Monkeys in the mirror and the nature of science

Charles Darwin wondered if animals were aware of themselves. Allowed to visit a rare orangutan in the London Zoo, he brought a mirror and observed the ape apparently make faces at its own reflection. It’s hard to say for sure that the orangutan really was aware that its reflection was its own. Over a century later, a scientist named George Gallup turned Darwin’s idea into a more rigorous test. He would secretly put a mark on an animal’s forehead and see if it noticed the difference the next time it passed a mirror.

Human adults pass this test, but young children don’t, suggesting that our self-recognition takes time to develop. Some chimpanzees (our closest relatives) seem to pass this mirror test, but others fail it. Orangutans also show mixed results. Beyond the primates, studies have indicated that magpies, dolphins, and elephants pass the mirror test. In a new paper in PLOS One, Luis Populin of the University of Wisconsin publish what may be the first compelling evidence that monkeys pass the mirror test too.

It’s a surprising result because people have tried to find evidence of self-recognition in monkeys before. Most scientists failed. The Harvard primatologist Marc Hauser claimed in 1995 that the cotton-top tamarin could pass the mirror test, but that paper was one of several that Harvard now claims were tainted by Hauser’s misconduct. Populin and his colleagues came across their first clues of self-recognition by accident. They had implanted electrodes in the skulls of rhesus monkeys for a different study. They keep mirrors in the monkey cages just to stimulate the animals, and they noticed that the monkeys started spending a lot of time looking at themselves in the mirrors after surgery.

To see if the monkeys were really aware that their appearance had changed, the scientists put different mirrors into the cages. Some were big and some were small. Some were made of ordinary glass, while others were painted glass. The monkeys looked much more often into the real mirrors than the blackened ones. The scientists then introduced an even bigger mirror into the cages, which allowed the monkeys to see their whole bodies. The mirror hung from the top of the cage so they could turn it around. The video above shows a monkey without an implant inspecting one of these big mirrors. It doesn’t use any of the gestures it might if it met another monkey. Instead, it seems to be inspecting its body. Every nook and cranny, in fact.

I asked two experts on self-recognition, Lori Marino and Frans de Waal, both of Emory University, what they thought of the paper.

Marino was enthusiastic:

I’ve been reading this article over and over again all morning and have looked at the videos. Here is my reaction: I think that this is potentially a very important study. The videos are absolutely convincing. At first I thought it was the saliency of the implanted head device that made a difference but the videos show that monkeys without the device also use the mirror and, even more importantly, the monkeys are using the mirror to explore OTHER parts of their bodies, i.e. genitals. That can’t be explained away in some sort of ‘stimulus saliency’ explanation, as far as I can tell. There are two areas that I wish I had more information about. First, exactly what was their prior mirror exposure? Is there anything they can report about how long that was and if there were any interactions with the monkeys during that prior time that could have made a difference in terms of the outcome of this study? Second, they report that the monkeys “failed the mark test” but there is no information given about the methods used in the mark test. It would be important to know how that was conducted.

With all that said, this is – by far – the most compelling evidence for MSR in monkeys to date. I have been trying to find an alternative explanation for the results – and haven’t come up with one yet. I’ll continue to return to it to see if any come to mind. I think that these findings show that self-awareness is not an all-or-nothing phenomenon as Gallup has asserted. I don’t find this conclusion particularly surprising because it has been extremely difficult for anyone to come up with an explanation for the supposed “discontinuity” between great apes and monkeys. I think the reason for this is that there is none.

De Waal was more circumspect:

It is hard to say what is going on as the cap may be seen as a “supermark,” as the authors call it, but it is of course a mark that is not only seen but also felt. As a result, the monkeys have two sources of feedback at the same time, the image in the mirror and the sensation of something new on their head. This is different from the classical mark test, which has only one source of information (visual). It seems clear that the convergence of these two sources of perception is helpful to achieve self-inspection, and this is an interesting finding, but the authors still need to explain why rhesus monkeys apparently cannot do the same with just the visual information.

The hallmark of the mark test is its spontaneity, purely on the basis of visual input, so in this sense this is different. Rhesus monkeys have been tested many times and never pass the test. I think you need to talk to Dr. Gordon Gallup and see what he thinks. It is unclear to me what to conclude, but this study is quite different from e.g. the magpie study, which applied a purely visual test.

