Chapter 3 - Animal Emotions and Subjective States

Back to Main Page Previous chapter Next chapter Conclusions References Junk

Chapter Outline

In this chapter, I discuss animal emotions and other feelings such as pain. I deliberately confine myself to "basic" emotions such as fear and anger, whose occurrence in at least some non-human animals is fairly uncontroversial. Thus I ignore emotions such as jealousy, envy and Schadenfreude, which involve high-level cognitive processes. The aim of this chapter is to answer four key questions regarding animal emotions:

(1) What are the cognitive pre-requisites of animal emotions?

(2) What are animal emotions "about", and what is each basic kind of animal emotion about? (The problem of intentionality.)

(3) How do we identify basic emotions in animals, and which animals can be said to have them?

(4) In which animals are basic emotions (and other feelings) phenomenally conscious? (The Distribution Question.)

I propose that the reality of emotions in at least some non-human animals should be taken as a methodological "given". I then attempt to identify the distinctive features of animal emotions. The literature on the emotions is vast, so I begin with a representative list of features of human emotions about which a solid consensus exists, and attempt to apply these to animals.

One of these features is that human emotions are typically (if not always) conscious. However, the question of whether phenomenal consciousness is an essential feature of animal emotions sparks debate, not only among philosophers, but even between different scientific disciplines. Within some disciplines, the very concept of an unconscious emotion is an oxymoron, while in others, the notion that consciousness is merely the icing on the cake of emotion (LeDoux, 1998, p. 302) is not uncommon (but see Panksepp, 2003). At the outset, I adopt a broad definition of animal emotions: animal emotions, whether conscious or not, must be mental states which are amenable to scientific investigation. Thus I reject both subjectivist theories, which envisage emotions as essentially private inner states and thereby exclude them from the domain of science, and behaviourist accounts, which define emotions purely in terms of outward behavioural dispositions and thus fail to explain why emotions should be treated as mental states. I address the criteria for animal consciousness later in the chapter, but I allow for the possibility that a large number of animals may have genuine emotions, but live their entire lives at the unconscious or preconscious level. Even if animal emotions are not conscious, however, they must satisfy certain minimum cognitive requirements to qualify as mental states.

Intentionality is another problem I address in this chapter. Emotions typically have intentional objects: they are "about" something. However, some emotions, such as an ill-defined feeling of depression, appear to lack any kind of object whatsoever. Any adequate theory of animal and human emotions must account for both their "aboutness" and their lack of "aboutness". The solution I endorse here is that whereas individual emotions may lack an object, the kind of emotion they instantiate must always be "about" something, which I call the generic intentional object of the emotion. Some authors (de Sousa, 2003) describe the object of each kind of emotion as a formal object which serves as a yardstick for judging the appropriateness of an animal's emotional response on a specific occasion. I argue that for animals that lack language, the only standard for judging the appropriateness of its responses is the animal's own well-being. Accordingly, I envisage the generic intentional object of each kind of emotion not only as a formal object, but also as a teleological object that explains what the emotion is "for" and how it helps the animal to survive and/or flourish.

I address the first of my four questions regarding animal emotions by examining the relationship between emotion and cognition (discussed in chapter two), and conclude from the available neurological and behavioural evidence that although emotions presuppose cognitive states, they are fundamentally distinct from them. Cognitivist and appraisal theories of the emotions, which construe emotions in cognitive terms - e.g. as judgements or evaluations - are therefore inadequate.

If an animal acts intentionally, its behaviour is appropriately explained using the mentalistic agent-centred intentional stance discussed in chapter two: we invoke the animal's relevant beliefs and motivations to account for its actions. In chapter two, we described these motivations as "desires", but other emotions (e.g. fear or anger) are capable of motivating animal agents equally well. If we define an emotion as any internal state that is capable of motivating an animal in intentional agency, it follows that animal emotions, even if unconscious, can be considered as minimal mental states. I propose that the animals to which we can ascribe emotions are simply those whose behaviour is best described in terms of the agent-centred intentional stance.

Because the agent-centred intentional stance presupposes beliefs, emotions can only be identified in animals that are capable of forming beliefs. However, I argue that these beliefs need not be propositional: strategic beliefs (whether conscious or unconscious) about the best way to attain some goal suffice to manifest emotion. The preference beliefs described by Regan, on the other hand, appear to lack the requisite cognitive structure for proper beliefs.

Some philosophers, however, argue that even "basic" emotions such as fear, anger and desire require their possessors to be capable of certain feats that presuppose the use of language. If they are right, then at best, animals that lack language can only be said to have these emotions in an analogical sense, as part of some "attentuated language-game" (Leahy, 1994, p. 136). I argue that on the contrary, there are certain core cognitive requirements for these emotions, which non-human animals can and do satisfy.

