Dog experiment shows that dogs don't have shape bias when learning
This exercise has profound implications for training methods of all dogs, though much more research will undoubtedly be taken to confirm and expand on findings.
THE core question of the experiment was to investigate the "key feature of human word comprehension in a five year old Border Collie: the generalization of a word referring to an object to other objects of the same shape, also known as shape bias."
" Two experiments showed that when briefly familiarized with word-object mappings, the dog did not generalize object names to object shape but to object size.": That is, that all objects about the same size of a ball, that the researchers told the dog was called a ball, the dog would assume to be a ball - such as if the dog was told to fetch a ball and a can, a ball, a cube might be considered to be a BALL by the dog. Assuming that human interaction (pointing) and smell of the objects are the same.
To complicate matters " A fourth experiment showed that when familiarized with a word-object mapping for a longer period of time the dog tended to generalize the word to objects with the same texture."
This naturaly let the reserachers to conclude that the dog tested did NOT display human-like word comprehension. It could not easily define the difference between one word and objects of similar size, and texture. Whereas humans can tell that a ball is a ball, regardless of size, or texture.
It is true that some dogs can learn in excess of 200 words and what they mean, but those words are often command driven, and not trying to define the dogs ability to discern between similar size/ colour/ texture objects. The fact that humans can do this, suggests anoother evolutionary path in the develppent of their brains and how they can categories objects, something unique that has allowed the evolution of language and all of the cooporative feats that have flowed since then.
How humans learn object names
Landau, Smith and Jones (LSJ) previouslyshowed that when 3 year old children or adults learnt to connect a new object name with a new object, they generalized the meaning of the new name to objects that are similar in shape, but not to objects that are similar in size or texture.
For instance, children and adults were told that a U-shaped solid object (previously unseen shape) was called a "dax" (a previously unknown name). When they were asked to select a DAX object among a number of objects, ... it was found that many children and adults choose objects that only differed in size or texture but not in shape. That is they associated the name of an object with its shape, but not with its size or texture.
How dogs learn
Previously it was found that "a domestic dog is able to use object shape but also object size to identify objects " However this 2012 experiement goes on further and " investigated which object features a dog will spontaneously select for named objects: shape – as in humans – or object size or texture.
The fact that the subjects was a five year old boder collie (named Gable) should be no surprise (one of the smartests dogs around. But this dog also had a history of word learning, so it was pre-primed so to speak.
Before the experiment it was claimed by the owner, and tested by the researchers that the dog knew 54 different words referring to 54 different objects. These results show that Gable reliably knew the words for 43 objects.
Testing was also down for one experiment over a ten minute learning period, and another over a 39 day learning period. Would length of study improve his abilities or was the type of associate ingrained.
bias in word generalization after brief familiarization (ten minutes learning)
The experiments aim was to replicate LSJ's standard experimental paradigm. That is "LSJ found that word generalization is based on object shape, but not on object texture or object size in both young children and adults."
In this experiment a new word (for example, dax) was used to name a new objec,t after which Gable was presented with pairs of objects and asked to select the DAX object.
The texture variations used on the NEW SHAPE objects were based on three different cloth textures covering foam cut-outs. Gable was familiarised with the DAX object for a ten minute interval before testing,
" In order to avoid scent cues every object was covered with the same two layers of cloth applied with different textures outermost, so that every object was always covered by the same kinds of materials, and the objects and the materials they were made of were always stored together."
"During the familiarization phase the owner paired the word dax with the DAX object by asking Gable to fetch the DAX object by using the get dax command several times, after which the DAX was put into several sets of objects for which Gable knew the names, and asking him to fetch the DAX"
"Gable linked the word dax to DAX-sized objects in ten out of ten cases in which he was given the choice between a DAX-sized object and a larger object, thus confirming that Gable generalized the meaning of the word dax to other objects that were of the same size as the DAX object, irrespective of their shape or texture.
Experiment 3: investigating dogs size bias after 10 minute familiarization
The experiment aim was to determine whether the size bias found in experiment 2 (able to select similar sized objects all considerd a DAX), was based on "Gable preferring the smallest object (for example, because smaller objects are easier to pick up) or whether the size bias was based on a preference for generalizing word meaning to objects of the same size as the standard object."
In this experiment Gable learnt to link the word gnark with standard object.
Gable demonstrated a word generalization bias for objects with the same size as the GNARK but not for objects with a GNARK shape.
" This experiment results confirmed a size bias in word generalization for the dog after a brief new word-new object familiarization period, as well as the absence of a shape bias as found in humans."
Experiment 4: bias in word generalization after extended familiarization (39 days)
This was a repeat of the DAX experiment. " Gable took the DAX home for 39 days, and was taught to link the word dax with the DAX object on a daily basis in different contexts afforded by the home environment and by inserting the DAX into many different sets of toys, and asking Gable to get the dax."
" When Gable was familiarized with the dax word-DAX object combination for 39 days he selected the DAX in 6 out of 6 trials from among 9 or 10 of his other toys"
HOWEVER, Gable tended to generalize his word knowledge to objects of the same texture but not to objects of the same size or shape
Gable did not show that he could reliably distinguish between a DAX and other artificial objects that only differed in size, texture or shape.
These results show that after a long dax word-DAX object familiarization period, word generalization was qualitatively different for Gable compared to a short familiarization period, but also compared to word generalization in humans.
While the results seemed to show a marked difference between how adult humans learn shape names, it may also have been tainted by previous word training for objects over the past three years.
It was shown that Gable's reliance on object size was replaced by a tendency to rely on object texture when he was familiar with a word-object mapping for a longer period of time. That is when confronted by a tricky choice (objects appearing as too similar in shape to discern between) he would then select an object based on a texture similar to the texture of the original object that he learned the word for.
So the main learning or future study question from these experiments are "Why do object names refer to object shape for humans and to object size or even texture for Gable, or any other dog?"
It is noted that humans primarily rely on vision for object identification while dogs use scent first. Thus our cognitive system rely mostly on shape for object naming.
It is also possible that in the absence of clear scent cues (objects were made to smell as similar to each other as possible) it is possible that object size and object texture were important for Gable because these cues were available to him from manipulating the objects with his mouth (bite size and felt texture),
Article by Bruce Dwyer. If you wish to use any of this information please refer to the article as a reference and provide a link to http://www.dogwalkersmelbourne.com.au
If you are interested on being kept up to date on dog article releases or healthy dog treat specials, Join us on http://www.facebook.com/HealthyDogTreatsShop
Word Generalization by a Dog (Canis familiaris): Is Shape Important? open access journal PLOS ONE by Emile van der Zee and colleagues from the University of Lincoln, UK.