The idea that mirror self recognition is a black & white distinction (you either have it or you don’t) was first challenged in another monkey study that we conducted, in which we showed that capuchin monkeys do not seem to see a stranger in the mirror: they seem to distinguish the monkey in the mirror from another monkey, strange or familiar. As a result, we proposed a gradual scale of self awareness. The piece of intriguing information presented here may support this view, but I am sure many scientists would want more tests and more controls.

[Update: New Scientist reports that Gallup shares De Waal’s reservations.]

[Update: Peter Roma, who published a 2007 paper in which he failed to find evidence for self recognition in monkeys, rejects the new one:

Although the video samples are provocative, I cannot agree with the conclusion (and title) of the paper.

The lack of social behaviors towards the mirror is irrelevant because the monkeys all had an extensive history with mirrors prior to the study, so there was no reason to expect social responses after years of habituation to reflective surfaces. To anthropomorphize, they may still think the monkey in the mirror is another animal, but over the years they’ve learned that he’s harmless.

The examples of putative genital viewing were not convincing either. The authors repeatedly asserted that the monkeys used the mirrors to view areas they could not see directly, but monkeys can see their genitals unaided, and they play with them all the time with or without mirrors! Even the video samples show the monkeys looking at their genitals directly then viewing the same area(s) in the mirror. This is why scientists do the mark test!

In my view, the most compelling evidence was the first video of the monkey touching the head implant while holding the mirror. There is no doubt that the monkeys could not see the implant without a reflective surface, but the key here is whether or not this self-examination behavior occurred more frequently in the presence of the mirror vs. without. The authors report increased incidence of touching “unseen” areas in the presence of the mirror (figure 2C), but these data include touching the cranial implant and the genitals. I suspect these data are artificially inflated by what the authors perceive as mirror-guided genital examination, which even in the video examples did not appear to be anything more than typical stereotyped “acrobatic” behaviors often seen in individually-housed rhesus monkeys. The authors provide no data on the frequency of just cranial implant touching with vs. without mirrors, and no visual evidence except for the single incident from the video. Why wouldn’t they report the number of implant explorations independently of the genital viewing?

My primary concern is that all monkeys failed the mark test, and the strongest apparent evidence of mirror self-recognition (MSR) was only seen in two monkeys following cranial surgery–a manipulation with strong tactile cues that could elicit exploration regardless of the mirror’s presence. Their argument rests largely on the assertion that the cranial implant is a “super mark” that somehow awakened a latent ability in the monkeys to self-recognize, but it’s unclear why the implant would be more visually salient than a brightly contrasting color marking on the face. The more parsimonious conclusion is that the tactile sensation of the implant was enough to elicit exploration, but even then, the authors provide no evidence that implant exploration occurred more frequently in the presence of the mirror vs. without.

If the authors’ hypothesis is true that a cranial implant serves as a “super mark,” then their procedures warrant replication, which frankly they should have done before making such a bold assertion. Currently within the Order Primates, the overwhelming preponderance of evidence still limits MSR and the fundamental cognitive precursor to a “sense of self” to the apes.

This back-and-forth is not just interesting in itself, both for what it says about monkeys and what it says about ourselves. It also says something about science. If you read some of the more extreme comments about the Hauser affair, you’ll find some people trying to indict the entire line of research into how human and animal minds evolved. This new paper, and its reception, shows just how absurd that radical rejection is. Science is bigger than individual scientists, and the mirror test will survive.

[Update: PLOS One paper linked fixed]

29 thoughts on “Monkeys in the mirror and the nature of science

  1. This is very interesting. There is another feature of the mirror reaction that might not indicate self-recognition. Every time the monkey moves, or generates the brain process that initiates movement, the image in the mirror changes. This correlation may be innately fascinating, and is the type of correlation that any learning organism should pay attention to, like learning to use a new tool. The curious animal will likely explore the self-movement/mirror correlation.

    Harder to explain is exploration of an unexpected mark on the animal’s body. Like seeing a pimple on your face in the mirror.

    Overall, it would be surprising to me for an animal to pass tests of “theory of mind” (as monkeys and other animals do) and fail tests of self recognition. But things aren’t always what you expect.

  2. I’m glad to see you drive this point about the danger and folly of rejecting Hauser’s entire line of research b/c of the misconduct issue. As this bit of work shows, this is an extremely fruitful and sound way to plumb the past. Hell of an interesting study, too.

