aspect of a dog's language ability: the ability to link a word for an object to the shape of the object. Their work was published recently in the journal PloS ONE.


It comes down to a question of priorities. Humans tend to be shape-biased in our language learning; when we learn a new word for a novel object, we tend to associate that word with things that are the same shape, but not things that are similar in size or texture to that object. When another group of researchers taught children and adults that the word for a certain solid U-shaped object was "dax," and asked them to pick out other "daxes" from various objects, they picked similarly-shaped objects, but not objects of similar texture or size.


(You can even see traces of shape-bias in the previous paragraph; would reading "solid shiny object" be as clear to you as seeing "U-shaped object"?)


In the current study, the researchers worked with a five-year-old border collie named Gable, who was known to be relatively sharp -- his owner claimed that the dog knew at least 54 different words.


After the researchers had confirmed Gable's word knowledge by testing his ability to fetch toys by name, they re-enacted the "dax" experiment with the dog, using a 3-inch wide U-shaped object made of wood and covered in furry cloth.


Once Gable learned to associate "dax" with that particular object, the scientists started mixing in different objects: differently shaped pieces, ones covered with smooth cloth instead of fuzzy cloth, and objects nearly identical to the original "dax" object but differently sized.


To make sure that Gable wasn't using his sense of smell to discern the difference between objects, they let him play with all the items before testing, thus getting his saliva over all the test objects. The objects were also stored together and handled by a single experimenter.


When asked to fetch "dax" from pairs of various possible objects, Gable picked up an item of similar size to the original U-shaped object any time he was given a choice between an object of similar size and one that was larger than the original. When the dog was asked to choose between objects with the same texture as the original object and ones with different texture, he did not seem to have a particular preference for one or the other.


To investigate Gable's size bias further, the researchers did another experiment, this time with an L-shaped object that the dog was taught to associate with the word "gnark." They then offered alternative objects of varying sizes and shapes but kept the texture the same. When then asked to fetch "gnark," Gable preferred to bring back objects that were around the same size as the original.


“We have shown that word generalization and the developmental path for acquiring words is qualitatively different in Gable compared to humans,” the scientists wrote.


The larger question, then, is why human and dog comprehension differs. One possible explanation could be that while humans tend to rely on vision, dogs have enhanced hearing and smell that makes the shapes of things less important to a dog.


“The evolutionary history of our sensory systems -- with vision taking priority over other sensory systems -- seems to have primed humans to take into account visual object shape in object naming tasks,” the scientists wrote.


SOURCE: Van der Zee et al. “Word Generalization by a Dog (Canis familiaris): Is Shape Important?” PloS ONE 7(11): e49382.