There is some brand new research on those New Caledonian Crows that seem to be able to do all sorts of things we thought only smart Humans could do (like making tools, understanding physics, that sort of thing). I've written it up here: The Incredulous New Caledonian Crows, on 10,000 Birds, where, as you know, I write a monthly entry on bird evolutionary biology. It turns out they can do something that I'm not so sure all humans are great at. Other than flying.
Please go have a look!
I would be surprised if many intelligent animals didn't do the same thing; it just makes sense as a way of interacting with the world. My cat will put up with or sleep through almost any amount of familiar noise, but if there's even a small atypical noise (e.g., last night, an envelope sliding off a pile of stuff while she was napping), she will hop up to investigate it at once. In a potentially dangerous environment it would not be rational behavior to ignore unexplained events, but it also wouldn't be rational to burn energy repeatedly investigating things that you already believe to be harmless. So there ought to be some kind of logic built in to avoid those outcomes.
Likewise, while animals may not score well in the tests behaviorists use to demonstrate "theory of mind", as your essay mentions, I have to believe that any animal who practices stalking-type hunting has at least an unconscious, intuitive understanding of the fact that what they know and what the other guy knows may not be identical. My cat has engaged in various actions that seem to rely on this point. For example, she often comes into the bedroom when I get up in the middle of the night, entering simultaneously with me by sliding around my ankles in the doorway. (This is voluntary, as the door is always kept open a crack for her.) She used to get accidentally bumped or stepped on, until she developed the habit of giving a little murmur (resembling her "good morning" murmur) as she was coming by. She never does this when she passes someone in a doorway in the daytime. The simplest explanation of this behavior is that she understands that I only punt her at night by accident because I don't see her; hence a warning is useful at night and needless in the daytime. While she certainly knows that she's down by my feet, she comprehends that I do not know it until I'm told.
Jane, there's a difference between comprehending the UCA/HCA distinction and habituation and that difference is what's being investigated.
With regard to ToM, I'm more in agreement with you in that what ToM is, as a characteristic of cognition, is ambiguous and, more to the point, in my opinion most likely something that isn't integral with well-defined boundaries. It's probably not correct to say that an animal does or does not possess it in some absolute sense.
That said, I think you're not quite seeing how ToM, by definition, differs from merely having a complex predictive model of behavior.
In my opinion, the distinguishing characteristic is that a ToM is teleological — it postulates an agent with goals, and those goals are presumed to explain and predict behavior. Therefore, identifying an agent's goals is an efficient heuristic for understanding and predicting behavior, as opposed to, say, a long list of past behaviors and their exceptions coupled with a probabilistic analysis. A mind lacking a ToM will use the latter regarding the behavior of others, to greater or lesser degrees of sophistication. A mind with ToM will use this teleological shortcut — it will still require observation and list of exceptions, but presuming the existence of a goal, and presuming that it is among a class of familiar goals, makes this task much easier.
The degree to which an animal needs to predict complex behavior in other animals (its own species or otherwise) is the degree to which a ToM becomes more valuable. It's not necessary.
My own sense is that ToM is strongly coupled with socialization — that is to say, it's far more likely in highly social animals than in less social animals. And those social animals need to be, though social, relatively independent agents capable of complex cognition and behavior. That is to say, quite a lot like humans. Why? Because the social relationship aspect puts it in a context of minds that are evolutionary already very similar — they're of the same species, interacting in a (somewhat, at least) cooperative shared environment. That makes the comprehension of another mind much easier — it's a mind like the comprehending mind, in a similar local environment. Presumably, it will have similar goals. Being relatively independent, though, is important too because eusocial animals, for example, have less need to comprehend each other's actions through a ToM because, well, they're not that different or unpredictable. Or, put differently, there's just not the need. You have more incentive to comprehend another mind when it's working against your own goals than with them. In other words, the social environment needs to be a mixture of cooperation and competition. And, finally, the cognition and behavior need to be complex because, otherwise, you'd not need the shortcut of teleology to efficiently guess at future behavior.
One thing I think is extremely thought provoking about this is that this implies that non-human intelligences might be very alien to us. That is, they won't necessarily require a ToM to be highly intelligent, but lacking a ToM-oriented intelligence, they would be very, very different from us and difficult to understand.
