Biology & Philosophy

, Volume 28, Issue 4, pp 657–673 | Cite as

Mirror self-recognition and symbol-mindedness



The view that mirror self-recognition (MSR) is a definitive demonstration of self-awareness is far from universally accepted, and those who do support the view need a more robust argument than the mere assumption that self-recognition implies a self-concept (e.g. Gallup in Socioecology and Psychology of Primates, Mouton, Hague, 1975; Gallup and Suarez in Psychological Perspectives on the Self, vol 3, Erlbaum, Hillsdale, 1986). In this paper I offer a new argument in favour of the view that MSR shows self-awareness by examining the nature of the mirror image itself. I argue, using the results of ‘symbol-mindedness’ experiments by Deloache (Trends Cogn Sci 8(2):66–70, 2004), that where self-recognition exists, the mirror image must be functioning as a symbol from the perspective of the subject and the subject must therefore be ‘symbol-minded’ and hence concept possessing. Further to this, according to the Concept Possession Hypothesis of Self-Consciousness (Savanah in Conscious Cogn 2011), concept possession alone is sufficient to demonstrate the existence of self-awareness. Thus MSR as a demonstration of symbol-mindedness implies the existence of self-awareness. I begin by defending the ‘mark test’ protocol as a robust methodology for determining self-recognition. Then follows a critical examination of the extreme views both for and against the interpretation of MSR as an indication of self-awareness: although the non-mentalistic interpretation of MSR is unconvincing, the argument presented by Gallup is also inadequate. I then present the symbol-mindedness argument to fill in the gaps in the Gallup approach.


Self-awareness Self-consciousness Mirror self-recognition Symbol Symbol-mindedness Mark test Concept 


In a landmark 1970 paper Gordon Gallup Jr. describes a test he had devised to demonstrate that chimpanzees are able to recognise themselves in a mirror (the mark test). Since then Gallup has consistently maintained that mirror self-recognition (henceforth MSR) is a sufficient demonstration of self-awareness1 (e.g., Gallup 1970, 1975; Gallup and Suarez 1986). The general idea is that one needs a self-concept in order to recognise oneself in the mirror. However, MSR is not universally accepted as an indicator of self-awareness. Furthermore there are objections to both the validity of the mark test itself as a demonstration of MSR and to the experimental methodologies applied. In this paper I address these objections and present a positive case for MSR as a valid demonstration of self-awareness using a new approach that focuses on the nature of the mirror image itself. The result is an argument founded on solid theoretical grounds and consequently more robust than the argument presented by Gallup.

To achieve this aim requires several steps. Firstly, I establish that the experimental methodology employed in the mark test is sound (i.e. is rigorous enough to eliminate false positives/negatives) in order to validate the mark test as a definitive demonstration of self-recognition. Next, I analyse the extreme inflationary and deflationary views on self-recognition and conclude that neither provides a satisfactory account. Although I agree with Gallup’s ultimate conclusion that MSR implies self-awareness, his argument is inadequate in that it is not founded on theoretical grounds beyond simply assuming that self-recognition implies some level of self-awareness and therefore full-fledged self-awareness. On the other hand, the accounts based on associative theories alone (those not involving a self-concept) do not stand up to scrutiny. On the contrary, I conclude that one core argument on the deflationary side (Mitchell’s (1997a, b) kinaesthetic-visual matching account) actually favours the conclusion that MSR implies self-awareness.

Finally, I present my own argument that MSR does indeed imply self-awareness. The argument is composed of two independent but individually significant parts. Firstly, based on the studies of ‘symbol-mindedness’ by DeLoache (2004), I show that MSR subjects are symbol-minded. I argue that a mirror image recognised as the self cannot be taken as an actual object behind the mirror (i.e. an extension of the body) and so most plausibly is taken as a representation of the self. Thus, it functions as a symbol. As such, species that have passed the mark test, such as chimpanzees, must be considered ‘symbol-minded’, a capacity that develops in human infants at around 18 months of age (DeLoache 2004). Secondly, based on the Concept Possession Hypothesis of Self-Consciousness (CPH; Savanah 2011), symbol-mindedness implies concept possession and hence self-consciousness.

The concept possession hypothesis of self-consciousness (CPH)

Informally, the notion of self-consciousness I am considering here is just what we all recognise in ourselves as human beings—the question being, of course, whether any animals have something similar. However, if pressed for an explicit definition of self-consciousness, I give something like the following: an understanding of one’s own existence as a psychological subject with intentional agency. This more-or-less captures what I think many would accept as a self-concept. Contrast this with organisms that can happily interact with their environment without a self-concept—or indeed without possession of any concepts at all. That is, they do not understand anything but act according to genetically encoded (‘hard-wired’) predispositions (some might say ‘instincts’) and/or through associatively learned behaviour. Of course there are other views of ‘self-consciousness’ and ‘self-concept’ but what I have given will suffice for my purposes here.

Given the above definitions, it is not controversial to say that self-consciousness requires concept possession, since obviously possession of a self-concept requires the capacity for concept possession generally. However, the further claim made in the Concept Possession Hypothesis of Self-Consciousness (CPH; Savanah 2011) is that concept possession alone is sufficient evidence of self-consciousness. Another way to put this is that if you possess any concepts at all you must already possess the self-concept. Several arguments are presented in Savanah (2011) to support CPH. First, conception is likened to perception to illustrate that, just as perception always involves the self-percept,2 conception always involves the self-concept. Second, giving examples, I argue that concepts do not exist in isolation and that every concept can be linked to the self-concept via a web of necessarily associated concepts. Finally, I use Bermũdez’s (1998) theory of non-conceptual content to support an argument that the only factor separating organisms that are conscious but not self-conscious from organisms that are both conscious and self-conscious is the fact that only the latter are capable of mental states with conceptual content.

