Essay On My Pet Animal Fishers
In an earlier test, Sarah had been shown a real apple (as the sample) and four pairs of alternatives:
a red plaque vs a green plaque
a square with a stem vs a square
a circle vs a square
a square with a stem vs a circle (see Figure 2).
Sarah consistently chose the alternatives indicated in the figure: the red plaque, the circle, the square with a stem. In the choice between the square with stem vs the circle, she chose the circle on 70% of trials, and on 30%, the square with a stem.
The apple was then removed and replaced by a small, blue, plastic triangle, Sarah's word “APPLE”. The test was repeated. Sarah assigned the same features to her word “APPLE” (the blue plastic triangle) as she had assigned to the round red fruit (the apple) --including the 70/30 choices of circle vs square with a stem.
Suppose we turn the question around and ask what the animal knows about an object's “name”. The apple was used as the sample once again, but now the alternatives were relevant, NOT to the features of the apple, but to the features of the “name-of” apple, the small blue plastic triangle.
Sarah was then shown 15 pairs of alternatives consisting of both objects and words. The first five consisted of wood blocks cut in various shapes, the second ten of words. The wood blocks included these shapes: square vs triangle; triangle vs circle; large triangle vs small triangle; and small triangle vs small circle. The second ten alternatives included such “words” as: “big” vs “small”, “square” vs “triangle”, “round” vs “triangle”, “green” vs “blue”, etc. Sarah made one error in 15 trials, choosing “big” over “small”. In other words, in the presence of the apple, Sarah selected those features that described her word “apple”, even as earlier she had done the reverse, i.e. in the presence of her word “apple” selected the features that described the apple.
Displacement: The Information Retrieval of Words
Language has been glorified on grounds of its displacement which enables us to discuss “things that are not there”. To talk about, say, tigers and elephants from the safety of our living-room. But how do plastic words work for the chimpanzee? Do its “words” too enable the chimpanzee to retrieve stored representations, and “talk about things that are not there?”
Two issues need to be resolved in order to answer this question. We must know, first of all, if the chimpanzee's mental representation system is an effective one. Does it do justice to the animal's perception and retain what the animal perceives? Secondly, how much of the retained material can the animal retrieve with its plastic words? To answer the first question, we divided seven different fruit into a number of components and features and then required the animal to match one feature or component of the fruit with the other.
For instance, the animal received a taste of a peach, and had then to choose between the white outlines of a peach and an apple; between a yellow and a green plaque; a peach pit and a grape seed, etc. The logic of the test can be viewed as a kind of game in which photographs of family members are cut into identifiable pieces, and the purpose of the game is to match grandmother's ear to her mouth, eye, or etc.
In the animal version of the test, all four of the chimpanzees performed well above the chance level. While three of the animals were less accurate in the use of some features than with others, Sarah matched every feature and component with every other. The tests were a clear demonstration that all four animals had well-formed mental representations of the large variety of fruit contained in their daily diet. Would the plastic words be equally effective in retrieving the animals' representations?
This question was tested by substituting plastic words for the features of the fruit, and requiring the animal to match, say, the taste of peach with the words “peach”, “apple”, etc. The four animals had been about 83% correct (on average) when matching features to features - they made still fewer errors when the alternatives were the names of the fruit! The animals were as effective in matching parts of the fruit to the “names” as they had been in matching the parts to the whole fruit! In short, for the chimpanzees, the “names” of the fruit were essentially as informative as were the actual fruit.
Can the chimpanzee use its words, not only to retrieve information, but to solve problems, as children can? When given containers holding concealed fruit that can be identified by either their place or features, rats and apes use place. So do young children. But when children are three years old, they shift to features, presumably because they can now use language to identify features. Requiring children to count backwards when doing this problem confirms the hypothesis. Children revert to the use of place, for counting backwards interferes with the use of language. This is the question: while an untrained chimpanzee behaves like a rat and uses place, would a language-trained chimpanzee behave like a three-year old child, and shift to the use of features?
The Shift Effect
Young children tend to sort objects on a thematic basis, for example, they pair a picture of a dog with a picture of a bone; older children however sort on a taxonomic basis, placing a picture of a dog with a picture of a cat. Curiously, however, young children will switch from thematic to taxonomic sorting as soon as a nonsense word is introduced : “….if shown a dog, and told it is called a “dax”, and then told to find another “dax”, …young children now tend to put a dog and cat together…” This shift is made, according to Markman and Hutchinson (1984), because for children, words are the names of categories, not of individual objects.