I propose that the second and third of my four philosophical questions regarding animal emotions can be resolved by understanding how they are realised within animals' brains. A brain may sound like an unpromising place to look for "aboutness", and I wish to make it clear that I am not espousing any theory which reduces emotions to brain states. My point is rather that the brain embodies the evolutionary history of the emotions, and accounts for their intentionality, at least in generic terms. I make use of Panksepp's (1998) neurophysiological account of animal emotions, according to which the basic emotions in animals arose as different patterns of responding to the various kinds of environmental challenges that their ancestors had to confront. Put simply, the "generic intentional object" of each kind of emotion - what it is "about" - is the environmental challenge it evolved to meet. The brains of human beings and other mammals contain several neurologically distinct, well-described "emotion systems" (Panksepp, 1998), each regulating its own kind of emotional response, which reflect the way in which these animals' brains evolved to respond to these challenges. I claim that the emotion systems in these animals' brains are, generically, "about" the environmental challenges they evolved to meet, in two robust senses. First, the environmental challenges have (over millions of years) caused the evolution of behavioural capacities which are directed at them. Second, each of these emotion systems is capable of motivating intentional acts (which require a mentalistic description), we can say that the environmental challenges have caused the evolution of mental capacities which are directed at them. I contend that rival "feedback" theories of the emotions (LeDoux, 1996; Damasio, 2003) which refine the original James-Lange theory on which they are based, and reduce emotions to internal bodily states, fail to explain the intentionality of the emotions, despite the fruitful scientific research they have generated.

Studies of animals' brains can thus reveal (i) the original motivational context of each kind of emotion, (ii) the "generic intentional object" of each kind of emotion, (iii) the proper taxonomy of the basic animal emotions, (iv) the evolutionary history of these emotions, and hence (v) which animals possess these emotions, at least at an unconscious level. I also formulate criteria for the kind of behaviour which is sufficient for us to identify all of the basic kinds of emotions in animals that have been described by researchers.

Turning to the final problem of how we identify phenomenal consciousness in animals, I contend that most philosophers have been looking for it in the wrong places. Animals' reactions alone cannot unambiguously manifest consciousness on their part, and even intentional agency is not a sufficient condition for its occurrence. However, I argue that a special kind of behaviour by animal agents, namely hedonic behaviour, does require a first-person account. Utility theory allows us to apply first-person concepts to non-human animals (Dawkins, 1994; Berridge, 2001, 2003a, 2003b), but I propose that phenomenal consciousness in animals can be identified unambiguously by affective distortions - by which I mean the kind of emotional mis-judgements that only a creature precoccupied with its own welfare would make. Recent research by Cabanac (1999, 2002, 2004) and Panksepp (1998, 2003) points to ways of measuring these affective distortions.

My specific proposal is that because consciousness can alter human beings' perceptions of risk (Slovic, Finucane, Peters and MacGregor, 2003), it should be possible to identify erratic behaviour in animals blinded by their feelings, which reflects their faulty risk assessments of events in their environment.

The available behavioural evidence suggests very strongly that reptiles, birds and mammals do indeed possess subjective states or phenomenal consciousness. (The case for affective states in other animals is much more problematic.) This finding clashes with the opinion of many neurologists, that even simple consciousness is confined to mammals. However, other neurologists do not share their view, argue that the grounds commonly adduced for this mammalocentric view are inconclusive.

I also suggest that animals' ability to control their emotions - especially fear, anger and desire - (so-called "cortical over-ride") is a sign of consciousness on their part. Research in this area is still in its infancy.

Even if the majority of non-human animals lack phenomenal consciousness, they still matter. To claim that only beings with subjective states are ethically significant is a form of moral myopia. Most unconscious animals still have first-order desires that can be frustrated (Carruthers, 2004). Additionally, they, like other living things, have interests that can be harmed in measurable ways by stressful events.

Finally, I examine Midgley's argument that human beings and companion animals are emotionally symbiotic, and propose that the way in which ordinary people usually identify emotions in human beings is inseparable from the way they identify emotions in animals.


Emotions, and feelings in general, are often used to drive an ontological wedge between a subjects, which possess some kind of mental life, and are described using first-person terminology, and mindless objects, which are more appropriately described by third-person terminology. The argument of this thesis is that the "I versus "it" (first-person vs. third-person) divide is dwarfed by the gulf between what I shall call the "it" vs. "they" (third-person-singular vs. third-person-plural) divide: that is, between organisms, which behave as co-ordinated wholes, and artefacts, which lack intrinsic unity and behave as assemblages. As we argued in chapter one, the unity of a living thing is reflected in the nested functionality of its parts, which subserve the good of the whole.

Insofar as they pursue those things which are in their interests, all life-forms have something analogous to emotions. The humble E. coli bacterium, which is attracted to lactose but averse to benzoate, and which prefers glucose to lactose, lacks "feelings", but it does possess a feature of emotions that is much more important than mere subjectivity: it can be moved towards or away from some object that in which it has an interest.

Methodological presuppositions

At the outset of this enquiry into animal emotions, I shall attempt to set forth my background assumptions.

(a) Animal Emotions are Real

In this chapter, I shall take it as a "given" that at least some animals - in particular, some mammals and birds - have emotions. (Whether these emotions are phenomenologically conscious is a matter I will discuss later.) This decision can be defended on linguistic and psychological grounds, even before we attempt to define emotions.

Animal emotions are a linguistic "given". Ceasing to recognise emotions in mammals and birds would do violence to the way we talk about emotions, because these animals often serve as primary referents for words describing emotions. If I were teaching a child the meaning of anger, I could do no better than to point to a hissing, snarling cat and say: "This is what anger looks like." Here, the angry cat functions as an exemplar. Additionally, the fact that all human languages possess an abundance of animal metaphors for emotions makes it impossible to divorce the meaning of these emotions from animal behaviour.