The upshot of CPH is that any conclusive demonstration of concept possession in a subject is sufficient evidence of self-consciousness in that subject. For example, if it could be demonstrated that a subject has thoughts in propositional form, or is capable of inferential thinking, or is symbol-minded, then that subject possesses concepts and, according to CPH, is self-conscious. The key conclusion in this paper is that MSR implies symbol-mindedness. Given that conclusion my further claim, relying on CPH, is that MSR is sufficient evidence of self-consciousness.

The mark test

In Gallup’s original experiment on chimpanzees, two female and two male wild-born subjects with little or no presumed prior exposure to mirrors were tested. After a couple of days of habituation to mirrors, the chimpanzees were observed engaging in mirror-aided self-directed behaviour, such as mirror-guided self-examination and experimentation with facial gestures. Gallup took these observations as evidence of self-recognition by the chimpanzees. However, in order to create a more objective measure of MSR Gallup devised the mark test. In this protocol, subjects are marked on parts of the face visible only in a mirror. Subjects who see these marks on their reflection in a mirror and react as though the marks are on their own face (i.e. by touching them) are deemed to have passed the mark test and shown evidence of self-recognition. The number of touches within a set period of time provides a quantifiable measure. Gallup marked his chimps on the uppermost portion of an eyebrow ridge and the top half of the opposite ear. Gallup went to great lengths to avoid any possibility of his subjects knowing about the marks before seeing them in the mirror. His subjects were marked under anaesthesia with an odourless, non-irritating dye and so had no information about the marks due to olfactory or tactile cues. Furthermore, after the subjects recovered from the anaesthesia, they were observed for a suitable period of time before further mirror exposure to ensure no behaviour indicating subject knowledge of the mark application (Gallup and Suarez 1986).

Gallup applied the mark test to confirm self-recognition in subjects only after previous observations of self-directed behaviour in front of mirrors. In his studies with chimpanzees, Gallup reported that for several days after initial exposure to mirrors his subjects reacted to their own reflections as if they were seeing conspecifics; that is, they displayed social responses such as bobbing, threatening and vocalizing. After a few days the chimpanzees began responding to their reflections as themselves, evidenced by observations of mirror-guided self-directed behaviours as listed earlier. To pass the mark test the subjects must inspect the marks while watching their reflection, though Gallup further reports some instances of visual and olfactory inspection of the fingers that touched the marked areas (Gallup 1975).

Since Gallup’s original 1970 paper the mark test has been applied, albeit with some variations of technique, to several other species and also human infants. Of primates, great apes (save wild-born gorillas) tend to pass the mark test (see for example Schilhab 2004; Patterson and Cohn 1994; Swartz and Evans 1994) while old world monkeys do not.3 Regarding non-primates, much more experimentation needs to be done to convince the mainstream; nevertheless, there are promising results for dolphins (Marten and Psarakos 1994; Marino et al. 1994); elephants (Plotnik et al. 2006); and Magpies (Prior et al. 2008).4 In human infants the question focuses on the age at which they pass the test, as eventually they must in the course of normal development. This appears to be in the 18–24 month range (Amsterdam 1972; Schulman and Kaplowitz 1977; Bertenthal and Fischer 1978; Lewis and Brooks-Gunn 1979; Bigelow 1981).

Objections to the mark test

Objections to the methodology

Although some researchers failed to replicate Gallup’s findings with chimpanzees (Swartz and Evans 1994) sufficient evidence has now accumulated that chimpanzees as a species pass the mark test (Schilhab 2004). Various reasons explain individual differences in mark test results within a species; in Gallup’s original study the control subjects—chimpanzees who were given the mark test without prior habituation to mirrors—failed to pass. Also, Povinelli et al. conducted a study showing ontogenetically based differences, whereby most chimpanzees pass the mark test after about 4.5–8 years of age (Povinelli et al. 1993). This of course should come as no surprise as even human infants do not pass until around 1.5–2 years (Amsterdam 1972; Schulman and Kaplowitz 1977; Bertenthal and Fischer 1978). One would expect that development of higher cognitive capacities in species less developed than humans would, ontogenetically speaking, lag human development.

There is an apparent anomaly in the putative phylogenetic development of the capacity for self-recognition in that among the great apes only (wild-born) gorillas fail the mark test (although captive-bred gorillas have passed Patterson and Cohn 1994). It has been suggested that the cause of the failure has more to do with methodological problems of the mark test (such as a lack of controls for species-specialisation) than a true lack of self-recognition in this species (Swartz and Evans 1994; Patterson and Cohn 1994). For example, some animals that are too impatient in front of the mirror or too distracted by the presence of experimenters might fail the mark test only because they miss any opportunity to grasp the relation between mirror image and somasthetic sensations (de Veer and van den Bos 1999; Schilhab 2004). Gaze-aversion or threat stares in primates might prevent subjects from making enough eye contact with their mirror images to establish self-recognition5 (de Veer and van den Bos 1999). Thus, perhaps captive-bred gorillas can pass the mark test because they are unencumbered with the species-specific impediments of their wild-born cousins. If this is the case then perhaps gorillas as a species do indeed have the capacity for MSR but (except for captive-bred gorillas) have not yet been able to demonstrate it in experiments, in which case the expected correlation between phylogenetic development and self-recognition may hold. That is, ‘advanced’ species such as gorillas may have the capacity for MSR despite an inability to demonstrate this in the field. However, the same argument of species-specificity might also be used against this correlation: less advanced animals might also have the capacity for MSR and yet fail the mark test for the same reasons. This highlights the importance of not forming conclusions too early. In a classic example of this, elephants failed to pass the mark test in early experiments (Povinelli 1989) but (at least in one case) did pass later when elephant-sized mirrors were used (Plotnik et al. 2006). Overall, Gallup’s methodology appears to be sound.

