Doggie concepts defended

« previous post | next post »

[Marc A. Pelletier wrote to me after reading this post about canine concepts (or the lack of them). He offered a somewhat more pro-canine perspective. What he says is quite reasonable (not that I necessarily agree with all or any of it), and it may mollify a few dog lovers in the Language Log readership who continue to hate me if I present what what he said as a Guest Post. So I herewith do that. And you can comment on it if you wish. —GKP]

Guest post by Marc A. Pelletier

I am wondering why Geoff Pullum seems so insistent that dogs are unable to attach semantic meaning to words uttered by humans beyond the level of conditioned reflexes. Ethology has, in my opinion, contracted the disease of "reverse anthropomorphismitis": the desperate compulsion to avoid ascribing common cognitive mechanisms to animals other than Homo sapiens sapiens, even when doing so requires contriving many additional assumptions and evoking ad hoc hypotheses — I'm surprised that linguists feel the need to do the same (or at least, one linguist does).

Allow me to illustrate my position with an anecdote.

I have a young (three year old) Labrador, with which I have a number of games — two of which are interesting for this discussion. One is the arch typical game of "fetch", where I throw one of several allowable objects some distance away and Vic runs to get it, then returns it to me. The other is sort of a treasure hunt, where I hide a specific chew toy of his which he then searches for and retrieves.

I initiate the first of those games by telling him to fetch his "baballe". He then rushes to find the closest object of the classes of allowable fetching objects (two are actual balls, one is a Frisbee) to bring it to me. On at least two occasions, he returned with objects that are not the usual objects but that obviously fit the correct "class" of objects: a discarded plastic plate that was definitely "Frisbee-like", and a ball forgotten by a visiting nephew that was much larger than his.

The second of those games is initiated by asking where his "nonos" is. This sets him on an immediate treasure hunt where he starts looking in common hiding places, and sniffing around, to find his chew toy (which happens to be and oft-replaced rawhide bone, hence his reliance on smell to help locate it).

I did an experiment by placing all his toys in a large box along with a number of non-toy objects. In all cases, requesting that he gets a "baballe" or a "nonos" will have him return with, respectively, one of the allowable "fetch" toys or his current rawhide bone. Being unable to find a suitable object on request will have him run around in circles trying to locate one. Additionally, any other person can make the same requests with the same results (Vic is very sociable, and will readily play even with a complete stranger as long as one of the members of the household is present and not in distress).

As far as I can tell, there are only two explanations for this behavior. Either he associates those two words with specific classes of objects (and only objects of those classes), or he associates the words with the specific play activity and simply knows which object class is required for the game — in both cases there is clear semantic meaning associated with the word, and neither can reasonably be explained by "reflex" alone: the response is far from automatic and requires active searching as well as minimal classification (Whether the classification is "Is object in class X?" or "Is object suitable for activity X" is immaterial).

So, when Geoff Pullum says that "The concept of being lower (closer to the ground in some appropriate sense) than one's present position was not there in the doggie brain at all", I think he is correct, but only because the statement is vacuously true: the semantics associated with "down" in the humans' brain was (obviously, from what happened) not the same as that in the dog's. But that does not say anything about the capability of the dog for associating semantics with words in general, which I believe was his point in the first place.

Marc A. Pelletier


  1. drjon said,

    July 30, 2008 @ 7:52 am

    Seems to me that the Schnauzer should have been taught the command "off" (in a sense which would have done what the owner wanted) in addition to the command "down" (which the schnauzer performed admirably, as trained to).

  2. Mark P said,

    July 30, 2008 @ 8:04 am

    It's clear to me that dogs as well as most other organisms communicate. It's not necessarily clear to me that their communication fits the definition of language used here. On the other hand, the "down" anecdote from the previous post isn't really relevant to that question. The "down" command obviously meant something very specific to the dog, and it should have come as no surprise to the owner when the dog obeyed the command exactly as it had been taught and learned. What the anecdote demonstrated was not that dogs don't understand language like humans but that the owner didn't understand dogs. Another explanation is that dogs have a peculiar sense of humor.

  3. Tobia said,

    July 30, 2008 @ 8:32 am

    Sorry, couldn't help think of the Gary Larson's cartoon:

  4. Tobia said,

    July 30, 2008 @ 8:32 am


  5. Andy Hollandbeck said,

    July 30, 2008 @ 9:35 am

    My great grandfather had a standard poodle named Charlie. I have heard from my mother and both of her sisters that, when they went for a ride in the car, my GGrandfather could say, "Charlie, there's a squirrel on the left side of the car," and Charlie would go to the left window. Same for the right.

    Whether the command was accompanied by any sort of gesture, intentional or not, I don't know. (He died when I was 8.) But my mom and her sisters swore that Charlie could respond to simple spoken comments — as opposed to commands.

    I do think the distinction between teaching a command and teaching the meaning of a phrase is gray.

  6. Rob Gunningham said,

    July 30, 2008 @ 9:37 am

    You may not be surprised when I say that this is probably the most significant post that Language Log has so far, well, posted (and if you are surprised then you probably don't know that I am in fact the so-called "Jeremy Hawker" of Geoff's earlier post). The reason it's significant is that there is clearly semantic meaning in one's communication with other animals — I mean, it's so bloody obvious: you only have to come into daily contact with any animal with a brain larger than a mosquito — but as a simple non-scientist I have been unable to explain it like the great Marc Pelletier.

    And now all the linguists are going to read the words dog and language in the same sentence, and start salivating like Pavlov's…subjects, and they're going to tell you that you're the one who's barking, Marc. Well, don't worry. You are in fact on the way to a Nobel prize, big feller — I know, I live in Norway. Nice work.