  3. The video linked here is ambiguous in that it seems to show a monkey displaying stereotopies from being housed singly in a cage. The monkey appears agitated and the seeming genital inspection is something I have often seen in caged male monkeys over the years when a mirror is not present. We used to refer to it at the “moon illusion” for obvious reasons. If I had only this evidence I would be unconvinced as it is not easy to separate the stereotypies from the use of the mirror. However after reading the article and looking at other movies I am convinced that these authors have presented evidence that rhesus monkeys pass the mirror test. I am not surprised, but it is nice to have evidence that supports what seemed very likely from the start. It is very difficult to imagine how a rhesus monkey could operate within a group of 100 or more monkeys without being aware of what was the monkey itself and who was a separate monkey. Self-awareness, in the sense of knowing you are you would seem to be a basic requirement of living in complex social groups. For me the issue with rhesus monkeys has not been whether they knew themselves from others, they obviously did as evidenced by their behavior in complex groups, but why they failed the “mirror test”.
    The first part of the explanation came when I visited the Davis Primate center where mirrors attached to monkey’s cage fronts are used as enrichment. I was struck be the eeery sensation of looking through a small windo into a housing room with two parallel ros of cages and seeing 20 or more hands reach out, take the mirror on their cage front and turn it so that they could see how was peering in their window. I knew that this was what they were doing as I could now see 20 or more monkey faces staring intently at me. There could be little doubt that these monkeys knew that they weren’t me and had learned the contingencies of mirrors. I note in this current study that mirrors had been used as enrichment in this case too. It may take rhesus monkeys a long time to overcome their intiatial reaction to mirrors as Gallup had shown that they didn’t show such mirror instrumentality after 75 hours of mirror exposure, if I remember correctly. However, it now seems clear that monkeys do use mirrors when they have had long-term interactions with mirrors. It also seems clear that they recognize themselves in the mirror for who they are and once they learn mirro contingencies do not confuse the monkey in the mirror as anyone but themselves.

    I think it likely that it would be impossible to form complex hierarchicalsocial groups if the individuals did not know themselves from others. The fact that the vast majority of monkeys have never had the chance to see themselves in a mirror and fail to use the mirror to inspect themselves after short exposure to the mirror and its contingencies, is not evidence that they confuse themselves with others. This study nicely demonstrates that given sufficient experience with mirrors and using more naturalistic evidence, that rhesus monkeys do indeed know that the monkey in the the mirror is sometimes themselves and sometimes someone else. Congratulations to the authors.

  4. I don’t understand why Marino thinks a discontinuity between monkeys and great apes requires explanation. It is possible that self-awareness arose in an ancestor of the apes, with magpies, elephants, and dolphins being cases of convergent evolution. Of more concern to me is the supposed discontinuity between gorillas and other apes. Why would chimps and orangutans have this ability, but not gorillas?

    A minor correction to the comment about Hauser: The 1995 mirror study was not one of the ones in which Harvard found misconduct. Gallup challenged the study and years later Hauser acknowledged that he could not replicate the results, although he never retracted the 1995 report.

  5. I’m uncomfortable doing an evolutionary or cladistic analysis relying on something like the mirror test. Mirrors are not part of the natural habitat of monkeys or apes. There is no pressure to recognize yourself from a mirror image. From what I see, passing a mirror test, with appropriate controls, shows a positive ability, an aspect of self recognition. But failure does not establish anything. There are numerous reasons members of a species might fail. A mirror test result does not equal a cognitive capacity, let alone a trait that arose due to evolutionary dynamics.

  6. I’m no scientist, but it seems I read somewhere recently that bonobo monkeys pass the mirror test as well & that we have some common descendency (is that a word?) to them. Can’t remember where I read this. Does anyone know about the bonobos? They are a fascinating species!

  7. I would be interested in a digital display (like facetime for Iphone) that showed a mirror-like image, with different short video’s of other animal in the mix

  8. Apologies for the self-link here … but I just want to toss in my two cents: if the results hold up, then we need to think more about how monkeys are used in research.

    This may also be the case even if the results don’t stand. IMO, the mark test is a behaviorist leftover; to quote neuroscientist Jay Neitz from an interview that didn’t make it into my piece, “I would bet that anyone not trained in psychology who works with the monkeys would guess that they have a sense of themselves as separate creatures.” Perhaps that doesn’t say much for psychology. “Working with them every day it is obvious that they know the other monkeys individually and they recognize different people very easily. If they recognize other people and monkeys as individuals I believe most people would assume that they know themselves. “

  9. I disagree strongly with the logic of the Jay Neitz quoted by Brandon Keim.

    “If they recognize other people and monkeys as individuals I believe most people would assume that they know themselves.”