Doubts over interpretation of results

Schilhab (2004) casts some doubt over the validity of the mark test for MSR as applied to human infants. She referred to results obtained by Lewis and Brooks-Gunn (1979) in which some infants, on seeing their mothers with a rouge marked nose, would touch their own noses even without the benefit of a mirror being present. This makes the mark test seem inconclusive: the child might attend to the mark on his nose even if interpreting the mirror image as another child. The objection does not hold against tests that used an alternative behaviour to indicate MSR (Bigelow 1981). In Bigelow’s method a clown face suspended from above was silently lowered behind the subject and could therefore only be seen by the subject in a mirror behind his own self-image. If upon seeing the clown face in the mirror the subject turned to look behind himself the subject was deemed to have demonstrated MSR. According to Bigelow, the subject would have realised from what he saw in the mirror that it was his own image that the clown face was behind. This assumption may not be valid; as I argue later, an organism might be able to use mirrors effectively in certain ways without self-recognition having been attained. Indeed, Bertenthal and Fischer (1978) applied ‘Hat’ and ‘Toy’ tests that were essentially equivalent to the clown face test to 48 infants aged 6–24 months and also applied the standard mark test. In their sample, passing the Hat/Toy tests always occurred before passing the mark test rather than at the same developmental stage as assumed by Bigelow. Despite these variations of MSR test protocols on children the overall results all fall within the same broad age range. Children generally demonstrate MSR between 18 and 24 months.


With regard to the overall methodology of the mark test, including its variations, the objections do not appear strong enough to warrant invalidation of the claim that the mark test demonstrates MSR, at least for positive results. The combination of observed mirror-directed behaviour plus the protocol of the mark test seem sufficiently rigorous to accept positive results as true indications of MSR. There are some lingering doubts over negative results, however. Species-specific traits such as eye-contact aversion; insufficient habituation to mirrors or other effects may mask the potential capacity for MSR in some species, such as wild gorillas and old world monkeys. In human infants, despite variations of the mark test, the results from several studies are in broad agreement. On the balance of arguments I conclude that the mark test is methodologically sound and presents sufficient evidence to prove MSR. Nevertheless, failure to pass the mark test does not necessarily rule out MSR capacity.

MSR and self-awareness

In this section I discuss the ramifications of MSR for self-awareness. The extreme views here can be called the ‘Gallup approach’ and (after Schilhab 2004) the ‘non-mentalistic interpretation’.6 The inflationary Gallup approach is to simply equate MSR with full-fledged self-awareness including relatively sophisticated cognitive abilities such as theory of mind and even personal identity. At the other extreme is the deflationary non-mentalistic interpretation (e.g., Heyes 1998), in which associative theories are used to explain MSR without the need to invoke the existence of a self-concept. I do not review the full range of views but I briefly analyse these extremes as an introduction to a third approach that I name the ‘Indirect Interpretation’. In this approach I evaluate MSR based on the role of the mirror image itself. Although I conclude that MSR does provide sufficient evidence for self-consciousness, I find the Gallup argument inadequate. The Indirect Interpretation, explicated later, provides the missing theoretical grounding for a more robust argument.

The Gallup approach

The approach taken by Gallup obviates the need for reasoned argument. It relies on the view that one either does or does not possess self-awareness. Gallup insists, unjustifiably in my view, that self-awareness is ‘all-or-nothing’ in the sense that one cannot be partially self-aware. As Mitchell (1997a) says of Gallup: “…he identifies the self-awareness necessary for mirror-self-recognition with any and all other forms of self-awareness” (p. 23). These other forms include Theory of Mind (ToM; Gallup 1998; Gallup and Suarez 1986) and even personal identity: “…the ability to infer correctly the identity of the reflection requires an identity on the part of the organism making that inference” (Gallup and Suarez 1986, p. 4). By taking this position, Gallup then needs only to show any indication of some kind of self-awareness to close his case. Thus, solely on the basis that self-recognition must imply some level of awareness of the self, Gallup concludes the subject must therefore have an “integrated concept of self” (Gallup 1975, p. 331).

The Gallup approach is implausible in that it does not allow for levels of self-awareness or for a progressive development of cognitive capacities of which MSR is only one stage. For example, whereas a reasonable argument may be made that a demonstration of ToM implies the existence of self-awareness (on the grounds that ToM relies on modeling others based on knowledge of the self), the reverse is not necessarily so. It is conceivable that an organism has developed self-awareness without achieving ToM. Furthermore, in at least some experiments used to gauge ToM abilities (such as ‘false belief’) this cognitive capacity appears to develop in children at a generally later age (3–4 years; Wellman et al. 2001) than does MSR (18–24 months). Also by this age infants have not yet developed other significant cognitive capacities such as episodic memory, which some see as necessary for a sense of personal identity (e.g., Suddendorf and Corballis 1997).

Given the lack of both substantial argument and supporting evidence, I cannot endorse the Gallup approach. MSR itself does not indicate the full range of cognitive capacities associated with self-consciousness and indeed there is no reason to simply assume that self-consciousness is all-or-nothing. I do believe, however, that MSR provides sufficient evidence for a level of cognitive capacities that we can comfortably characterise as self-consciousness, as I explain later.

The non-mentalistic interpretation

If we are to attribute any level of self-awareness to nonhuman organisms on the basis of their passing any type of test we must take care to rule out accounts that explain the observed behaviour without the need to invoke the possession of a self-concept, for example as merely stimulus–response reactions. Heyes (1998) claims that this condition has not been met with respect to MSR studies on nonhuman primates. For example, Heyes cites the pigeon studies of Epstein et al. (1981) to argue for a non-mentalistic interpretation: the finding that pigeons can apparently be trained to pass the mark test.7 Chimpanzees’ passing of the mark test might be essentially no different to the trained pigeons’. Since grooming plays a crucial role for chimpanzees it is argued that self-grooming might count as positive reinforcement and hence substitute for the food reward used to train the pigeons. Thus, passing the mark test is here explained using a non-mentalistic, associative account.