    By the way, why is it odd that a dog would roll in a pizza? Dogs love pizza, and if they roll in it they get more. Seems pretty sensible to me, but then I'm not a linguist.

  7. Jorge said,

    July 30, 2008 @ 9:42 am

    This might be more about humans than about dogs, but I think it's still language related. when Geoffrey wrote:

    "(And the pizza, in case you were wondering about this, for it worried me,"

    I was indeed worrying about something about the pizza at this point, given all the preceding genito-talk. But the rest threw me off a bit:

    "was not any longer regarded as humanly consumable. I don't know what they had for dinner instead that night.)"

    What I was worrying about instead was how hot the pizza was. I guess this shows I was not empathizing with the same being as Geoffrey. :)

  8. Marc A. Pelletier said,

    July 30, 2008 @ 10:28 am

    Despite the glowing praise (Thanks Rob), I think it's important that not too much is read into my anecdote either.

    I don't argue that canines are able to formulate anything resembling syntax, and I very much doubt they can understand anything resembling our complicated linguistic communication.

    What I do believe, however, is that they can associate meaning to words (symbolic stand-ins for what they represent), which is arguably the first step up the linguistic ladder; and that attempts to automatically reduce that to conditioned behavior is needless (and harmful) oversimplification.

  9. TootsNYC said,

    July 30, 2008 @ 10:40 am

    I still say, one of the reasons Mr. Pullam closed the comments on the other thread was that he didn't want to hear all the cute "dog on the table" stories.

    I'm w/ Mr. Pelletier on his last comment.

    @Jorge: I assumed the pizza was delivered, which means it certainly wasn't hot enough to burn anybody w/ fur. In fact, I'm not sure it would have been hot enough to be humanly consumable–or humanely consumable?

  10. Peter said,

    July 30, 2008 @ 10:58 am

    Reading these exchanges I conclude that this is just another example of the modus operandi of western science: when there's a natural phenomenon that can't be explained by current scientific theories, start by denying the phenomenon exists.

  11. Jonny Rain said,

    July 30, 2008 @ 11:02 am

    "What I do believe, however, is that they can associate meaning to words (symbolic stand-ins for what they represent), which is arguably the first step up the linguistic ladder."

    What's the difference between a complex neural network of perceptual-motor associations and a symbolic representation?

  12. John Cowan said,

    July 30, 2008 @ 11:49 am

    I think it's plain from these dog stories that we are still in the domain of "hear a sound, do a behavior". The behaviors associated with "baballe" and "nono" in one story, and "left" and "right" in the other are obviously complex ones, which is reasonable: dogs are smart. But there is no need to talk of meaning here, a concept we introduce when it's unmistakable that we don't have any such direct coupling between sounds (or other observables) and behavior. If I point at a window and say "Please open that sklarknick", you can probably do what I want even if you don't know the meaning of the neologism sklarknick. But if I ask you out of the blue "What's the typical height and width of an American sklarknick?", you can't answer unless you know (among other things) what the word means.

  13. Alan Gunn said,

    July 30, 2008 @ 12:35 pm

    It's certainly true that dogs have a way of seeming to understand more than they actually do. My springer spaniel "learned" to open the gate to the back yard with his nose–looked just like Lassie, breaking out to go get help for Timmy. But, after a winter inside, he tried again by standing a foot away from the gate and repeatedly lifting his nose in the air. Typical "I did this and got good stuff so I'll do it again" dog "learning." However, we (i.e. the humans) had to learn not to use the word "walk" in his presence, lest he think he was going to get an outing. I don't see how to interpret that except by saying that he associated the word "walk" with his favorite activity. Seems like a sort of understanding of that one word.

  14. Timothy M said,

    July 30, 2008 @ 12:41 pm

    It's clear that animals have the ability to associate a "stimulus" with "something." Pavlov's dogs associated the ringing of a bell (stimulus) with food. Many dogs associate the sound of the word "down" with lying down and receiving a reward (whether that reward be food or just the happiness of their master). Marc Pelletier's dog associates "baballe" with a game using a certain class of objects. My dog associates me picking up his leash with us going for a walk.

    People seem to be making a mistake in thinking that when the "stimulus" is what we humans consider a "word," then the animal is understanding that stimulus "as a word." But if Marc Pelletier's story is evidence that his dog understands "baballe" as a word, then the examples I gave in my first paragraph are also evidence that dogs can understand the ringing of a bell as a word, or me picking up my dog's leash as a word. But of course those aren't words. …See the leap of logic we tend to make?

  15. Mark Liberman said,

    July 30, 2008 @ 12:52 pm

    Three relevant posts from the Early Bronze Age of blogging:
    "The strange, new sight" (6/11/2004)
    "Canine intelligence" (6/12/2004)
    "Signs or symbols? Words or tools?" (6/15/2004).

    I especially recommend the last post, which quotes extensively from an excellent paper from late neolithic times, Seidenberg, M.S., & Petitto, L.A., "Communication, symbolic communication, and language" Journal of Experimental Psychology: General, 116: 279-287, 1987.

    The points that Seidenberg and Petitto make are simple and persuasive, it seems to me, but it seems to be very hard for people to understand or remember them. It's interesting to consider why this should be so.

  16. Marc A. Pelletier said,

    July 30, 2008 @ 1:16 pm


    Yes, that's an interesting argument; but explain to me *why* exactly you think that the ringing of a bell is less of a word than one in, say, English? If you ring the bell whenever you intend to evoke supper, for instance, there really isn't any mean for a dog to expect that *this* sound your produced is any less of a word than *that* sound.