Put differently, ToM is an idea often discussed within the context of the autistic spectrum. I've had (weird) conversations with AS folk who feel that there's something invalid and even offensive about ToM. That is to say, they believe that it's a shortcut that is both misleading and sometimes oppressive. They feel that their avoidance of teleological thinking about other people, their avoidance about attempting to get inside other peoples' heads and discern motives, enables them to think more clearly and accurately about peoples' behavior. Which, to some degree, they're right. That is to say, they're right precisely to the degree to which people use ToM and are mistaken about others' inner-self and motivations. The more unlike someone else is from ourselves, the more likely we'll go astray. If we limited ourselves to "just the facts" of past behavior, we wouldn't. Which is the accuracy part of the AS person's argument. The oppressive part is that non-AS folk assume everyone is like themselves when, in fact, AS folk aren't.
Anyway, my point in bringing that up is to illustrate the possibility that ToM is not dependent upon intelligence but is rather related to a particular kind of cognition. Your cat can comprehend complex behavior on your part without a ToM. On the other hand, I'm very disinclined to assume that non-humans necessarily lack a ToM — as Greg writes, some other primates are known to have ToM, as well as some other species. Just presuming that ToM is exceptionally human is far too anthropocentric past and proven erroneous assumptions about humans for my liking. Nevertheless, I think there's good reasons to believe, as I argue above, that ToM is rare and strongly related to characteristically human kinds of cognition.
I'll point this out to clarify/muddy the waters a bit as the case may be: Well designed experiments clearly indicate that ToM is not a property found in human children below a certain age. The "toy in the box" experiment shows that a young child assumes that everyone else has in their brains what the child has in his or her brain. Actually, spend a few days with a toddler and that becomes apparent in a more casual sense.
Object permanence is a phenomenon involving the ability to know if something still exists when it is hidden. Below a certain age a child seems to act as though a teddy bear vanishes from reality when covered with a cloth; they don't expect the teddy bear to be there when the cloth is removed. Above a certain age, they are shocked to find that the teddy bear has vanished when the cloth is removed.
These two phenomena are probably related, as they both require thinking about things you can't see.
Other well designed experiments, however, show that an infant is actually surprised (in a sense) when the cloth is removed and the teddy bear is gone. They don't make inferences or reason anything out with the assumption that a hidden object exists, but they stare longer at the space where the object was supposed to be.
If you designed an experiment in a non-linguistic species like a crow to see if they had a sense of object permanence, all they would have to do is to pass the lesser of these two tests, because that is all you could supply. Looking time, glancing frequency, etc. would be the measure, like in the crow experiment linked to above.
What this means to me is that humans that are typical and adult have assembled a set of cognitive abilities into a repertoire that we understand in the context of typical human operations, but that at least some of these parts are present but not employed in relation to each other from much earlier. This, therefore, implies, maybe, that one can imagine a creature that has roughly the same set of "parts" of a cognitive system but assembled differently, or at different life stages, or differently by sex, or whatever.
To me this implies that applying a set of carefully designed similar experiments across a range of animals will give you only a first order look at what is there and not there.
This has actually been demonstrated; If I recall correctly, Elephants have certain cognitive abilities that are found in humans and some apes...they pass a Gallup test. But, the Gallup test as applied to humans and chimps does not work on elephants. They don't care if there is a blotch of red paint on their forehead (they don't see red, I think, and are typically covered with piles of crap and stuff anyway). They are tested in some other way (I can't recall the details well enough to safely cite them here... here's a paper: http://www.pnas.org/content/103/45/17053.abstract ).
Bottom line: No matter how important research like that cited here with the crows is, we can't understand the whole system of cognition without a larger scale contextualized look at the system.
One more item: This may not apply to crows, but beware the primates. Human cognitive abilities are a function of the cultural milieu in which we are raised perhaps as much as any other source of information. Humans raised without "culture" or a highly limited culture do not turn into functionally typical humans. This is also true of chimps. Sadly, chimps and bonobos raised in the absence of normal chimp or bonobo culture are not even close to normal or typical. Since these two things are certainly true, one can imagine this model:
H = gH * cH (Humans are a product of human genes and human culture)
P = gP * cP (chimps are a product of chimp genes and chimp culture)
So, then, what would this be?
X = gP * cH
Thus, the amazing feats by a bonobo like Kanzi, raised mostly among humans, are very important and interesting, but are not direct analogs for presumed Last Common Ancestor behavior.
Note to self: Turn this into a blog post some day.