However there are several problems with this argument. Firstly, self-grooming (i.e., mark-touching) was not the only behaviour chimpanzees displayed in front of the mirror. They also engaged in facial gesturing and examination of body parts visible only in the mirror, behaviours not observed in pigeons. Furthermore, grooming is a social behaviour so self-grooming is unlikely to have the same reward value. Others also show similar scepticism, especially as the results have not been replicated (Thompson and Contie 1994). These objections taken together give me cause to reject the pigeon data as evidence to support the non-mentalistic interpretation of the primate data.

Mitchell (1997a, b) claims that MSR could be better explained by kinaesthetic-visual matching rather than the ascription of a self-concept to chimpanzees. This argument is intended to show that a subject passing the mark test might only be recognising its own body. Since all organisms can differentiate self from non-self, passing the mark test might thus represent nothing more than this type of ‘own body recognition’ with no concept of the self involved. In kinaesthetic-visual matching the organism facing the mirror is mapping its proprioceptive sensations onto the mirror image of its moving body. In this way, the organism purportedly demonstrates what Heyes (1998) loosely describes as a ‘body concept’.8 According to Heyes this is not what we refer to when using terms like ‘self-concept’ and ‘self-awareness’. Thus, Heyes and Mitchell both give an associative account of the organism’s mirror experience, whereby the organism correlates sensory input from the operations of its body with sensory inputs from elsewhere (visual data from the mirror image).

What is not made explicit in this argument is whether the subject takes the mirror image as its actual (physically extended) body, or as a representation of the body. It seems unlikely that the proponents of this line of argument are proposing that the subject is taking its mirror image as a physical extension of its own actual body, given that the image is distal and unconnected. However, I argue that it is only if the mirror image were taken as a physical extension of the actual body that the argument could support a non-mentalistic interpretation, as it is the proprioception of the actual body that corresponds to a body schema, not the image. Given that a primate will see much of its body directly as well as simultaneously reflected in the mirror, we should assume that it not only does kinaesthetic-visual matching but also ‘visual-visual’ matching. The movements of (say) an arm and the arm’s identical mirror image will be perfectly synchronised and both simultaneously visible. Any visual matching of kinaesthesia going on is more likely to correspond to the sight of the primate’s actual arm moving in the (non-mirror) visual field rather than its mirror image, as it is the actual arm’s position in space that is accompanied by proprioception, not the mirror image.

That the arm and the arm’s mirror image are both simultaneously visible might then lead to an alternative interpretation to the one defended by Mitchell and Heyes. Certainly I agree that the kinaesthetic-visual matching argument implies the organism recognises the mirror image as its own body, but the point I want to emphasise is that it does not recognise the image as part of its physical body. The organism learns to associate its kinaesthetic sensations with both the sight of its actual body and its identical visual image in the mirror. However, as only its actual body parts match the sensations of proprioception, and as the mirror image is distal and unconnected, it acquires the knowledge that the mirror image is only a representation of its actual body. The implications of this fact for self-consciousness is explained below.

The indirect interpretation

I think of this approach as ‘indirect’ because, unlike Gallup, I do not make a direct link between the self that is recognised in MSR and the possession of a self-concept. It is significant that the self is recognised, but not in the way Gallup and his supporters say; the route from MSR to self-awareness requires intervening steps. As I explain below, the reason MSR provides an excellent paradigm as a positive demonstration of self-awareness is because it is incontrovertible evidence of symbol-mindedness. Symbol-mindedness requires concept possession and hence (according to CPH) implies self-consciousness. I begin with a brief discussion on symbol-mindedness and its connection with concept possession.

Meaning, concepts and symbols

According to Deloache (2004), a symbol is something that someone intends to represent something other than itself. What denotes something as a symbol in DeLoache’s view is the symbol’s intentional9 nature. That is, what makes a symbol a symbol is the specific intention of the symbol’s creator that it is taken as a symbol. Of course this means it is a symbol in the eye of the symbol creator. One can make an equivalent definition to represent the viewpoint of a symbol perceiver. I offer the following definition to capture this perspective: a symbol is something that someone has taken to represent something other than itself.10 The key element here is the understanding that the perceived object is not an instance of the represented object but a representation.

When I refer to concepts I am using the term in its common, everyday usage. That is, to have a concept of something just means to understand what type of thing it is. Concept possession can thus be characterised as the ability to understand or to ascribe meanings to things. ‘Meaning’ here can be taken in the Gricean sense of ‘non-natural meaning’ (Grice 1957), which, according to Grice, is a tendency to produce a belief in the subject. To have a propositional attitude such as belief requires conceptual capacities. Thus, the ability to ascribe a meaning to something implies concept possession.

Concept possession is necessary for symbol usage, as recognising something as a symbol implies ascribing a meaning to it. For example, an organism might see a pie and detect its affordance11 to be eaten—possibly without ever having a concept of food or eating. The organism does not need to ascribe the meaning ‘food’ to the visual image of the pie in order to appropriately interact with it. The pie is not taken as a symbol; it does not represent something other than itself. But another organism might see π and recognise it as standing for a number with particular geometric properties—that is, standing for something other than itself (a squiggle12). In this case, the organism is symbol-minded and has treated the visual image as a symbol. In order to treat it as a symbol—that is, to ascribe a meaning to the squiggle—the organism must be concept possessing.

Before presenting the argument that MSR shows symbol-mindedness it will be instructive to review DeLoache’s symbol-mindedness experiments on human infants. Firstly, this will elucidate the nature of symbol-mindedness, which will be useful when later applied to chimpanzees. Secondly, it will provide an age-based comparison with the onset of MSR in children.