Emotions in some mammals and birds are also a psychological "given". Recent research has established that humans and their companion animals are emotionally symbiotic: each needs the other to thrive emotionally. Put simply, our own feelings feed off those of our pets, and therefore cannot be defined apart from them. Midgley (1993) summarises the evidence:

Pet therapy programs... caused a disturbance as soon as they began to be reported some years back, because they call in question the human race's boasted independence and autonomy. Could the undignified suggestion that people actually needed and welcomed the gifts of these unworthy, alien beings really be true? Repeated investigations have confirmed that it is indeed true... The idea that pet-keeping was some sort of pointless aberration, a meaningless, sentimental, perverse fad of the affluent West, can scarcely now be defended. The therapeutic effectiveness of pet-keeping, along with anthropological data showing that pets have been kept in all kinds of human societies, is gradually forcing attention to the meaning of such customs.

I am not claiming here that all pets have emotions simply because their owners bond with them. Rather, my point is that the vast majority of pet owners find that their own emotional well-being is enhanced by interacting with their pets. The most reasonable explanation of this fact is that most pet owners' feelings are reciprocated to some degree by their pets.

It should not be assumed, however, that animal emotions share all of the properties of human emotions.

(b) General Properties of Emotions

Emotions are the subject of long-standing philosophical controversy. However, one recent positive development is that philosophers have come to agree on several key points, highlighted by de Sousa (2003):

(Ronald de Sousa on a good day)(and on not such a good day)
Photos courtesy of University of Toronto.

A broad consensus has emerged on what we might call adequacy conditions on any theory of emotion. An acceptable philosophical theory of emotions should be able to account at least for the following nine characteristics...

  • emotions are typically conscious phenomena; yet
  • they typically involve more pervasive bodily manifestations than other conscious states;
  • they vary along a number of dimensions: intensity, type and range of intentional objects, etc;
  • they are reputed to be antagonists of rationality; but also
  • they play an indispensable role in determining the quality of life;
  • they contribute crucially to defining our ends and priorities;
  • they play a crucial role in the regulation of social life;
  • they protect us from an excessively slavish devotion to narrow conceptions of rationality;
  • they have a central place in moral education and the moral life (de Sousa, 2003, italics mine).
I have argued that emotions in at least some mammals and birds must be genuine. However, to impute rationality, morality or language to all of these animals would be clearly anthropomorphic. I therefore conclude that any features of emotions in de Sousa's list that make explicit mention of rationality (i.e. the third and eighth items), morality (the last item) or language (the fourth item, which appears to rely on the subject's verbally reported "quality of life") cannot be necessary features of emotions per se, although certain emotions (e.g. remorse and pride) obviously require them. (I shall respond below to arguments claiming that the attribution of any emotion to an individual presupposes that it is rational or can use language.)

That leaves five features of interest to us:

With the possible exception of the first property, all of these features of human emotions can also be said to apply to animal emotions. There are, however, good reasons for querying the notion that animal emotions are typically conscious.

(c) Animal emotions, whether conscious or not, are mental states amenable to scientific investigation

While human emotions are indeed typically (if not always) conscious, the question of whether phenomenal consciousness is an essential feature of animal emotions remains a contentious one among philosophers and scientists. The consensus view of most psychologists is that the very concept of an unconscious emotion is an oxymoron (Berridge, 2003c).

Joseph LeDoux. Photo courtesy of New York University Center for Neural Science.

There are several reasons for querying this consensus. First, within the discipline of neurology, unconscious emotions are not regarded as an anomaly. For instance, LeDoux (1998) argues that emotions are adaptive because of the way they work within the body, not because of the way they consciously feel:

Emotions evolved not as conscious feelings... but as brain states and bodily responses. The brain states and bodily responses are the fundamental facts of an emotion, and the conscious feelings are the frills that have added icing to the emotional cake... (1998, p. 302).

Second, research has conclusively demonstrated that normal human beings are capable of acting on emotions that they are not consciously aware of. Berridge (2003b) describes a study in which subliminal exposure to happy or angry faces - which the subjects were later unable to recall - had a dramatic influence on their liking for a fruit beverage, how much they they wanted to consume, and how much they would be willing to pay for it if it were sold (i.e. its monetary value). Berridge refers to these phenomena as "nonconscious 'liking' and 'wanting'".

Third, emotional facial reactions are known to occur even in human infants who, according to the American Medical Association and American Academy of Neurology (Shewmon, 1999), are congenitally incapable of consciousness, as their brains lack cerebral hemispheres and possess only a functioning brainstem. Thus the parts which are supposed to mediate cognition, the social emotions and the evaluation of the emotional significance of stimuli are missing. Research cited by Berridge (2003b) shows that anencephalic infants display positive facial reactions (e.g. lip sucking, smiles) to sweet tastes and negative reactions (e.g. gapes, nose wrinkling) to bitter tastes. Berridge concludes that the core of the facial reaction to sweet and bitter tastes is a nonconscious one. On the other hand, Shewmon (1999) argues strongly that these infants do indeed possess a rudimentary consciousness.