This claim applies to chimpanzees as well (Premack, 1990). Chimpanzees show a developmental change in the sorting of objects - young chimpanzees sort on a thematic basis, while older ones sort on a taxonomic basis. And when a nonsense word is introduced, the young animal shifts to the taxonomic alternative. It is notable that the animal makes this switch only under these conditions: the object added to the sample is a “nonsense” word, that is, a piece of plastic that, while never used before as a word, nevertheless conforms to the properties of the word class. The use of such foreign objects as bolts, dominoes, etc.—does not produce a shift effect.
Dogs and “Words”
The first reported paper on the teaching of words to dogs appeared in 1884 in Nature, written by Sir John Lubbock. Because Lubbock wrote of “conversing with animals” it seemed at first that he had done the obvious: taught his dog the names for a few objects, for a few actions, and then combined them to produce dog “sentences”. A sentence in the dog's language then could have been any object-action combination, such as “fetch bowl,” “sniff slipper”, “fetch slipper”, etc. In fact, if Van had been taught just four object names—ball, slipper, bowl, stick—and the names of just four actions—fetch, sniff, bite, push—Van could have carried out sixteen commands! Moreover, the dog need not have been trained in all sixteen of the commands. He might have been trained in just four and have transferred to sixteen. For, in principle, Van could have learned the rule: Sentence = action + object.
And having done that, Van should, in addition, have been able to apply the same rule to any NEW names of objects and actions added to his lexicon. But Lubbock's “conversing with animals” was merely a high-spirited claim. He simply taught his dog the names of a few objects and a few actions, but never combined them! But can a dog learn the rule: action + object = sentence, and use the rule productively?
There are two main alternatives: A species learns category-level associations; or it learns item-specific associations. Species that learn item-specific relations will not perform productively. Although they may ultimately learn many commands—fetch slipper, push box, bite ball, etc., - each command will be learned individually. Only species that learn category level associations will perform productively.
Children and chimpanzees, we know from various sources, use rules productively; moreover, as the shift effect suggests, both learn object names as names of categories. Probably only species that show a shift effect – species for which, therefore, words are the names of categories - form category-level associations, and use rules productively.
Although both child and chimpanzee ultimately acquire new words in a single trial, this is not how chimpanzees learn their first words. How quickly children acquire their first words is not entirely clear, but it is not until they are about three years old that they begin to add new words at a high rate (Golinkoff & Gordon, 1983). The naïve chimpanzee is extremely slow to acquire its first words, though ultimately, like the child, it too acquires new words in a single trial.
When being taught its first words, the chimpanzee was required to first place the name of the fruit on the writing board in order to obtain the fruit. The plastic names were backed with metal, and adhered to a magnetized writing board. Initially, the animal fails to use the correct name for the correct fruit, being as likely to use one name as another. What impedes the animal's learning at this early stage? Tests suggest there is interference from the learning of category-level information.
Indeed, the first rule that the chimpanzee uses reflects category-level learning, that is: any fruit can be obtained with any word. The rule assumes that the animals have divided the situation into two classes, words and referents. Words consisted of plastic names, and all referents were fruit in early training, and the animals associated an operator with each class. They were to: place words on the board, but place referents in their mouth . In a still earlier period, they occasionally squashed a piece of fruit on the board before eating it.
Rather than learn item specific associations such as: a blue triangular piece of plastic “goes with” apple, can be used to obtain apple, is the “name of” apple, or a red rectangular piece “goes with” banana, etc.—the animal learns a category–level association: that ANY piece of plastic can be used to obtain ANY fruit.
Tests confirm that, at a time when the chimpanzees had yet to learn a single word—had failed to learn a single item-item relation—they had nonetheless learned abundant category-level information. For the animal could distinguish the following:
1. a word (an object that conformed to the class of objects used as words) from a non-word.
2. a word that had been used from one that had not been used.
3. a potential referent from a non-referent (for instance, they choose fruit rather than candy in the language-training context, though they preferred candy, because fruit not candy was associated with words).
4. a referent that had been used from one that had not been used.
Children, because they have extensive experience with speech sounds, do not have a problem comparable to that of the chimpanzee. Long before children learn words, they experience the sounds of speech (even in the womb). They know speech, not only as the category to which words belong, but as the medium in which people express their social feelings toward one another. But suppose children were given, not speech, but the sounds of a violin as words, would they react to the violin sounds as chimpanzees react to plastic words? Would they learn the category properties of violin-sounds before learning individual words?
It is of some interest to note that tests with global aphasics show that adults continue to retain much of the kind of information which the chimpanzees acquire initially (Velettri-Glass, Gazzaniga, & Premack, 1973). Even though the stroke victims had completely lost both the comprehension and production of their language, they could nonetheless distinguish: a potential or nonsense word from an actual word, a familiar from an unfamiliar word (high probability of use from low probability of use). This similarity between what the chimpanzees learn and what the global aphasics retain supports the view that forgetting retraces the course of learning - that the last things forgotten are the first things learned.
