Recently, non-human empathy has also become the focus of attention spurred on, to some extent, by the influential paper of Preston & de Waal [3]. Interestingly, these authors outlined a sequence of progressively complex levels of empathy across animals that parallels the development of empathy in young humans. According to their proposal, early on in development, an automatic mechanism causes the state of an individual who experiences an emotional state (the object) to elicit a relevant, or similar, state in a perceiver (the subject)—the more interrelated the subject and the object, the more the subject will attend to the object, and the readier the emotional reaction. This limits empathy to emotional contagion because the subject cannot distinguish their own emotions from others' and has no control over emotional reactivity. Later on, as cognitive abilities are progressively layered upon automatic emotional activation, empathy goes beyond emotional contagion and develops, first into sympathetic concern (implying that the subject has the ability to discriminate between internally and externally generated emotions) and then into empathic perspective taking (implying that the subject can form a cognitive representation of the object's emotional situation and needs, similarly to representing non-emotional states) [3]. In humans, the tendency to comfort others who are expressing emotional distress (probably grounded in sympathetic concern) develops around nine months of age and is followed by the emergence of spontaneous helping (probably driven by empathic perspective taking) around 14 months of age [4].

Although observable examples of emotional contagion in non-human animals have long been mentioned in the literature, only recently has experimental evidence started to emerge. Langford et al. [5] showed that mice intensify their responses to pain when perceiving cage mates, but not strangers, in pain; Parr [6] found that chimpanzees respond to video images of socially close conspecifics in distress by a vicarious emotional response and Davila Ross et al. [7] reported involuntary facial mimicry in orang-utans. Note that facial mimicry has been linked to emotional contagion in humans, although it is still being discussed whether emotional conveyance results from mimicry or vice versa.

As suggestive evidence that sympathetic concern must also have emerged in a pre-human basis, some species (e.g. [8,9]) show consolation-like behaviours, that is, post-conflict affiliative interactions directed from a third party to the recipient of aggression and assumed to have a stress-alleviating function. Similarly, the basic forms of instrumental help demonstrated in chimpanzees have been associated to the capacity for empathic perspective taking ([4] but see [10]). Curiously, chimpanzees that grow up interacting with people not only appear to console and help conspecifics but also humans [4].

What about other animals living in close contact with humans, such as domestic dogs? Although the special relationship that dogs have developed with humans has been attracting increasing research interest, and a profound level of mutual understanding and shared emotion has already been suggested [11], no stringent proposal can be found in the literature on whether, and at what level, dogs may empathize with humans. This is even more remarkable in view of the strong possibility of convergent evolution between dogs and humans—sustained by the fact that dogs possess certain human-like social skills that non-human apes do not [12].

Although dogs tend to fail the mirror self-recognition test [23] some authors, who have questioned the use of a single technique, based solely on visual cues, as the only valid test of self-awareness, claim that it might not indicate the absence of self-awareness required for empathic levels beyond emotional contagion (e.g. [11]). Indeed, a study showing that pets, namely dogs, behave as ‘upset’ as children when exposed to familiar people faking distress, strongly suggests ‘sympathetic concern’ [24]. Also, it has been reported that untrained dogs may be sensitive to human emergencies and may act appropriately to summon help [18], which, if true, suggests empathic perspective taking. The one study testing this idea in two experiments (in one dogs' owners feigned a heart attack and in another they experienced an accident in which a bookcase fell on them and pinned them to the floor) concluded, however, that dogs did not seem to understand the nature of the emergency or the need to obtain help [25]. We believe, however, that this does not discard the possibility for empathic perspective taking, as the emergency scenarios might not have been sufficiently dramatic or realistic to be interpreted by the dogs as a real emergency. Also, olfactory cues may be important for dogs to accurately assess and respond to others' emotions. For instance, pheromones produced by a person suffering the pain and the stress of a real emergency may contribute to dogs' sense of a real emergency [11].

Since no additional studies have been conducted, one way to further discuss dogs' empathic potential is to employ research on mental state attribution (as in Koski & Sterck [26]). Studies show that when faced with a piece of forbidden food, dogs are quicker to take it if the experimenter cannot see them [27]. Contrastingly, dogs preferentially beg from a person that can see them [27]. While, to some authors, such findings are indicative of a simple foraging strategy based on the greatest chance of reinforcement (e.g. [28]), to others (including ourselves), they suggest that dogs may have the capacity to infer what humans see, know and feel (e.g. [27]). Likewise, there is a seemingly intractable debate around studies showing that dogs can use human pointing gestures to find hidden food (e.g. [29]). While some authors stress that this capacity can be accounted for by means of straightforward associative learning (e.g. [30]), others (including ourselves) defend the idea that dogs may actually understand that the person knows the location of food and is trying to convey this information to them (e.g. [27]).