Keith (and Greg), that's a very cogent explanation - thanks. I still think my cat has an effective understanding that others have goals - for example, she knows that I wish to avoid stepping on her. If it might be that I just didn't care if I hurt her, it could be counterproductive to alert me that she was within convenient punting range, so she'd just stay quiet and keep a safe distance.
But I'm intrigued by your suggestion that autistic people do not have a theory of mind. How is this determined? I am certainly more likely to draw conclusions about people's future behavior from their observed past behavior than to try to guess what they may be thinking, which I cannot know. But it would seem very difficult to accumulate enough data for a "probabilistic analysis" of the behavior of each person you met, except, say, someone who is obviously dangerous, if you did not have an underlying assumption that their behavior was connected to some kind of intention. Without that, it would seem that autistic people would have to evaluate the behavior of others in the way that we would evaluate the behavior of someone who was doing crazy things apparently at random. Is that really true? That would certainly make life stressful, if so!
As for the UCA/HCA distinction, I didn't adequately defend my cat's superior abilities :-) - but let me try again. In the case where something fell and made a noise while she was sleeping, that was an unknown cause. If something slides off a pile of papers and makes noise when she is awake and looking at it, she does not alert. (Yes, this has also been observed repeatedly, and yes, we're messy.) Is this not perhaps comparable to the "mysterious sight" vs. "adequately explained sight" in the crow study?
"But I’m intrigued by your suggestion that autistic people do not have a theory of mind. How is this determined?"
Well, I didn't mean to suggest that this was a case. Rather, I meant to say that arguably, as I've heard from both AS folk and non-AS folk, some people on the AS have a more narrow ToM or, perhaps, don't rely upon it as strongly. The AS interlocutor I mentioned seemed to be arguing that a ToM is more like one of a number of tools in a cognitive toolkit that he mostly avoids using and, when he does use, he uses it very carefully. But someone else — and that would be me, for example — might argue that no human that could have such a conversation with me lacks an active ToM that he's utilizing. My argument is that a ToM is a predictive model of behavior built around the presumption of an agent with comprehensible goals, where those goals are the primary heuristic within the model.
My interlocuter, though, would probably reply that his critique of ToM is not a denial of it in that absolute sense, but rather is a critique of it used beyond the very most basic and safe assumptions about another mind.
There's a saying among those of us who write and teach about Behavioral Biology: "Everybody has a cat. Or a dog."
But seriously, I caution people against inferring anything by reference to their cat or dog for two reasons; the first reason makes the scientists happy, the second one makes the cat/dog owner happy, and they are both perhaps valid.
1) Even common, repeated, and intense casual observations that seem to show a pattern (of anything in nature) need to be tested under experimental (sometimes natural) or controlled conditions before conclusions can be drawn in a scientific framework.
2) Domestic animals are strange. They do things not seen in non-domestic animals that are similar. If any sort of human-ness rubs off on domestic animals, that contaminates them and while they remain interesting, things they seem to be able to do should not be added to the repertoire of what animals do.
Keith - I'd say your "AS interlocutor" has a more rational viewpoint than most, if others would disagree with it. I was pondering this issue further last night and realized that not only is it impossible to know which of several possible goals motivates average people's behavior, but other people's behavior does not match anything that they would think of as their goals. For example, I have one colleague whose goals almost certainly are NOT "To screw up major research projects and give everyone around me a migraine," but sadly, behavioral observation allows the conclusion that that's what to expect of him!
Greg - Your second point is fair, but domestic animals are reported to be duller than wild relatives, and they lead often more constrained and certainly safer and less challenging lives. So while they certainly pick up new behaviors from us, I wouldn't assume that they have much more fundamental cognitive capacities than animals in the wild.
Answering your first point risks starting one of those pointless arguments over whether there are ways of knowing things other than science, and then whether all observation is science or whether some peoples have no knowledge, etc., ad nauseam. Most of what humans have learned about the behavior of wild animals has been from observations that were not "controlled," not "replicable" at will, and hence perhaps not scientific knowledge - but in my opinion, still knowledge. Last night, my cat for some reason pried open the drawer in the bathroom that holds the fabric swiffer covers and spread them all over the floor (then came to notify us that there was a mess that should be cleaned up). I can't make her open that drawer again, but even if she never does it again I have adequate evidence for me to believe that she can do it, even if you do not have adequate evidence for you to believe the same claim.