To avoid biasing my methodology, I shall adopt a broad definition of animal emotions: animal emotions may or may not be conscious, but at a minimum, they must be mental states which are amenable to scientific investigation. Whatever they are, emotions must in some way be psychological states as well as physical ones, for that is how the word "emotion" is used in our language. Also, the fact that there are several well-established scientific disciplines that deal with emotions attests to their amenability to scientific investigation.

Ludwig Wittgenstein.

Many people, on the other hand, adhere to a "subjectivist" account of feelings, claiming that emotions and other feelings (such as pain) designate private sensations. These private inner states are said to be inaccessible to scientific investigation. However, Leahy (1994, p. 124) argues that this conviction arises from focusing exclusively on isolated feelings such as toothaches. On this point, Leahy cites a supporting comment from Wittgenstein and comments:

'A main cause of philosophical disease - a one-sided diet: one nourishes one's thinking with only one kind of example' (PI 593). The question we must ask is, 'How does a human being learn the meaning of the names of sensations? - of the word "pain" for example' (1994, p. 124).
Similar observations apply to animal emotions (Leahy, 1994, p. 128). The key point here is that if emotions were inherently private, it would be impossible for us to learn the meanings of the words used to designate them. But in fact, emotions are often defined ostensibly in a public context: "This is what anger looks like". We may even use animals as paradigm cases of these emotions.

The subjectivist account of emotions also fails to explain how outward bodily states are able to manifest these allegedly private inner feelings, or why there should be different kinds of emotions, or what makes private inner states possess the property of "aboutness", or how such essentially private states can have an important social role.

At the other extreme, I reject any accounts of emotions which define them purely in terms of outward behavioural dispositions and thus fail to explain why emotions should be treated as mental states at all. Dispositions do not require a mentalistic explanation: as we saw in chapter two, a mind-neutral goal-centred intentional stance can be used instead.

Thus I propose to re-write the first of our five selected features of animal emotions as follows:

Animal emotions, whether conscious or not, are mental states which are amenable to scientific investigation.

We therefore have to consider the possibility that there may be animals who have non-phenomenal emotions - emotions without any conscious feelings whatsoever.

(d) Animal emotions have intentional objects

Daniel Dennett. Photo courtesy of University of California.

Much of the argumentation in this chapter rests upon the assumption, which I defended in chapter two, that Daniel Dennett's intentional stance can be applied to all kinds of mental states, including emotions. In other words, emotions, like beliefs, desires and other mental states, have to be "about" something: they require intentional objects. As Leahy (1994) puts it:

[S]omeone who announced that they were afraid, or hoping, or feeling, but seemed unable to tell us any more would indeed be a source of perplexity. A person should be able to say what they are afraid of, hoping for, or feeling guilty about... The emotions, or most of them, are said to be directed, and to have objects (1994, p. 130).

Even individuals (including animals) who cannot tell us what the objects of their emotions are can still be said to have directed feelings:

The targets of an animal's fear or anger are usually clear enough and it is perfectly natural and necessary to speak of them... Herring gulls with wings held 'akimbo' are poised to hurl themsleves at an intruder" (Leahy, 1994, p. 135).

There are, however, situations where we experience emotions without any apparent object. Sometimes, for instance when we experience an undefined feeling of fear, there may be an object but we may not be consciously aware of what it is, because the brain registers it at a subconscious level. Other emotions, such as an ill-defined feeling of depression, resist characterisation in intentionalist terms, as they lack any kind of object whatsoever.

We could deal with these object-less emotions by differentiating them from emotions proper, and placing them in a separate category, e.g. "moods" (de Sousa, 2003), but this seems a rather artificial manoeuvre.

Alternatively, we could assimilate object-less emotional states to emotions of the same kind that have an object, by virtue of the strong "family resemblance" between the former and the latter, but at the cost of having to deny that the class of emotions designates a natural category. The decision to include a mental state as an emotion would be made within the context of a "language game". This flies in the face of neurological evidence (to be presented below) indicating that emotional states are sharply distinguished from cognitive states, and that the different kinds of emotions are also fairly well-defined.

Finally, we could differentiate between specific emotions and kinds of emotions (de Sousa, 2003). The former may lack an object; the latter cannot.

Philosophers speak of each kind of emotion as having a formal object which defines what it is. De Sousa (2003) describes this object as "a property implicitly ascribed by the emotion to its target, focus or propositional object, in virtue of which the emotion can be seen as intelligible". For instance, my fear of a snake only makes sense because I construe certain properties of the snake as frightening.

But it does not help us much to say that we fear snakes because we think they are frightening. To make sense of our fear - or indeed, any kind of emotion - we need to know what it is for. The only standard by which we can judge the appropriateness of an animal's emotional responses is its own well-being. Since we cannot ask non-human animals lack language, we have to assess their well-being from an external, scientific standpoint. Accordingly, I envisage the object of each kind of emotion not merely as a formal object, but as a teleological object that explains what the emotion is "for" and how it helps the animal to survive and/or flourish. Each kind of emotion is "for" responding appropriately to a certain kind of object, which I propose to call the generic intentional object of the emotion. This generic object is what the emotion is about. Fear, for instance, is "about" objects recognised as dangerous, and it is for responding appropriately to them. That is what makes it different from anger.