Symbol-mindedness in human infants

DeLoache (2004) studied infants up to 3 years of age and performed tests designed to indicate whether they were symbol-minded. When the infants treated symbol-objects as if they were ‘real’ objects, they were considered to be not yet symbol-minded. For example, a 9-month-old infant placed his lips on a photograph of a baby bottle. There are other behaviours that also correspond to a lack of symbol-mindedness, such as manually exploring the symbol-object or even attempting to grasp the depicted object in a picture.

DeLoache describes one experiment where infants observed experimenters hiding a miniature toy using a symbolic object such as a realistic scale model of a room and were then asked to locate the real object in the real room. Children who have achieved symbol-mindedness recognise the miniature toy and scale model as symbols13 representing a real toy and a real room and are able to locate the real hidden toy within the real room. The symbol-minded child is able to act upon the meanings of the symbols in this experiment: the miniature toy hidden behind a miniature chair in the scale model of the room means that the real toy is hidden behind the real chair in the real room.

As one might well expect, there is no definite cut-off age at which an infant suddenly becomes symbol-minded. Apart from the obvious variation in developmental progress between different children, the same child will perform differently on symbol-mindedness tests depending on a variety of factors. For example, in the hidden toy experiment children who did not fully appreciate the symbol-referent relationship between the scale model and the real room obviously performed poorly on the test. However, their performance could be improved by decreasing the salience of the model as an object by placing it behind a window (and hence presumably increasing its symbolic nature).

Despite the inability to narrowly define the age range at which children become symbol-minded, the evidence suggests an age range centred more-or-less around 18 months. I do not think it a coincidence that this age range is about the same as when infants generally pass the mark test, as explained below.

Symbol-mindedness and MSR

Earlier when discussing kinaesthetic-visual matching theories of MSR, I proposed that organisms demonstrating MSR could take their mirror images in one of two ways. Either the mirror image is an extension of their own actual body, or it is a representation of their body. I then argued that the former alternative is implausible, given that proprioception matches body movements with visually accessible actual body parts but not with their mirror images (and, of course, because the images are distal and unconnected with the actual body). If so, then MSR subjects understand their mirror images as representations of their own bodies. In effect, they are treating the images as symbols—they are not their actual bodies but representations of them.14 A reviewer suggested a third alternative, that the mirror image is being used as a natural sign of positional and configurational information rather than as a representation, in which case concept possession need not be assumed. The natural sign/representation distinction is indeed important, and for most mirror images the explanation of the MSR data in terms of natural signs is plausible. However, I argue that when the mirror image is taken specifically as the self the representation interpretation is more plausible.

A natural sign correlates with aspects of the environment but, unlike a symbol, is not ascribed any meaning by an apprehender. I will use the example provided by a reviewer: footprints in the sand. These can be considered natural signs rather than symbols as they correlate with an aspect of the environment (i.e., the path walked). An animal might just detect the affordance of paw prints to lead to (say) a waterhole. This could be just a learned association due to many exposures to the advent of paw prints connected to waterholes (or conceivably a genetically hard-wired pre-disposition). An organism need not possess concepts in this case. Nevertheless, when an organism interacts with a natural sign, this does not mean that they are not concept possessing. For example, a human can undergo reasoning along the lines of “those are my footprints from when I started the walk, therefore by following them backwards I can find the trailhead”. To undergo such reasoning implies concept possession.

On the other hand, footprints can be actual symbols. For example, in some buildings such as hospitals or pre-schools, artificial footprints can be placed on the floor deliberately with the intention of leading people in particular directions. In such cases they act just like arrows on a signpost. These artificial footprints are then not natural signs but symbols. In some cases, conceivably, an apprehender might not even know whether the footprints are natural or artificial. Nevertheless, either way, the apprehender might react to the footprints in exactly the same way, i.e. to follow them (as in the earlier example of paw prints). Deacon (1997) emphasises that no objects are intrinsically symbols but are interpreted as such (p. 71). This point is crucial for my argument: it matters not whether the sign is intrinsically a natural sign or a symbol but how the apprehender interprets it.

In almost all cases involving non-linguistic organisms determining how the apprehender has taken the natural sign is extremely difficult as all we have to go on is behaviour. Animal behaviour can almost always be interpreted in multiple ways with varying levels of assumed cognition involved, just as the foot/paw prints example shows. Another simple example is when a lab rat presses a lever to gain a reward. One might wish to argue that the rat knows (that is, has a conceptual understanding) that pressing the lever will result in a food pellet. But, as per the oft-quoted Morgan’s canon, we should explain animal behaviour according to the lowest psychological faculty feasible, and assume no more than that the rat has encoded an association between the lever press and a reward via training. Indeed, non-mentalistic accounts of MSR present just such associative types of explanation. However, in the case of MSR I argue that ascribing concept possession to the subject is indeed the more plausible explanation.

A mirror image correlates with aspects of the environment and so can be considered intrinsically a natural sign. But a mirror image, like footprints, could still be interpreted by an apprehender as a symbol. Also like the footprints example, in most cases we could not tell from observation of a subject’s interactions with the mirror whether it is using the image as a natural sign or interpreting it as a symbol. However, when the image is specifically of the self, we can tell. To start, let’s examine the non-self case, say, the image of a banana.

Imagine a set-up whereby a mirror is angled in such a way that a primate can see the reflection of a banana but initially cannot see the actual banana due to a removable barrier. It is easy to predict that the primate will first seek the banana behind the mirror where it appears to be located. Most likely it will eventually locate the actual banana, and I think we can safely assume that after repeated trials the subject will learn to locate objects using a mirror. Now, it is tempting to say that the primate has learned about mirror properties and knows that the image in the mirror is not an actual physical object (though it indicates the proximity of the actual object). If so then I would be committed to ascribing self-consciousness to it. The reasoning is as follows. If the mirror image is understood to not be a physical object located behind the mirror then it must be a representation of the actual object. In effect, the image is being treated as a symbol much as the toy and room models were in the Deloache experiments: it stands in for something else (the actual object). Thus, if a subject takes a mirror image—any mirror image—as not a physical object behind the mirror but standing in for one, then it is symbol-minded.