    For that matter, "picking up the leash" might be somewhat word-like as well (although, in that case, it's obviously much less symbolic thus less clear whether there could be meaning associated with it or just observation). You cannot discount non-verbal messages as "not word"s simply because they were not vocalized: nobody would suggest ASL isn't linguistic because it involves no vocalizations.

    A "word" is any communicative symbol that stands-in for, but isn't, a concept (oversimplified here). If a sound I produce, "baballe", is associated with a concept that isn't the sound *itself* then it's symbolic and it has meaning.

    — Marc A. Pelletier

  17. Kris Rhodes said,

    July 30, 2008 @ 1:39 pm

    Is there anyone who thinks the concept of "word" is an inherently syntactic concept? By this would be meant the idea that you haven't understood a symbol "as a word" unless you have understood how it can be used in a syntactic system?

    Is there anyone who argues for something like this?


  18. Marc A. Pelletier said,

    July 30, 2008 @ 1:41 pm

    @Mark Liberman

    Funny you should have dug up that particular paper on your neolithic dig; that's the very paper that finally decided my to move away from ethology as a field of study entirely.

    When I read that paper, I understood "We notice that there are a number of behaviors that are shared between a toddler and a chimp. We have no evidence whether the same processes and types of knowledge underlie the behaviors, so we're going to posit that they are necessarily caused by entirely different processes, for which we have absolutely no evidence whatsoever."

    That pretty much summarizes the almost unanimous reasoning underlying all the research in the field.

    — Marc A. Pelletier

  19. Marc A. Pelletier said,

    July 30, 2008 @ 1:52 pm


    That's an arguable concept, but I think it would require that you then accept that very young toddler, using their first words, are not in fact using words at all yet.

    Personally, I think "language" is a continuum, associating symbols with concepts is on the low end, the ability to discuss the nature of those symbols and how they work together is on the high end.

    A toddler starts by learning vocabulary alone, and it's an amazing testament to our species ability at linguistic communication that we learn syntax by simple observation afterwards (although there is a school of thought that posits that the ability for syntax is innate, and is simply "tuned" during learning).

    So no, I'm quite sure that a dog could not understand the *concept* of a word as a syntactical unit; but that does not mean they cannot understand them as words. (Nor does it mean, for that matter, that they do — but they behave as if they do at least to some extent and I see no reason to invent a difference without evidence).

    — Marc A. Pelletier

  20. Mark P said,

    July 30, 2008 @ 3:36 pm

    Since I have always doubted the ability of the other primates to use language in the same way that humans do, I would certainly never argue that dogs can do it. But I also think it is possible to define distinctions into existence. It seems to me that there is a continuum of communication among organisms, and the closer you get to the human species, the closer is the communication to that of humans. It's obvious that certain animals can understand some types of human communication to them, but it's equally obvious that it's not exactly what we mean when we talk about human language. But given the great similarity in many areas between humans and other animals (especially the other primates, but dogs as well) I think the danger of anthropocentrism is greater than the danger of anthropomorphism.

  21. S Onosson said,

    July 30, 2008 @ 3:45 pm

    If we have a word "language", does that mean that "language" really exists?? (semi-serious question)

  22. dr pepper said,

    July 30, 2008 @ 4:00 pm

    Can dogs actually parse sounds in enough detail to recognize words? Or is it the accompnying tone? Consider how much trouble many humans have recognizing the exact sounds in a stream of foreign language.

    This reminds me of how my brother liked to amuse himself by calling cats with "hey stupid stupid stupid!", with the intonation normally used with "here kitty kitty kitty".

  23. Jonny Rain said,

    July 30, 2008 @ 4:14 pm

    "I think "language" is a continuum, associating symbols with concepts is on the low end"

    If tones and bells can be considered symbols, and associating symbols with concepts constitutes low end language use, then all pavlovian learning is low end language use. This means that pigeons are linguistic and perhaps Aplysia too.

  24. Rob Gunningham said,

    July 30, 2008 @ 4:19 pm

    I think the danger of anthropocentrism is greater than the danger of anthropomorphism

    I think I pointed out the last time this came up that anthropomorphism is first and most importantly a comparison between us and god, rather than between us and animals, trees or inanimate objects. Ever since the counter-reformation, since Descartes' cogitoI think, therefore I am (and, by implication animals don't think, therefore they aren't, aren't of any worth spiritually, at any rate) animals have been treated by the academy, people like you lot in other words, as if they have little value in comparison to mankind. Those of you in this post who are now going to such lengths to objectify animals, to show how very, very different they are from humans (the otherwise very bright John Cowans, Alan Gunns and Timothy Ms) are really still just toeing the line of the catholic church in the mid-sixteen hundreds, and it's kind of pathetic that you don't seem to realise it. The great Dr Marc A. Pelletier puts you all to shame. (Now you see why Professor Pullum closes the Comments section, but I can't contain myself on this).

  25. Mark P said,

    July 30, 2008 @ 4:57 pm

    Rob, I intended the use of anthropomorphism to refer to the attribution of human qualities to animals. I believe that in the absence of evidence to the contrary, it is perfectly reasonable to compare certain types of behavior observed in animals to the same type of behavior observed in humans, rather than insisting that it cannot be so simply because an animal is not a human. But it is also perfectly reasonable to explore the ways in which that behavior differs from the behavior of a human. All of this serves to further our understanding of both animals and humans.

  26. Marc A. Pelletier said,

    July 30, 2008 @ 5:03 pm

    This is going to be my last response post on this thread, I wouldn't want to abuse the hospitality of my hosts. While the conversation is interesting, it's Language Log and not Ethology Gripe.

    @Mark P: Yes, that's pretty much my position as well, but I don't think the problem is anthropocentrism so much as cultural fear of blurring the line between humans and the other animals — especially primates.