The purpose (i.e. intrinsic finality) of each kind of emotion can be explained in Darwinian terminology. Below, I shall defend the view that the generic intentional object of each kind of emotion is simply the kind of environmental challenge that the emotion evolved to meet.

Object-less feelings are therefore derivative upon directed feelings. I propose that the ascription of object-less feelings to a non-human animal can only take place after we have identified:

(i) emotions of the same kind in that animal, that are "directed" at something;

(ii) the generic intentional object of the kind of emotion the animal is feeling; and

(iii) non-arbitrary, teleological criteria which allow us to identify the different kinds of emotions as natural categories.

The advantages of the approach defended here can best be seen by contrasting it with a behavioural approach to the emotions. The latter approach would allow us to define individual emotions as behavioural dispositions towards their intentional objects, conceived as external to the animal - e.g. an animal's movement towards (or away from) X for the sake of obtaining (or avoiding) X, is "about" X. Although this account incorporates intentionality (by stipulation), it cannot explain object-less emotions, such as undirected feelings of fear or depression.

Another drawback of this account is that it defines "aboutness" in purely non-mentalistic terms: bacteria can also move towards their objects, for the sake of obtaining them.

But perhaps the most serious problem with this behaviourist approach to intentionality is that it says nothing about the different kinds of emotions. Distinctions have practical significance here. For instance, the question of whether fear and panic are different emotions or two versions of the same emotion is not a semantic one, but a substantial one, if you happen to be a researcher looking for a drug to treat panic disorders.

Finally, although a dispositionalist account of emotions overlooks their social context. How one expresses an emotion like rage will vary considerably according to the company that one is keeping and the nature of the occasion, for instance. One may even refrain from showing one's feelings at all.

We can now re-write the third of our five selected features of animal emotions as follows:

Emotions come in different kinds, and each kind of emotion has a range of intensities, and a generic intentional object. Emotions on specific occasions typically have intentional objects too.

Even with these revisions, our list of the features of animal emotions remains an incomplete characterisation. Although it stipulates that they are mental states, it fails to explain what makes them mental. The only feature it lists which is pertinent to mental states is their intentionality, but as we saw in chapter two, behaviour describable in intentional terms can occur in the absence of mental states. Something more is needed.

In this chapter, I shall endeavour to answer four key questions regarding animal emotions:

(1) What are the cognitive pre-requisites of animal emotions, and what makes them mental events?

(2) What are animal emotions "about", and what is each basic kind of animal emotion about? (The problem of intentionality.)

(3) How do we identify basic emotions in animals, and which animals can be said to have them?

(4) In which animals are basic emotions (and other feelings) phenomenally conscious? (The Distribution Question.)

In the following sections, I shall investigate emotions from the standpoint of the two mentalistic intentional stances I identified in chapter two: the agent-centred stance (which will be used to address the first three questions) and the first-person stance (which will be used to answer the question of when feelings can be said to be conscious).

1. What are the cognitive pre-requisites of animal emotions?

In this section, I argue that contrary to cognitivist and appraisal theories of emotion, animal emotions cannot be reduced to cognitive states. I propose that we can identify emotions in animals because of the role they play in intentional agency, even though animal emotions are usually not accompanied by intentional action. Specifically, all animals belonging to species whose behaviour is best described in terms of the agent-centred intentional stance discussed in chapter two - and only those animals - can be said to have emotions. Because the ability to have emotions goes hand in hand with the capacity for intentional agency, emotions can only be ascribed to animals capable of forming relevant beliefs. However, I argue that these beliefs do not have to be propositional.

(a) Is emotion fundamentally distinct from cognition?

Certain influential philosophical theories of emotion characterise emotions using cognitive terminology. For cognitivists, emotions necessarily involve having propositional attitudes towards certain statements. Some cognitivist theories construe emotions as judgements. In appraisal theories, emotions are typically characterised as evaluations.

Regardless of what sort of (conscious or unconscious) cognitions these theories use to characterise emotions, they all collapse in the face of the weighty neurological and behavioural evidence that has been amassed, showing that emotion and cognition are indeed fundamentally distinct.

Neurological arguments for a distinction between emotion and cognition

Joseph LeDoux. Photo courtesy of New York University Center for Neural Science.

LeDoux (1998) marshals several lines of argument to the effect that the neural processing for cognitive and emotional reponses is quite distinct, which refutes the view that emotions are either conscious or unconscious cognitions. For example:

Behavioural arguments for a distinction between emotion and cognition

The main behavioural argument for a distinction between emotion and cognition is that cognitive processes, in the absence of emotions, are incapable of motivating animals to act. Aristotle seems to have made essentially the same point in his De Anima 3.10, where he argued that of the two things that seem to cause movement in animals - desire and practical thinking - desire is the crucial factor, as thinking cannot move an animal in the absence of an object of desire.

Antonio Damasio. Photo courtesy of University of Iowa Health Care.

Damasio (1994) has also documented several case studies in which subjects who had a diminished capacity to experience emotion, owing to injuries sustained in the prefrontal and somatosensory cortices of the brain, were severely hampered in their ability to make intelligent practical decisions.