However, in the banana scenario a plausible explanation exists that does not assume such a high level of cognitive capacity. It could just be (for instance) that the primate has learned to associate the mirror image with the existence of a similar object in a different location. Even though it may eventually stop searching behind the mirror, we cannot assume that it realises that the image it sees is not a real banana—it has simply stopped looking because that activity has never been rewarded by success. Quite plausibly, the subject takes a mirror image as an actual, physical object behind the mirror, but just one that it is not able to retrieve. None of this requires any conceptual understanding; the subject may simply be using the mirror image as a natural sign.

For most mirror images, as just argued, we simply cannot assume the subject takes them as symbols—even after they have stopped searching for them behind the mirror. But now, this is where MSR comes in. The one case for which we can be sure that the subject has taken the image as a representation rather than an actual object behind the mirror is when it has recognised the image as itself. Mirror images are not actual objects behind the mirror, but it is likely that subjects that are not capable of MSR continue to take them to be so. Monkeys seeing their own mirror images continue to take them to be conspecifics. But a chimpanzee cannot be taking its mirror image as a physical object behind the mirror. This is because (as argued earlier) it has definitively taken the image as itself. All organisms can differentiate the physical self from non-self, perhaps mainly by proprioception. Since the image is not (part of) the physical self but is still understood to be the self, the most plausible interpretation is that it is representational of the self (in effect a symbol). A stark demonstration of this interpretation is that when a chimp looks at the distal image in a mirror, it reaches for its own forehead to touch the mark. It understands that what it sees in the mirror is not a physical object behind the mirror (a conspecific), but a representation of itself (its own body) onto which a mark has been applied.

At this point objectors might propose that this still unnecessarily ascribes too high a level of cognition to the chimpanzee; it might still just be (non-conceptually) interacting with a natural sign (or, put another way, detecting the mirror’s affordance to locate marks on its forehead). The analysis provided in the first part of this paper strongly argues in favour of the validity of the mark test as conclusive evidence of self-recognition, and I have also argued that an MSR subject cannot take its mirror image as a physical extension of its body. Nevertheless, the argument would go, the subject does not recognise the image as a representation but just learns a simple association between the mark on the image and the existence of a similar unseen mark on its own forehead. But this interpretation is implausible: one must question how such an association could be formed in the first place (that is, between the seen mirror image of the mark and the unseen actual mark on the forehead). Compare this with the association that could be formed with images of non-self objects as in the earlier banana example. Furthermore, on the associative account, one must also question why chimpanzees learn this association while the morphologically similar monkeys never do. It is more likely that the phylogenetically more advanced chimpanzees are exercising a more developed cognitive capacity, one that allows them to grasp mirror images as representations. The human example supports this view: symbol-mindedness is a capacity acquired as children become more cognitively developed.

It is interesting to compare the results of experiments on infant MSR with those of DeLoache’s (2004) experiments on infant symbol-mindedness. As mentioned earlier, the age for the onset of infant symbol-mindedness (about 18 months) is similar—or perhaps prior to—that for infant MSR (18–24 months). This is consistent with the claim that MSR is an indication of symbol-mindedness. More specific experiments are needed to verify this. If my analysis is correct, infants cannot pass the mark test if they are not already symbol-minded. Therefore, the same infant subject should pass tests of symbol-mindedness at an earlier age than when they pass the mark test. I suggest longitudinal experiments be carried out to establish that symbol-mindedness arises prior to or concurrent with MSR in the same subjects.


I have used the ‘Indirect Interpretation’ to argue that MSR does provide sufficient evidence of self-consciousness. The argument is based on MSR being a demonstration that the subject has essentially taken its mirror image as a symbol representing its own body and as such is symbol-minded. Being symbol minded means being concept possessing, which, according to the Concept Possession Hypothesis, is sufficient evidence of self-consciousness.

Non-human animals that have so far convincingly demonstrated MSR are the great apes (save wild-born gorillas) while of the non-primates there are promising signs of MSR in dolphins, elephants, and magpies. When MSR is confirmed beyond reasonable doubt, these animals must be considered self-conscious organisms in some fundamental yet highly significant way. Exactly how significant remains for further discovery and debate, however, we can get some appreciation of it by comparing with a known standard: humans. Human infants demonstrate MSR at around the 18–24 month age range, at which age it is probably safe to say they have a sense of their own existence. However, by this age they may have not yet developed other significant cognitive capacities such as episodic memory and theory of mind. It appears not unreasonable to equate the cognitive capacities of animals that have demonstrated MSR with those of (at least) 18–24 month old children.


  1. 1.

    I use the terms self-awareness and self-consciousness interchangeably in this paper.

  2. 2.

    See, for example, Gibson (1979): “One perceives the environment and coperceives oneself.” (p. 126).

  3. 3.

    A report that cottontop tamarins passed the mark test (Hauser et al. 1995) failed to replicate (Hauser et al. 2001) and is questionable. A claim has been made that rhesus monkeys possess MSR despite failing the mark test (Rajala et al. 2010), however a critical examination of the results reveals them to be at best inconclusive (Anderson and Gallup 2011).

  4. 4.

    Gallup et al. (2011), though impressed with the magpie research in particular, caution against accepting any of these results too readily, especially as in the dolphin and elephant cases only one individual convincingly passed and the results are yet to be replicated.

  5. 5.

    Shillito et al. (1999) conducted mark tests on two adult lowland gorillas at the Washington D.C Zoo to test this hypothesis. They used angled mirrors such that the gorillas could see their own reflections without making eye contact. The results were negative: indicating that gaze aversion probably was not the explanation for lack of MSR. The issue remains, however, that some other as yet unknown species-relevant properties or behaviour may be preventing a positive mark test result.

  6. 6.