    @dr pepper: I don't know how much, and I can't seem to find a published paper that examines the question. Anectodally, I am confident that dogs can train to differentiate a number of phonemes but trying to find "minimal pairs", as it were, would be illuminating about how dogs parse sounds.

    @Jonny: That raises the interesting question on how actually involuntary Pavlovian learning really is, or how voluntary "human" learning is. Personally, I have no doubt that there is a vast quantitative gulf between humans and "lower" mammals, let alone other classes, but I remain unconvinced that there the presumption that there is a significant qualitative difference is warranted. There certainly is pithy little evidence to that effect.

    @Rob: Incidentally, I only have a Masters and it's not in linguistics. The regular writers here are, I think, interested in communication outside the human species mostly as a hobby, as I am. Good faith all around, right?

    — Marc A. Pelletier

  27. Rob Gunningham said,

    July 30, 2008 @ 5:05 pm

    @ Mark P, I'm sorry, I certainly wasn't aiming my rant at you, but your use of the term reminded me.

  28. Rob Gunningham said,

    July 30, 2008 @ 5:16 pm

    interested in communication outside the human species mostly as a hobby, as I am
    @Dr Pelletier (you'll always be a PhD to me, sir): It's too bad, but those who do it professionally don't appear to have much credibility. I'm not sure why that's so, but I noticed it last time this subject came up.

  29. Timothy M said,

    July 31, 2008 @ 12:08 am

    To Marc Pelletier: Since the above was apparently your last comment, I don't know if there's much point in posting a response, but, well, I want to. Seeing as how we're on topic and this post was made for comments specifically, I can't see much harm in it.

    So, point taken on how non-verbal symbols can be words, too. Still, I don't see how conditioning can be taken as evidence that an animal has understood a stimulus as a word. In your original post, you write:

    Either he associates those two words with specific classes of objects (and only objects of those classes), or he associates the words with the specific play activity and simply knows which object class is required for the game — in both cases there is clear semantic meaning associated with the word

    I was with you up until the hyphen, but where is the evidence that there was semantic meaning involved? If you condition a human by pairing the ringing of a bell with blowing a puff of air in their eye, then they will learn to blink at the ringing of the bell. Would you say that the human associates semantic meaning with the ringing of that bell?

  30. Marc A. Pelletier said,

    July 31, 2008 @ 12:41 am

    @Timothy: I'll bite. It's hard to resist a real discussion isn't it? :-)

    "Would you say that the human associates semantic meaning with the ringing of that bell?"

    That's a difficult one, and a good point. I'm going to say probably not, because you are talking about what amounts to an involuntary response. If there is such a thing as true operant conditioning (which I am not necessarily conceding; that's an interesting question in itself) then definitely not.

    But my example was selected specifically because that wasn't it: the response to my utterance requires voluntary action, and needs to engage at least some minimal cognitive ability (recognizing an activity and classifying objects according to their use, or recognizing an object classes proper).

    Now there is some interesting question as to whether "word" is the right word. It's definitely symbolic (the utterance has no direct relationship to the desired object/activity beyond one that was learned associatively), and it is without a doubt a communication event. Is a symbol used to communicate, by definition, a "word"? I'm certain there are definitions of "word" we can use that would exclude it, but it is what most people mean by the term.

    — Marc A. Pelletier

  31. Rob Gunningham said,

    July 31, 2008 @ 1:46 am

    I'd promised not get cross on this subject, so I'll bow out with apologies, especially to messrs Cowan, Gunn and Timothy M for being rude to them. I think it's important to keep an open mind on these questions — they are still questions — more open than I've been. René Descartes was still an asshole, though.

  32. misterb said,

    July 31, 2008 @ 1:51 am

    Until some humans learn to speak dog, we have no justification in feeling superior because no dogs speak human.

    Dogs and humans live in different worlds perceptually so it's no surprise that we have evolved different ways of communicating. However, we are both social species and therefore have evolved effective methods of interspecies communication. Trying to fit doggie communication into human rubrics is approaching things backwards – after all, we're the smart ones – it should be easier to go the other way around.

  33. Joe said,

    July 31, 2008 @ 3:16 am

    One thing I would caution with an anecdote like this is the importance in animal behavior studies of having strict controls on things. There have been instances of trainers unknowingly giving the animals cues which trigger behavior. I believe one of the more famous instances of this involved the horse that was "able to count."

    That said, it's entirely possible that the dogs associate words with actual meanings. I really don't know one way or the other. I just didn't see it mentioned and thought I'd throw this out there.

    Even so, I see that he had others try this, which may help control for it (though it might help if he wasn't present while they did the test), so there may be something to it.

  34. Jonny Rain said,

    July 31, 2008 @ 6:35 am

    "That raises the interesting question on how actually involuntary Pavlovian learning really is, or how voluntary "human" learning is. Personally, I have no doubt that there is a vast quantitative gulf between humans and "lower" mammals, let alone other classes, but I remain unconvinced that there the presumption that there is a significant qualitative difference is warranted."

    When a rat learns to pair a particular flavor (e.g. grape) with a particular unconditioned reflex (e.g nausea) the rat may learn that grape "means" nausea in some sense of how we understand what "meaning" means. And what the grape flavor means in this sense can probably be discussed by us without granting any voluntary control to the animal learning such an association. So this doesn't make the grape flavor a "word" in any sense (of the word). It may be useful to think of "meaning without language" where different levels of meaning might exist along a continuum even in the absence of language. If the meanings of words in human language find grounding in the areas of the brain responsible for processing perceptual and motor information it may be reasonable to talk about the associations between sensory representations and particular stimuli in animals as constituting a kind of meaning. I think most people here who are arguing in favor of animal language would also feel more comfortable calling these kinds of associations "meaningful" if the animal was aware of this association in a way that permitted flexible behavior, and could voluntarily choose to respond or not to particular stimuli (do sea slugs form meaningful associations?). But this changes the argument to one about animal consciousness or intentionality (or something). It's no longer about language per se. One can model complex associations between stimuli, mental representations and behavior without language.