Similarly, Evans (2001) contends that the vital role of emotions in agency can never be replaced by reason alone. He argues that a race of creatures with a capacity for logic but no emotions, like Dr. Spock in Star Trek, would be non-viable:

It should be clear by now that a creature totally devoid of any emotional capacity would not survive for very long. Lacking fear, the creature might sit around and ponder whether the approaching lion really represented a threat or not... Lack of disgust would allow it to consume faeces and rotting food. And without the capacity for joy and distress, it might never bother doing anything at all - not a good recipe for survival (2001, p. 62).

Conclusion E.1 Animal emotions cannot be reduced to cognitive states such as judgements or evaluations, whether these states be conscious or unconscious. Cognitivist theories and appraisal theories of emotion are therefore inadequate.

(b) The role of emotions in intentional agency

Daniel Dennett. Photo courtesy of University of California.

In chapter two, I argued that Dennett's intentional stance could be applied to all mental states, including emotions. However, it turned out that there were two ways of describing the intentionality of mental states: either in terms of the agent's beliefs and desires, or simply as first-person states.

This analysis suggested that an alleged mental state might fail to qualify as such for one of two reasons: it may be possible to adequately describe it in terms of information-states and goals rather than beliefs and desires, or it may be possible to adequately describe it in terms of third-person states rather than first-person states.

For instance, aversive behaviour towards noxious stimuli, which is found in all cellular organisms, including bacteria (taxis), is an insufficient criterion for identifying the occurrence of fear, and similarly, antagonistic behaviour (which occurs in insects) is an insufficient criterion for anger. A goal-centred intentional stance can explain these kinds of behaviour more parsimoniously than a mentalistic account, by appealing to the animal's information-states and goals.

Likewise, since (as we argued in chapter two) learning can take place in the absence of mental states, fear conditioning, or the ability of worms, flies, snails and other animals (LeDoux, 1998, p. 146) to learn to avoid a neutral stimulus which is associated with a noxious stimulus, is an insufficient criterion to establish the occurrence of a mental state such as fear, whether we envisage it as conscious or unconscious.

However, it was also argued in chapter two that while perception, memory, learning and conditioning were not necessarily mental states, certain kinds of learning could only be explained by adopting a mentalistic agent-centred intentional stance: operant conditioning, spatial learning, tool-making and social learning. These kinds of learning are manifestations of intentional agency and presuppose the occurrence of relevant beliefs and motivations. In chapter two, we described these motivations as "desires", but other emotions (e.g. fear or anger) are capable of motivating animal agents equally well.

We now have a criterion for genuine mental states, which emotions satisfy:

Conclusion E.2 Any positive or negative internal state that is capable of motivating intentional agency in a species of animal qualifies as a mental state in that species, and is therefore a genuine emotion (and not a mere behavioural disposition).

One might object that emotions typically accompany reactions rather than actions, and that most emotional reactions in animals are innate and beyond their control. Attempting to identify animal emotions by focusing on intentional agency seems like searching in the wrong place.

However, it needs to be kept in mind that emotions are by definition mental states, and can only be identified when they are manifested in states of affairs which are most appropriately interpreted in terms of mental states. We argued in chapter two that all mental states can be described by Dennett's intentional stance, and that a mentalistic interpretation of an event was warranted if (and only if) it was more scientifically productive to describe the event using either an agent-centred intentional stance or a first-person intentional stance.

Reactions, by definition, cannot be described in terms of an agent-centred intentional stance, as they are not actions. (Nor does it seem that they could require a first-person intentional stance to explain them. Animal reactions are either innate or acquired through conditioning. And as we saw in chapter two, both kinds of reactions can easily be explained by adopting a mind-neutral, third-person intentional stance which employs purely causal terminology.) The only kind of conditioning that requires a mentalistic explanation is operant conditioning, which (as we argued in chapter two) manifests intentional agency. Reactions alone, then, cannot establish the occurrence of mental states such as emotions in animals.

Conclusion E.3 Emotional reactions are bona fide mental states, even in the absence of agency. However, if we want to identify clearcut cases of genuine emotions in animals, then we have to restrict our search to a narrow subset of emotional behaviour and focus solely on intentional agency.

Conclusion E.3 narrows the range of animals to whom we may attribute emotions:

Conclusion E.4 Emotions can only be identified in those species of animals that exhibit intentional agency.

(c) What kind of beliefs do emotions require?

The foregoing conclusions help us to resolve our conflicting intuitions regarding the cognitive requirements for having emotions. On the one hand, it seems obvious that we can have feelings such as fear, even in the absence of beliefs. On the other hand, having an emotion (e.g. fear) seems to presuppose that one has certain beliefs about its intentional object (e.g. that it is frightening).

How fast one jumps away from a snake like this can make the difference between life and death

If we are talking about an animal's emotional reactions, then it is indeed possible that they may occur in the absence of beliefs. Certainly, they will be felt before the animal has had time to form relevant beliefs. LeDoux's example (1998, p. 166) of a hiker who recoils in fear from a thin, curved object in his path, even before his brain has had time to identify the object as a snake, is relevant here. The cognitive processing required to form a belief ("This is a snake") takes time, but in life-and-death situations, emotional reactions in humans and other animals have to be very fast. To make such a decision, the brain relies on crude cognitive processing in the thalamus which identifies the thin, curved object as a possible danger.