Initially, the chimpanzee uses plastic names, words, to request objects, subsequently it uses words as elements in “sentences”. In some cases the sentences take the form of questions, in other cases they are descriptions of simple events. The transition from words to sentences is smooth, the simple use flows easily into the more complex. In both cases, the word is at the center of a social exchange. In the simple case, words are exchanged for objects, in the complex case, words are exchanged for information. Questions serve to extract information from the listener, while descriptions impart information to the listener.
A recent article on the dog's learning of words falsely links one-trial learning (“fast mapping”) to words, children and dogs (Kaminski, Call, & Fischer, 2004). One-trial learning is not - in fact - distinctive of either a topic or a species. Moreover, although chimpanzees, and presumably children, do not learn their first or early words in one-trial, they learn subsequent words in a single trial. From the time language use becomes autonomous in the two- to three-year old child, word acquisition continues at a formidable rate, culminating in the 60,000 word lexicon of the college graduate. Since language use never becomes autonomous in the chimpanzee, it must continually be taught new words. But the chimpanzee learns them in one trial—either with the instruction “X is the name of Y”, or even more directly using a method devised by a tired trainer, weary of formal procedures.
Taking advantage of two containers, one filled with unnamed objects, the other with pieces of plastic not-yet-used-as-words, she took one item from each container, clapped them together calling the animal's attention to the pair, and henceforth the piece of plastic became the name of the object (Premack, 1976, p 166).
Crediting the dog with the use of “exclusion learning” to infer the referent of a new word is again dubious, since the test combined an unfamiliar (unnamed) object with an unfamiliar (unused) word (Kaminski, Call, & Fisher, 2004) . This procedure conflates unnamed with unfamiliar. Which of the two conditions is the dog using? To prevent the dog from merely matching unfamiliar object with unfamiliar name, one must combine either familiar (but unnamed) object with unfamiliar word, and/or vice versa. Nonetheless, the dog might still pass this test without making any use of inference whatsoever. While matching an unnamed word with an unused word is more complex than matching two instances of familiar/unfamiliar, it is not equivalent to inference.
If an individual, from the very start, learns words in one trial, there will probably not be a shift effect. The quick-learning of words—of item-item relations—suggests little interference from category-level information. It indicates that words are labels for objects, not names of categories. If so, the combining of action-object words would not lead to productive rules. Category learning seems a nuisance, an imposition on the all important item-level learning, until it is recognized that productive rules depend on category-level learning.
There may be a correspondence between category-level vs item-specific learning on the one hand, and the learning of relations vs the learning of absolute stimulus values on the other. While all vertebrates must certainly learn both relations and absolute stimulus values, species vary in the amount they invest in the two levels. Chimpanzees, for example, make a large investment in the relational level, larger than in the absolute stimulus level; whereas rats do the opposite. How large an investment a species makes in the two levels can be determined by putting the two levels into competition. For instance, rats taught to jump left when the top stimulus is brighter than the bottom-- say, intensity 10/6—will not jump right (as they should) when given the inverse, 6/10. The stimulus values 6 and 10 are both associated with jumping left, and absolute stimulus value takes priority over relation in the rat, therefore they jump left (Lawrence & DeRivera. 1954). This is not a mistake chimpanzees would make. Humans differ further in that their investment in the two levels is flexible: they can focus on relation in some problems, individual stimuli in others. The investment made by other species is frozen, i.e., is not subject to variation from problem to problem (Premack & Premack. 2002). Probably the two parameters are related. Species strong in category-level learning invest heavily in the learning of relations; those weak in category- level learning emphasize learning of individual stimuli.
Can a dog learn to name the concept “name-of”, and use it to learn new words? Can it do features analyses on words and referents, recovering the features of each when in the presence of the other? Can it use words to retrieve the mental representations of their referents, and thus carry out displacement? Will it show a shift effect? Can it form productive rules?
If the dog can do these things, then it is probably not a dog at all; but a chimpanzee in“dog's clothing” (see Figure 3).
"So, what's in it for the dog?"
This essay will be short and to the point. Not surprisingly, a piece by Roni Caryn Rabin in today's New York Times called "Need a Date? Get a Dog," resulted in my email inbox exploding with messages, all of which -- without a single exception -- asking something like, "So, what's in it for the dog?" There's nothing in this essay that speaks to the well-being of the dogs themselves, and there are many problems with dog-human relationships, some of which should at least have been mentioned.
There are many studies of dogs being social catalysts or social facilitators who act like "social glue" in bringing people together. For example, University of Pennsylvania professor and dog expert James Serpell sees dogs as mediators in three principle areas, namely, as social lubricants (catalysts of social relationships between people), social ambassadors (a moral link with other animals and nature in general), and the animal within (a sort of unconscious connection with other animals and nature).