  35. Rob Gunningham said,

    July 31, 2008 @ 7:22 am

    …the idea that parrots are…on the same developmental trajectory (as humans)…doesn't help anyone understand the situation.

    I'm back. I just wanted to re-post Mark Liberman's remark that animals are on a different trajectory from us. It's good to bear in mind, and it eloquently addresses some confusion.

  36. Kris Rhodes said,

    July 31, 2008 @ 10:41 am

    "That's an arguable concept, but I think it would require that you then accept that very young toddler, using their first words, are not in fact using words at all yet."

    Yes, actually I'd accept that implication. It seems pretty plausible to me that at the beginning, the kid does not understand words as words, but that as he develops, he grows into that kind of understanding.


  37. Timothy M said,

    July 31, 2008 @ 11:52 am


    So if I understand your argument correctly, you're saying it's voluntary action and/or a certain level of cognitive ability that makes you believe your dog understands "baballe" as a "word" or "symbol"?

    Because if that's the case, I don't see how the conclusion follows from the premises. When you give a dog a command such as "sit," it is a voluntary action for the dog to sit. But that is not enough to demonstrate that the dog understands "sit" linguistically; all it means is that the dog "knows" (for want of a better word) that when you say "sit," if it performs the action of sitting, it will get a reward.

    As for cognitive ability, I don't think anyone will disagree with you that dogs have a minimum level (or greater) of cognitive ability (as compared to humans). But the only cognitive ability that is relevant to this argument is linguistic ability. It seems obvious that dogs should be able to mentally group objects that are similar – otherwise every tree and house and telephone poll would seem a completely new object, and no animal would be able to live having to examine every object it sees to see what it is about. But if you think that ability somehow proves that dogs attach semantic meaning to human utterances, then I'm going to need you to connect the dots more explicitly for me, because I don't see how you can get that conclusion.

  38. Mark Reed said,

    July 31, 2008 @ 12:21 pm

    OK, I think there's some confusion over terminology, leading to folks talking past each other, and perhaps even agreeing violently.

    What the heck is "semantic meaning", other than a redundant phrase? Doesn't "semantic" *mean* "meaning"?

    Obviously dogs have a different concept space than humans. Smaller, much less thoroughly divided up, etc. Mappings between them are at best tricky, and certainly not one-to-one. But if the dog will retrieve different objects based on hearing the words for those objects, then he obviously has some idea that those words "mean" those objects. I would maintain that the dog has a mental concept of what a ball is (no doubt different in boundaries from a human's concept), and that concept is associated with the sound of the word for "ball".

    What is the additional requirement that the rest of you are imposing? Does the dog speak a language? No. Is the dog capable of understanding human syntax? No. Are the object names actually "words" as far as the dog is concerned? My previous answers would seem to imply "no", certainly in the linguistic sense. All granted! What's the argument? Is someone saying "yes" to theose questions? It doesn't seem to be M. Pelletier, from my reading…

  39. Marc A. Pelletier said,

    July 31, 2008 @ 12:38 pm

    @Timothy: Basically, yes.

    I do believe that a voluntary response to a command such as "sit" demonstrates that the dog understands "sit" in the same general way as you would understand "go away", say.

    If a policeman walked to you on the street and told you to "stop", you know it is a command to cease moving, and that it is given by someone with authority. You are very likely to cease movement voluntarily, knowing that you may be punished unless you comply. Or you may decide you don't feel like it, or be to distracted, and disobey; but you certainly have understood that the command meant that the policeman wanted and expected you to stop.

    Now, dog almost certainly have a theory of mind (even if quite simplistic) [grant that for the moment]; and one of the imperative of social animals is approval by the group and especially the leaders. In order to get that approval, the dog must know what is *currently* expected.

    When I utter "assis", the dog is informed that what is currently expected is to assume a specific body position and wait. The voluntary act of moving to that position and waiting is consistent with the dog's current model of the world having been modified by my utterance — pretty much the definition of a communicative act.

    Now "assis", or "sit", or whichever the utterance was must be symbolic: it does not position the body, does not resemble or directly represent the correct stance, and is only associated to the act of sitting still through association and memory. (I used the initial good training method of noticing when the dog was in the correct position on his own, giving the utterance and then congratulating the dog for already being in the right stance).

    Now, as far as I am concerned, an utterance that is symbolic and successfully conveys informationis a "word" by any reasonable definition of the term.

    Now, if you'll forgive my excursion to speculative land, I'm going to make the hypothesis that this behavior is entirely explained by the dog "understanding" a very specific subset of the linguistic array: nouns. They have (this is still hypothetical, remember), the ability to associate a number of symbols with such things as classes of objects, individuals, activities, smells, etc. To a dog, the utterance "sit" is the name of an activity. This explains the ability to understand what is desired of them, what is appropriate behavior, invitations to play, and every other bit of human-dog communication I have seen.

    And it explains it entirely parsimoniously: by only positing a very small and very simple subset of an ability that is known to exist in humans. That is a *much* more scientific mindset that discarding all that is known about human communication as qualitatively different and positing and entirely distinct mechanism to exist in other animals that cannot be known any better that the alternative.