    Schilhab (2004) uses this expression to describe an organism’s ability to distinguish its own body sensations from externally sourced sensory input, which does not relate to a mental category. In other words, this ability, common throughout the animal kingdom, need not imply the capacity for conceptual thought, but potentially could explain MSR (as discussed below).

  7. 7.

    Three previously mirror-naïve adult male pigeons were first trained to peck at various dots on the body and around the cage. A mirror was used in the case of dots around the cage, in which the dots were flashed when they could only be visible in the mirror—the pigeons were rewarded for pecking at the positions where the dot had been. Then a test was conducted whereby a dot was placed on the pigeon’s breast and a bib was used to render the dot invisible to the pigeon except in a mirror. If the pigeon bent its head forward even slightly the bib slid down the breast and covered the dot. The pigeon did not peck at the dot unless a mirror was present to allow the pigeon to see it, in which case the pigeon pecked at the position on the bib that corresponded to the position of the covered dot. This result was taken as a passing of the mark test. The experimenters did not conclude that pigeons are self-recognising; rather, they claimed that the mark test was insufficient evidence of self-recognition.

  8. 8.

    I think the term 'body concept' is an unfortunate choice by Heyes as no concept is involved. What she refers to is better described as 'body schema' (Gallagher 1995).

  9. 9.

    Used here in the ordinary sense; not to be confused with Brentano’s (1874/1973) ‘intentionality’ or ‘aboutness’.

  10. 10.

    DeLoache (personal communication) has indicated that her definition is meant to cover both the creator and perceiver perspectives. Here I emphasise the perspective of the perceiver since this is crucial for my argument.

  11. 11.

    In the Gibsonian sense (Gibson 1979).

  12. 12.

    This example highlights the point made earlier that we need to consider symbol usage from the perspective of the perceiver and not just the symbol creator, for the squiggle may have been unintentional. For instance, a squiggle resembling the Greek letter pi may have been nothing more than an infant’s doodle, yet later taken as pi by an adult.

  13. 13.

    Perner (1991), perhaps, would rather call the scale models ‘analogues’. Even so, these remain examples of symbols according to the way DeLoache and I have characterised them. That is, the subjects treat them as symbols by regarding what they see as standing in for something else.

  14. 14.

    DeLoache (personal communication) has indicated that, to her, mirror images are not symbolic as they do not stand for anything as other symbolic objects do. However, I argue that it depends on how the subject takes the image. Consider a picture hanging on a wall: DeLoache would agree that children above a certain age treat this as a symbol. But now substitute the picture for a mirror (at a certain angle) showing a similar image. Subjects might treat what they see just the same as for the picture—i.e., as a symbol. DeLoache herself used a similar trick as mentioned in the previous section: the ‘window’ view increased the symbolic nature of the scale model. A mirror should have much the same effect. Further, for reasons I spell out below, when the image is specifically of the self, as in a case of MSR, the subject must be treating the image as a symbol.



I am grateful to the following persons for their comments on earlier versions of this paper: John Sutton, Peter Menzies, Gordon Gallup Jr., Glenn Carruthers, Liz Schier, Mitch Parsell and Stuart Palmer. I thank Judy DeLoache for her responses to my questions on symbol-mindedness. I also thank an anonymous reviewer and Kim Sterelny for helpful suggestions for improvements to the original submission.