Ms. Rabin's essay is available online, so you can read the nitty-gritty details of her human-centric review of a number of different studies and quotations from others that extol the benefits, mostly to men, of using a dog to meet women. I fully realize that her essay is not a scientific review, but I would have expected even just a few words from her or others about what the dogs get out of it, or don't. That's why I received so many emails that still are coming in.
Choosing to bring a dog into your life has to be good for you and the dog
Ms. Rabin begins: "Something happens whenever Aaron Morrill takes his large and fluffy mutt, Donut, for her daily walk, and it’s something that always catches him a bit by surprise. Women gather." She goes on to use a number of quotes by Dr. Helen Fisher, a chief advisor to Match.com and senior research fellow at the Kinsey Institute. She quotes Dr. Fisher as saying, “Having a dog really says something about you ... It says you can care for a creature, that you can follow a schedule and get home to feed it, that you can walk it and love it and spend time with it.” This superficial generalization that ignores available data about what life is really like for far too many dogs is way too fast for me and others who actually study dog-human relationships.
Ms. Rabin also summarizes the results of a number of different studies that speak to ways in which "the messages sent by dog ownership influences others’ judgment and behavior." These include a study in which volunteers "ranked someone as happier, safer and more relaxed when they appeared with a dog" and another series of experiments in which men had more success in getting a woman's telephone number when they were accompanied by a dog, and both men and women were more successful in panhandling when accompanied by a dog. But, once again, what's in it for the dog? Do they also get to meet new dog friends?
Is dog ownership "a real honest message" about who someone really is?
After a brief review of some studies that show how dogs have positive effects on meeting someone and on positively influencing relationships, Ms. Rabin writes, "But ultimately, Dr. Fisher comes down on the side of pet owners, who must devote a lot of time to their animals. In a world full of messages, not all of them necessarily honest, she said, dog ownership is generally 'a real honest message.'”
It's more than reasonable to ask if dog ownership really is "a real honest message” about who one truly is and what sort of life they will offer their canine companion? And, indeed, far too often it's not. Facts actually show that far too many dogs want and need much more than they usually get from us. And, facts also show that rescued and adopted dogs don't do as well as many people claim they do when they find a home (for more discussion please see "Dogs, Cats and Scapegoats: Messes We Make With Companions"). Indeed, for far too many companion animals, it's really not the "forever home" they sorely want and need.
It might surprise people to learn that numerous companion dogs who are fortunate enough to share their life with a human are highly stressed, but when you think about it, they're always trying to adapt to a human-oriented/dominated world in which their wants and needs are secondary to those of their own and other humans.
Psychology Today writer Dr. Jessica Pierce provides an extensive discussion about this in her excellent book called Run, Spot, Run: The Ethics of Keeping Pets. And in her book, Love Is All You Need, Jennifer Arnold notes that dogs live in an environment that “makes it impossible for them to alleviate their own stress and anxiety.” (p. 4) According to Arnold, “In modern society, there is no way for our dogs to keep themselves safe, and thus we are unable to afford them the freedom to meet their own needs. Instead, they must depend on our benevolence for survival.”
Think about it: We teach dogs that they can’t pee or poop wherever they want. To eliminate, they must get our attention and ask for permission to go outside the house. When we go outside, we often restrain dogs with a leash or fence them within yards or parks. Dogs eat what and when we feed them, and they are scolded if they eat what or when we say they shouldn’t. Dogs play with the toys we give them, and they get in trouble for turning our shoes and furniture into toys. Most of the time, our schedule and relationships determine who dogs play with and who their friends will be. All in all, many people are helicopter guardians for the dogs with whom they choose to share their lives, and they say or command "No" or "Don't do that" far more than they give praise (for more discussion please see "For Dogs, Helicopter Humans Don't Balance Scolds and Praise").
So, are dogs good for meeting people and is choosing to live with a dog a real honest message about who someone truly is? Ample data show dogs are pretty good at being social lubricants, and a good deal of data also show that even people with good intentions far too often don't give their dogs what they want and need.
When we choose to live with other animals it's essential to pay close attention to what they need and want
When we're considering how dogs can be used to serve us, let's not be selfishly anthropocentric and let's also seriously consider what's in it for the dogs. Dogs are highly emotional and sentient beings, and the shared emotions between dogs and humans can function as "social glue" to bring us together. But, it's got to be a two-way street in which both the well-being of the dogs and the humans are given careful consideration.
Beware of false advertising. When a relationship between a dog and a human is a two-way affair, it's a win-win for all. However, far too often it's not. So, let's not use dogs as bargaining chips or billboards to say something about ourselves, perhaps peppered with a touch of false advertising, without considering what they get out of it too.