    — Marc A. Pelletier

  40. Mark Reed said,

    July 31, 2008 @ 1:02 pm

    Having read the other articles, I find the whole debate somewhat free of substance. It reminds me of the "Can a machine think?" question, and Turing's response "Can a submarine swim?" There seems to be some hypothesis that we humans "really understand" things in some way that goes beyond our ability to convince other people that we have such understanding. While it's certainly possible to understand and not be able to convince or convey that understanding, I do not believe the reverse to possible, under reasonable assumptions about the target of the convincing. In other words, as long as I'm on the subject of Turing, his test is a valid one (assuming a knowledgable tester). If Searle's Chinese room (mentioned in the earlier article) can pass a Turing test, then that room really understands Chinese, even if the man inside does not – and if the man attempts to internalize the algorithm, he will effectively learn Chinese along the way, even if he doesn't set out to do so.

  41. Alicia Casuso said,

    July 31, 2008 @ 2:49 pm

    We need to pose the question in a better way. It seems a lot of people are jumping back and forth between "word", "semantic ability" and "communication". There's a lot of people who have especially dedicated their comments to communication.

    There must be a distinction between communication and a linguistic ability. Even if we proved animals can communicate (and I do believe we have plenty of prove that at least some of them do) that does not mean they have any kind of linguistic ability as we define it. Communication is involved, no doubt, but only as a consequence. I don't think anyone can deny animals communicate. Bees can indicate the distance and direction of nectar to other bees, vervet monekys can climb up or climb down trees depending on the type of cry of another vervet.

    The real phenomena of language is being able to understand (or convey) an expression that has never being formed before, simply by understanding each unit in the expression and the relationships that they have with each other (and I'm sure there's other characteristics like this one, but let's use it as an example).

    The anecdote of the dog who knows left from right window is indeed remarkable, but it can only have linguistic significance if the dog could tell left from right door and left from right paw (given that it previously knows what a paw and a door are but not a left or right door and a left or right paw).

    The same happens with the dog who can get the baballes and the nonoes. It would only be of linguistic interest if the dog was taught to "Sniff" and "Fetch" and then give him the command of "sniff baballes" and "fetch nonoes". Could the dog know what to do when asked to reverse his searching methods for the item looked for? Or must he be taught again the meaning of the words in combinations such as "fetch-baball", "fetch-nonoes", etc. ?

    Lastly, I would point out that saying dogs probably can't use language the way we do should in no way offend dogs. It does not mean that there could not be other types of representational systems that could work differently, and could not be even imagined by us (and never understood by us). And who's to say language has really made us better than anyone? Under what standards? Our own? I'd say a dog's life could be more fulfilling for a dog than a man's (and viceversa). So what use could they have for our language? Thinking that it's unfair we don't expect dogs to be more like humans is assuming we are better. Let the dog be a dog, it's nothing to be ashamed of.

  42. Timothy M said,

    July 31, 2008 @ 3:39 pm


    Well, I think I've understood your argument a bit better this time! Before I get into it though I'd like to start with a disclaimer. In an earlier comment, you wrote that dogs behave as if they do understand words to some extent, and you "see no reason to invent a difference without evidence." I agree with you wholeheartedly. I feel it necessary to say this because people in this thread seem to be chomping at the bit to accuse others of wanting to believe there is a difference and inventing ridiculous explanations to support it. Believe me, I have no such want. The reason that I'm having this argument is because – to me – there is evidence that there is a difference between humans and dogs regarding understanding the meaning of "words" or "communicative symbols." …I'm just having a difficult time presenting that evidence convincingly to other people. ;) But I think I have a better understanding now of where we disagree in this argument, and I'd like to continue the discussion with that in mind, if you're up for it.

    When a human issues a dog a command – be it with a word, a motion, a tone, or whatever – we can call that an Event. If the dog responds to Event X with the appropriate Action for that event, then it gets a reward (whether that reward is tangible or intangible). I believe the crux of our argument is whether the Event is – from the perspective of the dog – just an event, or whether it is understood as a "communicative symbol." If it is just an event, then the reason dogs are able to follow human commands has nothing to do with language. The dog has simply learned that when Event X occurs, if the dog performs a certain action, then it will get a reward. However, if dogs understand these Events as "communicative symbols" or "words," then the reason they are able to follow human commands is because they understand their basic meaning in a linguistic way.

    So I'd like to suggest a short thought experiment: Imagine you are sealed in a room. You are walking around the room, exploring it, and every once in a while you hear a bell ring. You think nothing of it. Tired, you decide to sit down. The bell rings again while you are sitting, and immediately after, a machine in the room dispenses food for you. You stand at the machine and eat the food. You finish the food, but are still hungry. You hope that the next time the bell rings, the machine will dispense food again. Eventually, the bell rings again, but no food is forthcoming. You stand there, pondering. After several more minutes, the bell rings again, and you realize that the one thing that was different the time that the bell rang and you got food was that you were sitting. Immediately you sit down, with only about a second having passed since the ringing of the bell. And as soon as you sit down, the machine dispenses food again. Now you've learned that if you sit down when the bell rings, you'll get food.

    If you were the person in this story, would you think that the bell was actually a form of communication? Would you think that the bell was a command that meant "sit"? Personally, such an idea would be very unexpected to me. There is nothing in the story that would suggest that the bell ring was actually language. And even if you disagree with me on this point, there is still the fact that a person in such a situation could obtain food simply by understanding that when Event X occurs and they perform a certain action, they get food. There is no need to posit the existence of a higher-level understanding of the event – an understanding in which it was not just an event, but "communication." That is why, for example, when researchers teach monkeys to push a lever in response to the color of a computer display changing (which if done correctly will score them juice as a reward), no one says that the experiment demonstrated the linguistic abilities of monkeys. The color change was not communication and the monkeys did not understand it as communication – it was an Event, to which they responded, and got a reward.