    (Well, most people might assume that, but I wouldn’t).

    Individual recognition and self knowledge (or a “theory of mind”) are very different things. If memory servers, EO Wilson speculated in ‘Sociobiology’ that a major pressure for brain expansion in mammals is individual recognition, which permits the formation of complex social structures. I believe Carl has written on brain size / social structure correlations.

    But being able to discriminate numerous individuals, and learning interaction rules for each, does not suggest that the learner has a “theory of mind”. At least not to me.

  10. It seems logical, if not scientific, that if a Monkey or Ape recognizes others then said Monkey or ape would recognize themselves. They may not recognize themselves initially when looking into a mirror, but they will have a cognitive sense or who and what they are. Otherwise we’d have monkeys hearding with Zebras trying to mate with Crocodiles.

  11. This video arouses the kind of curiosity that Darwin no doubt felt. It begs questions rather than answers them. Gollop’s mark test is quite different from simply putting a mirror within view of an ape or monkey and watching what happens. Darwin’s mirror test is a million miles from Gollop’s mark test.

    The question is: how far away from Gollop’s test is the one used in this study? A purely visual test compared with a visio-tactile test.

    Gollop’s mark test is a test to see whether a primate has a mental representation of its own appearance. It tests for a specific ability: to be able to recognise one’s mirror image as being how one’s body appears. I recognise how I look.

    If an animal can tangibly sense the presence of a physical object attached to its head and is aware that it is possible to use mirrors to see things which are otherwise invisible, then it can use the mirror to explore the object without the awareness that what they are seeing is how they look. The mirror merely aids the monkey’s investigation of the cause of the tactile sensation on the top of the head.

    However, awareness that something physical is happening to the top of one’s head, and having the wit to use a mirror to check out the situation, does require a visio-spatial mental model of the body. It requires self-recognition of a less personal nature.

    Does this study give a better measure of the gap in self-awareness between monkies and great apes? Less of a gap than had hitherto been thought but a significant gap nonetheless?

    Are mirror tests sufficiently strong scientific instruments to support such weighty conclusions?

    How can we test the reliability of mirror tests?

  12. In response to Stephen Push’s comment: “I don’t understand why Marino thinks a discontinuity between monkeys and great apes requires explanation.”, I’d like to clarify. I didn’t make it clear in my original statement that I was talking about the apparent discontinuity between the history of failed marked tests in monkeys and other scientific findings that show macaques capable of meta-cognitive abilities and other forms of awareness that never jibed well with the mirror failures. So, to make a definitive black-and-white argument that animals as complex and intelligent as macaques are devoid of self-awareness entirely, as Gallup has proposed, is too simplistic. There are many levels and facets to intelligence and awareness. The present study adds to our understanding of those processes.

    Moreover, given the findings of self-awareness in dolphins, birds, elephants – clearly this is a convergent capacity that may have something to do with some animals having attained a certain level of cognitive sophistication. I have written many papers proposing this very idea.

  13. Is there any acknowledgment of stochastic variation in intellectual capability within a monkey species? If only small numbers of monkeys have been tested, one researcher could get some average subjects while another got a particularly capable one.

    Two extra-terrestrial researchers testing small groups of humans on mathematical ability could reach very different conclusions about our intellect due to random variation. It’s not clear to me whether different individual ability has been considered.

  14. Personally, I have always found the conclusion “can use a mirror” -> “is self-aware” to be inherently incorrect:

    o If we consider self-awareness in the sense of a relatively high cognitive ability of having an understanding of oneself as mental entity, as humans do, the mirror tests proves nothing at all.

    o If we consider it a limited ability to just know that one exists as a physical entity separate from, e.g., the other monkeys, the mirror test does not prove more than what could be concluded (with a similar degree of speculation) from other behaviours (e.g. self-cleaning without a mirror).

  15. There are likely many kinds of “self”. For example, it/she/he is looking at “me”. A mental map of a home range (or any location) is of little value unless “self” is also on the map. At least, we need to distinguish between “reflection” and “self”. Monkeys who are experienced with mirrors know the relationship between mirror reflection and other objects or persons. I am convinced that with experience, most anthropoids recognize the reflection of their own bodies (at least as they would process a shadow), but the use of that information to examine themselves requires processes more complex than recognition. The Gallup test has served as a convincing demonstration of these more complex abilities.

Leave a Reply

Your email address will not be published. Required fields are marked *