  1. Amsterdam B (1972) Mirror self-image reactions before age two. Dev Psychobiol 5:297–305CrossRefGoogle Scholar
  2. Anderson JR, Gallup GG (2011) Do rhesus monkeys recognize themselves in mirrors? Am J Primatol 73:1–4Google Scholar
  3. Bermũdez JoséLuis (1998) The paradox of self-consciousness. MIT Press, CambridgeGoogle Scholar
  4. Bertenthal B, Fischer K (1978) Development of self-recognition in the infant. Dev Psychol 14(1):44–50CrossRefGoogle Scholar
  5. Bigelow A (1981) The correspondence between self- and image-movement as a cue to self-recognition for young children. J Genet Psychol 139:11–26CrossRefGoogle Scholar
  6. Brentano F (1874/1973) Psychology from an Empirical Standpoint. In: Kraus O, McAlister LL (eds) (trans: Rancurello AC, Terrell DB, McAlister LL). Routledge, Kegan Paul, LondonGoogle Scholar
  7. Deacon T (1997) The symbolic species. Allen Lane, The Penguin Press, LondonGoogle Scholar
  8. DeLoache JS (2004) Becoming symbol-minded. Trends Cogn Sci 8(2):66–70CrossRefGoogle Scholar
  9. de Veer M, van den Bos R (1999) A critical review of methodology and interpretation of mirror self-recognition research in nonhuman primates. Anim Behav 58:459–468CrossRefGoogle Scholar
  10. Epstein R, Lanza RP, Skinner BF (1981) “Self-awareness” in the pigeon. Science 212:695–696CrossRefGoogle Scholar
  11. Gallagher S (1995) Body schema and intentionality. In: Bermũdez J, Marcel J, Eilan N (eds) The body and the self, a bradford book. MIT Press, Cambridge, pp 225–244Google Scholar
  12. Gallup GG Jr (1970) Chimpanzees: self-recognition. Science 167:86–87CrossRefGoogle Scholar
  13. Gallup GG Jr (1975) Towards an operational definition of self-awareness. In: Tuttle RH (ed) Socioecology and psychology of primates. Mouton, The HagueGoogle Scholar
  14. Gallup GG Jr (1998) Can animals empathize? Yes. Scientific American Presents, Exploring Intelligence, Jan 1998Google Scholar
  15. Gallup GG Jr, Suarez SD (1986) Self-awareness and the emergence of mind in humans and other primates. In: Suls J, Greenwald A (eds.) Psychological perspectives on the self, vol 3. Erlbaum, Hillsdale, pp 3–26Google Scholar
  16. Gallup GG, Anderson JR, Platek SM (2011) Self-recognition. In: Gallagher S (ed) The Oxford handbook of the self. Oxford University Press, New York, pp 80–110Google Scholar
  17. Gibson JJ (1979) The ecological approach to visual perception. Houghton Mifflin, BostonGoogle Scholar
  18. Grice HP (1957) Meaning. Philos Rev 66(3):377–388CrossRefGoogle Scholar
  19. Hauser MD, Kralik J, Botto-Mahan C, Garrett M, Oser J (1995) Self-recognition in primates: phylogeny and the salience of species-typical features. Proc Natl Acad Sci USA 92:10811–10814CrossRefGoogle Scholar
  20. Hauser M, Miller C, Liu K, Gupta R (2001) Cotton-top tamarins (Saguinus oedipus) fail to show mirror-guided self-exploration. Am J Primatol 53(3):131–137CrossRefGoogle Scholar
  21. Heyes C (1998) Theory of mind in nonhuman primates. Behav Brain Sci 21(1):101–148Google Scholar
  22. Lewis M, Brooks-Gunn J (1979) Social cognition and the acquisition of self. Plenum Press, New YorkCrossRefGoogle Scholar
  23. Marino L, Reiss D, Gallup GG (1994) Mirror self-recognition in bottlenose dolphins: implications for comparative investigations of highly dissimilar species, chap 25. In: Parker S, Mitchell R, Boccia M (eds.) Self-awareness in animals and humans: developmental perspectives. Cambridge University Press, New YorkGoogle Scholar
  24. Marten K, Psarakos S (1994) Evidence of self-awareness in the bottlenose dolphin (Tursiops truncatus). In: Parker S, Mitchell R, Boccia M (eds.) Self-awareness in animals and humans: developmental perspectives. Cambridge University Press, New York, pp 361–379Google Scholar
  25. Mitchell RW (1997a) Kinesthetic-visual matching and the self-concept as explanations of mirror-self-recognition. J Theory Soc Behav 27(1):17–39CrossRefGoogle Scholar
  26. Mitchell RW (1997b) A comparison of the self-awareness and kinesthetic-visual matching theories of self-recognition: autistic children and others. In: Snodgrass J, Thompson L (eds) The self across psychology, vol 818. Annals of the New York Academy of Sciences, New York, pp 39–63Google Scholar
  27. Patterson FG, Cohn RH (1994) Self-recognition and self-awareness in lowland gorillas. In: Parker S, Mitchell R, Boccia M (eds) Self-awareness in animals and humans: developmental perspectives. Cambridge University Press, New York, pp 273–290Google Scholar
  28. Perner Josef (1991) Understanding the representational mind. A Bradford book. MIT Press, CambridgeGoogle Scholar
  29. Plotnik JM, de Waal FBM, Reiss D (2006) Self-recognition in an Asian elephant. Proc Nat Acad Sci 103(45):17053–17057CrossRefGoogle Scholar
  30. Povinelli DJ (1989) Failure to find self-recognition in Asian elephants (Elephas maximus) in contrast to their use of mirror cues to discover hidden food. J Comp Psychol 103:122–131CrossRefGoogle Scholar
  31. Povinelli DJ, Rulf AB, Landau KR, Bierschwale DT (1993) Self-recognition in Chimpanzees: distribution, ontogeny, and patterns of emergence. J Comp Psychol 107:347–372CrossRefGoogle Scholar
  32. Prior H, Schwarz A, Güntürkün O (2008) Mirror-induced behavior in the magpie (Pica pica): evidence of self-recognition. PLoS Biol 6(8):e202. doi:10.1371/journal.pbio.0060202
  33. Rajala AZ, Reininger KR, Lancaster KM, Populin LC (2010) Rhesus monkeys (Macaca mulatta) do recognize themselves in the mirror: implications for the evolution of self-recognition. PLoS ONE 5(9):e12865. doi:10.1371/journal.pone.0012865 CrossRefGoogle Scholar
  34. Savanah S (2011) The concept possession hypothesis of self-consciousness. Conscious Cogn. doi:10.1016/j.concog.2011.02.019
  35. Schilhab TSS (2004) What mirror self-recognition in nonhumans can tell us about aspects of self. Biol Philos 19:111–126CrossRefGoogle Scholar
  36. Schulman AH, Kaplowitz C (1977) Mirror-image response during the first two years of life. Dev Psychobiol 10(3):133–142CrossRefGoogle Scholar
  37. Shillito DJ, Gallup GG, Beck BB (1999) Factors affecting mirror behaviour in western lowland gorillas. Anim Behav 57:999–1004CrossRefGoogle Scholar
  38. Suddendorf T, Corballis M (1997) Mental time travel and the evolution of the human mind. Genetic Soc General Psychol Monogr 123:133–167Google Scholar
  39. Swartz KB, Evans S (1994) Social and cognitive factors in chimpanzee and gorilla mirror behaviour and self-recognition. In: Parker S, Mitchell R, Boccia M (eds) Self-awareness in animals and humans: developmental perspectives. Cambridge University Press, New York, pp 189–206Google Scholar
  40. Thompson RKR, Contie CL (1994) Further reflections on mirror usage by pigeons: lessons from Winnie-The-Pooh and Pinnochio Too. In: Parker S, Mitchell R, Boccia M (eds) Self-awareness in animals and humans: developmental perspectives. Cambridge University Press, New York, pp 392–409Google Scholar
  41. Wellman HM, Cross D, Watson J (2001) Meta-analysis of theory-of-mind development: the truth about false belief. Child Dev 72(3):655–684CrossRefGoogle Scholar

Copyright information

© Springer Science+Business Media B.V. 2012

Authors and Affiliations

  1. 1.ARC Centre of Excellence in Cognition and its Disorders (CCD)Macquarie UniversitySydneyAustralia

Personalised recommendations