By contrast, intentional agency presupposes the existence of beliefs regarding the attainment or avoidance of the object, as we argued in chapter two. The animal forms and revises these beliefs in the process controlling and correcting its bodily movements in an effort to attain its goal. Since emotions can only be identified in animals capable of intentional agency, we may conclude:

Conclusion E.5 Emotions may occur in the absence of belief, but can only be identified in animals that are capable of holding beliefs.

But how should we construe these beliefs?

Emotions are not propositional attitudes

Cognitivist philosophers (e.g. Frey, 1980) have argued that emotions necessarily involve having propositional attitudes towards certain statements: for instance, one cannot be angry with someone without believing that she is guilty of doing something bad.

There are two issues at stake here: first, whether having emotions can be reduced to holding propositional attitudes, and second, whether emotions presuppose these attitudes.

The straightforward identification of emotions with propositional attitudes is fraught with problems (de Sousa, 2003).

First, emotions may run contrary to one's propositional attitudes, as when someone has a fear of flying yet believes it to be the safest means of transport. Thus believing that X is dangerous is neither necessary nor sufficient for experiencing fear of X (de Sousa, 2003).

Second, dispositional beliefs typically have a simple tailor-made form of behavioural expression: someone who believes that P will manifest her belief by publicly assenting to P. Dispositional emotions, by contrast, are much more open-ended, and are capable of being manifested in a diverse range of behaviour.

Conclusion E.6 Having emotions cannot be reduced to holding propositional attitudes.

Emotions do not presuppose propositional attitudes

The view that emotions, while not identical with propositional attitudes, presuppose these attitudes, merits more serious consideration. If it is true, non-human animals are excluded from having emotions at all, as they lack the linguistic sophistication to formulate propositions. Infants and severely cognitively impaired human beings would likewise be excluded.

Tom Regan. Courtesy of The Culture and Animals Foundation.

Regan (1988, p. 42) argues that this conclusion is absurd. He offers the counter-example of an intellectually impaired man who is incapable of learning a language, reacting in terror to the sight of a rubber snake. Our "common sense" intuition is that the man is afraid because he believes the snake is pursuing him and will harm him. If a non-human animal reacted similarly, we would say it was afraid too, and impute the same belief to it. Frey (1980, p. 90) adopts a contrary position, arguing that beliefs (unlike states of affairs) are capable of being true or false. To believe that P is to believe that the sentence "P" is true. Since non-human animals lack language, they cannot formulate propositions and are therefore incapable of having beliefs.

This argument assumes that language is required to distinguish true from false beliefs. It has been argued in chapter two that this is not so: self-correcting behaviour, displayed by animals undergoing operant conditioning, can achieve the same result. If an animal can make mistakes which it then tries to rectify, then it can be said to have beliefs about its goals. Intentional agency is the critical factor, not language.

Let us modify Regan's case a little, and suppose that the snake is not a rubber snake but a very life-like mechanical toy with a built-in thermal sensor, programmed to pursue the nearest warm object. Several people and one animal are in a locked room with the snake. The animal is standing nearest to the snake. Startled, it initially backs away but is pursued by the snake. If the animal proves capable of adjusting its strategy for avoiding the snake, by learning to position itself behind someone in the room when approached, then we can justifiably say that the animal believes that the snake will pursue (and harm) that person rather than itself. The attribution to the animal of the mental state we call fear would then be warranted, as the animal is using the content of its beliefs to avoid the snake.

But what should we say about an animal which is incapable of engaging in self-correcting behaviour, in any situation? What if it reacts to emotional stimuli, but is incapable of modifying its actions to pursue or avoid these stimuli? In that case, we would have no warrant for ascribing agency to the animal, and hence no grounds for ascribing emotions to it (according to Conclusion E.4). As we argued above, reactions alone cannot supply such a warrant.

We can now formulate the following conclusions:

Conclusion E.7 The attribution an emotion to an animal does not require it to have propositional attitudes.

Conclusion E.8 The attribution an emotion to an animal requires it to be capable of self-correcting behaviour that enables it to modify its strategies for attaining its goals.

In other words, propositional attitudes are not necessary. Strategic attitudes - "This works; that doesn't" - are what counts. That being so, there is no compelling reason why we should be reluctant to ascribe emotions to animals - even if specific emotions, such as remorse, presuppose linguistic abilities beyond the reach of non-human animals.

The propositional content of beliefs that accompany emotions

Even if emotions do not presuppose propositional attitudes, it could be argued that they must have some propositional content. However, this would entail the absurd consequence that the hiker described by LeDoux (1998, p. 166), who recoils from a thin, curved object in his path, even before his brain has had time to identify the object, does not experience the emotion of fear, as his emotion has no specific content until the visual cortex of his brain identifies the object as a snake.

On the other hand, any beliefs which accompany a particular emotion must surely have a propositional content - otherwise, they could not be described as beliefs. But what might this content be?

First, the animal has to be capable of forming strategic beliefs whose content relates to how it can pursue or avoid the intentional object of the emotion, as occurs in operant conditioning.

Second, the animal may be said to implicitly believe any propositions whose content is entailed by the strategic beliefs it forms. Returning to the case cited above, an animal that entertains the strategic belief that it can avoid a snake by positioning itself behind one of the people in the room, also implicitly believes that:

(a) there is a snake in the room;
(b) the snake needs to be avoided;
(c) there are people in the room; and
(d) it is not presently positioned behind any of them.