    So, whaddaya think?

  43. Marc A. Pelletier said,

    July 31, 2008 @ 8:53 pm


    It is an interesting thought experiment, and one that needed me to think hard for a while before I figured out why I disagreed with your conclusion.

    I think the problem is the nature of the event you describe: Having a bell ring will evoke in the human the image of a mechanical event, and will cause that human to discount the sound as a communication event.

    If you accompany the bell with a lighted display showing a glyph the human is not familiar with, or substitute it by an utterance that is obviously of human origin (even if not in an intelligible language), your human subject will almost certainly interpret the very same sequence of event as a command given.

    At least that's where I feel the difference is.

    — Marc A. Pelletier

  44. Andy J said,

    August 1, 2008 @ 5:51 am

    I would like to come at this subject from another direction. Dogs communicate with other dogs (and cats and humans) through a variety of means, principally barking or other simple verbalisations such as growling or whinning. Even humans can recognise the different meaning of a warning bark from, say, a "don't forget me, I'm part of the family too" bark. How they acquire these vocalisations (inherent or learnt) I wouldn't like to speculate, but clearly they understand something when another dog in their pack uses a particular vocalisation to mean "danger" or "intruder" or whatever. When they live with humans we become their pack, and so for them to associate some of our simple vocalisations (or other cues such as picking up a leash) with particular activities seems to me to be a straightforward extension of their own "language". If you accept that barking (in all its forms) operates as a primitive language then the idea of a dog accepting "baballe" into that vocabulary because it helps him co-operate with his pack leader doesn't seem too outrageous. He is limited by his lack of suitable vocal cords from articulating the word "baballe" or "walk" to show he wants to initiate either of these activites, but as any dog owner will attest, there are plenty of ways dogs can indicate they want certain basic things.

  45. Peter said,

    August 1, 2008 @ 6:24 am

    In this discusion we are forgetting that dogs may be able to communicate with their owners telepathically, as Rupert Sheldrake speculates:

    Of course, there is another whole argument to be had about whether telepathy, supposing it to exist, necessarily involves the understanding of concepts as words.

  46. Timothy M said,

    August 1, 2008 @ 8:45 am

    @Marc: The point I'm trying to make is that it is possible to perform the correct action when given a command whether an organism understands that command as language or not. The thought experiment above illustrates this – both you and I agree that under those circumstances, we would not have had any understanding of the bell as language, yet we would have been able to sit "on command" and gain a reward just the same. Surely this ability to respond to an event with an appropriate action is present in dogs as well as humans, and it seems to me that is a cognitively simpler task than understanding sounds, motions, etc. as words. Therefore, the fact that dogs are able to respond to human commands is not sufficient evidence that they understand those commands as words like we do. There is a cognitively-simpler alternative, and while it isn't always true that the cognitively-simpler alternative is the correct one, it would also be presumptuous to assume that the more cognitively-complex option is correct.

    @Andy J:

    If you accept that barking (in all its forms) operates as a primitive language then the idea of a dog accepting "baballe" into that vocabulary because it helps him co-operate with his pack leader doesn't seem too outrageous.

    Sounds reasonable to me, if barking really is a primitive language. But, to use the phrasing we've been using throughout this thread, do dogs understand barking as "words", or are they too just events that they respond to?

  47. Marc A. Pelletier said,

    August 1, 2008 @ 1:07 pm


    Perhaps we arguing from closer positions than was at first apparent, but I think we disagree on what the conditions for an event to be a "word" in the linguistic sense are.

    As far as I'm concerned, an event is a "word" if it is voluntary, symbolic and intended to convey meaning. Maybe "lexical token" would be a better term. Now, it is entirely possible that the intended recipient, not expecting a lexical token, might reject it as one (or be unable to understand it as one), but it does not change the objective classification of the event. This does not rule out that something that was not a lexical token be misinterpreted as one either.

    I'm quite sure you don't need to understand what a lexical token is in order to use them any more than you need to understand how an internal combustion engine works in order to drive a car— a dog is not limited in its use or understanding of lexical tokens by its lack of understanding that there may be several kinds, or that they could be used in a complex grammar.

    Now, I am convinced dogs do expect lexical tokens in many situations. If you give attention to a dog and make some unfamiliar sound/noise in a "command" way, you will see the dog act in obvious confusion— cock his head attentively, look around for something, try one or two variations on known commands or even whine in obvious frustration if the command is repeated while you appear unpleased. The dog obviously expected the sound to be a lexical token conveying information on what was expected of him, but is unable to associate a meaning with it.

    So yes, I would say that barks are lexical tokens in this sense; even if their entire grammar reduces to "statement: noun".

    — Marc A. Pelletier

  48. Andy J said,

    August 1, 2008 @ 3:00 pm

    @Timothy M. "But, to use the phrasing we've been using throughout this thread, do dogs understand barking as "words", or are they too just events that they respond to?"
    The reason I raised the barking issue is that between dogs there is some form of communication through barks; it's a two-way process: Dog A says "Ruff" and dog B knows what that means, and dog B can also say "Ruff" and be understood by dog A. That seems to me to be more than just a kind of Pavlovian response mechanism. Perhaps Marc's lexical token is a better description of the bark sounds, and there may be many other factors such as repetition, intensity, hierarchy within the pack etc which I'm ignoring here, but this still allows human words to be incorporated within a dog's "vocabulary" and represent a concept in the dog's brain beyond that of pure response.

  49. Chris Schoen said,

    August 1, 2008 @ 11:38 pm

    I am wondering why Geoff Pullum seems so insistent that dogs are unable to attach semantic meaning to words uttered by humans beyond the level of conditioned reflexes.

    –Marc P.

    What the heck is "semantic meaning", other than a redundant phrase? Doesn't "semantic" *mean* "meaning"?

    –Mark R.

    If there's a circle around the moon, it "means" that it might snow. If I find marine fossils on a mountaintop, it "means" that the peaks carry rock that was once seafloor. But these occurrences don't exist in order to convey this information. That's the difference between a sign and a symbol. Humans respond to signs all the time, both consciously and otherwise, as do dogs. Getting out of bed to the sound of alarm clocks, for example. But humans also respond to symbols, which create concepts in their minds, and there's no evidence dogs can do this.

    In the case given, babelle and nonos do not indicate concepts; rather they are a signal that certain games are about to start. Marc writes that there is "clear semantic association with the word[s]," and that the dog's response is not automatic. But neither is our response necessarily automatic when we wake to an alarm clock. If we're tired, we may hit the snooze button, or just sleep in. Even if we get up every day at the same time without fail, it would not be fair to describe this as automatic, in a Pavlovian sense.

    It becomes clearer that we don't attach any symbolic (representative) meaning to the alarm if we consider what happens if we use the radio alarm function. If asked what the semantic meaning of the radio announcer's words are, we would not answer "to wake us up." The words are a sign that it's time to wake up (a complicated enterprise that is far from automatic), but the semantic content of the words could be the weather, or the recent G8 summit, or anything else.

    Marc writes "explain to me *why* exactly you think that the ringing of a bell is less of a word than one in, say, English?" The difference is that the ringing of the bell may indicate that dinner is coming, but it can't be used to stand for dinner. We can't use the bell to talk about dinner, what kind of dinner might be nice, what some of the best dinners of days gone by might be, how much it cost to make or buy dinner, or any other like expression. The bell indicates only that dinner is coming; it cannot be used to represent the concept of dinner, which is what a true word does. This is much more than a mere "lexical token" (a term which begs the question, because it has the latin root for "word" in it. If you said "utterance" you'd be on firmer ground, but again, there is a big difference between utterances and words, as all primatologists know.)

  50. Timothy M said,

    August 1, 2008 @ 11:50 pm

    @Marc: Well, my initial intent was to argue that dogs do not understand human commands as words (or "lexical tokens", whatever – I'm not hung up on the term we use to describe them), because that is what my intuition is telling me. Ultimately I wasn't able to come up with any good arguments to support my intuition, but I was able to come up with an argument (if you accept it, of course) for why your argument does not succeed. The result of that is what I wrote in my comment before this one.

    Anyway, I have no issue with your definition of what makes an event a word. But since it is a description about the person who causes the event (i.e. the speaker), it doesn't change much with regard to the discussion about the listener.

    As for the evidence you give for why you think dogs expect lexical tokens, could it not be said that dogs instinctively know to give attention to humans (and especially their masters) when they utter command-like sounds? That sounds like something that dogs would have been selectively bred for. But just because you give attention to something doesn't mean you understand it as a lexical token.

    Secondly, if dogs have learned the connection between "master has said something in a commanding tone" and "now I have to do something," it doesn't seem so odd that they would try to perform the appropriate action even if they didn't know what that was. Again, is this alone really evidence that dogs understand commands as having a meaning?

    At the risk of writing too much, let me finish with one more example of what I mean by responding to a certain Event with a certain Action. Let's say there is a small animal and a larger animal that meet in the woods. The two keep their distance but neither one leaves the area. Then, the smaller animal notices that the larger animal has adopted a certain stance. At this point, the smaller animal is in danger, but it doesn't know that yet. All of a sudden, the larger animal attacks. The smaller animal is injured, but manages to get away. Now, you can be sure the next time that animal comes in contact with the larger one, when the larger animal makes that stance the smaller animal will know it's time to run away, or at least back off. The association has been learned – when Event X happens, I perform Action X. Certainly this is not proof of any linguistic ability, right? Now let us say that the larger animal was actually a human, and in that human's tribe they use that stance to command others to "get out of my sight or I'm going to hurt you" (they're a very violent people). If that human comes in contact with that same animal again (or one that has had a similar experience), when the human takes up his stance and the smaller animal runs away, the human may take that as evidence that the animal understands a part of his language. "Hey, I told that animal to get out of my sight, and it obeyed!", he may tell his friends. But in reality it is almost certain that the animal did not understand the stance as a lexical token.

    @Andy J:

    Dog A says "Ruff" and dog B knows what that means, and dog B can also say "Ruff" and be understood by dog A. That seems to me to be more than just a kind of Pavlovian response mechanism.

    What if it is instinct that Dog A say "ruff" whenever it senses danger? That wouldn't require any cognizance of "ruff" as having the meaning "danger." And then Dog B knows, through instinct or learning, that when it hears "ruff" it has to be on the alert, or run, or some such appropriate behavior.

    I may have just had an epiphany, but it seems to me that an animal with linguistic ability would eventually come to realize that "ruff" means "danger," whereas an animal without that ability would forever be unable to make that connection, even though it could perform the correct response when it heard the sound. Humans understand commands as language and perform the correct response, so we assume that because dogs perform the correct response, they understand commands as language too. But I think it's perfectly plausible that one could respond correctly having only learned Event –> Response. Furthermore, there are plenty of examples of Event –> Response (such as the one I gave in my comment to Marc) that we don't consider evidence of linguistic ability. Thus it still seems to me that we only think dogs understand certain Events as Language because we understand those Events as Language.

RSS feed for comments on this post