If my proposals regarding animal beliefs are correct, then an animal can only be said to have emotions directed at intentional objects if it is capable of forming beliefs about strategies for attaining or avoiding those objects. This implies that if an animal desires X as an end in itself, it must also be capable of believing that Y is a means to X, and consequently desiring Y.

Conclusion E.9 Any animal that is capable of desiring ends (e.g. food or sex) must also be capable of desiring means to these ends.

I would thus agree with Frey (1980, p. 104) in rejecting the possibility of an animal that only has simple desires, such as a dog's desire for a bone. Nor can I accept Regan's proposal (1988, p. 58) that a dog which desires a bone also has a preference-belief that there is a connection between its choosing a bone and satisfying its desire for a certain flavour. First, it is doubtful whether dogs can desire such abstract things as "flavours"; in Regan's example, the appropriate object of the dog's desire is surely the bone itself. Second, there is no genuine means-end behaviour in this case, as there is only one action (choosing the bone) and one physical object (the bone), with no manipulation. Finally, one can construct parallel examples with organisms lacking desires. Should we say that a bacterium chooses a glucose-rich solution as a means of satisfying its desire for sweetness? By contrast, Regan's example (1988, p. 70) of the dog digging in the garden to retrieve an old bone it has buried is a perfect illustration of the kind of behaviour that manifests a strategic belief. No bacterium could do that.

Although I agree with Frey's rejection of simple desires, I part company with him on the question of whether an animal's behaviour alone can manifest the beliefs it must be capable of entertaining, before it can be credited with desiring those ends. The examples cited by Frey to establish the inherent ambiguity of animal behaviour - notably the case of his dog, which wagged its tail in the same way when its master was outside the door, when lunch was imminent and when the sun was ecliped by the moon - are not good ones, as they do not include the kind of strategic behaviour discussed above.

(d) Do the basic emotions of fear, anger and desire presuppose a capacity for language?

Before we conclude our discussion of the identification of basic emotions in animals, we have to address arguments purporting to show that these emotions presuppose feats of rationality and language that non-human animals are incapable of. I argue that on the contrary, non-human animals can and do satisfy the core cognitive requirements for these emotions, without requiring language.

Cognitive requirements of fear

It has been argued that fear, in the fully-fledged sense of the word, has to be amenable to reason.

Leahy (1994, p. 136) claims that when we ascribe fear to animals and to rational human beings, we are playing two distinct language games: human fear is amenable to reason whereas animal fear is not. People can be argued out of their fears if they can be shown to be groundless, but animals cannot. The object of an animal's fear operates not as a reason for its behaviour, but rather as a cause of its behaviour (1994, p. 135). Indeed, Leahy considers the dissimilarities between animal and human fear to be so profound that one could justifiably use two distinct words to differentiate them. In the end, he decides to use one verb to describe both cases, but only because the similarities in the overt behaviour of frightened people and animals are so profound.

I have several comments to make in response to Leahy's claims. First, it would be grossly mistaken to view the similarities between human and animal fear as merely behavioural, as Leahy seems to do (1994, pp. 135-136). Animal fear cannot be cashed out in dispositionalist terminology which describes external behaviour. The fact that scientists routinely perform research on animals to discover the causes of and treatments for fear in humans (LeDoux, 1998; Hall, 1999) would make no sense unless the internal neurological and affective states accompanying fear were substantially the same in humans and other animals.

Second, when contrasting human fears with those in other animals, it is essential to compare like cases. Once we do so, we find that fear, properly speaking, is amenable to reasoning in humans only under restricted circumstances. The inability of other animals to reason their way out of their fears then becomes far less anomalous.

There is a growing consensus among psychiatrists (see Catherall, 2003, pp. 76-78) that there is a significant difference between fear and anxiety in both humans and animals: fear is a response to a present danger (e.g. a predator) that is triggered by perceptions (e.g. the sight of a lion), while anxiety (or worrying) is a response to potential threats that involves cognitions rather than perceptions. The point is that because fear, unlike anxiety, is processed within the brain independently of higher-level cognition, fear behaviour is typically involuntary, even in human beings.

Behavioural responses to innate fears, such as the fear of a predator, are involuntary in humans as well as non-human animals. These fears have evolved in response to stimuli that consistently threatened our survival during evolutionary history (Panksepp, 1998, p. 207). These innate fears are tailor-made to circumvent the need for reason and other cognitive inputs, in both humans and animals. In LeDoux's (1998, p. 166) earlier example of a hiker in the woods who abruptly encounters a long thin object in his path and jumps clear, - even before the visual cortex in his brain has had time to ascertain that it is indeed a snake and not a stick, - the ability to respond rapidly may make the difference between life and death.

Conditioned fears in humans and other animals are also largely involuntary, because fear memories are stored in the brain's amygdala, from which they can never be erased (LeDoux, 1998, p. 146; Hall, 1999). There is a good evolutionary reason for this: the brain's ability to recall stimuli associated with danger in the past assisted individuals' survival (LeDoux, 1998, p. 146). According to LeDoux (1998), the only way to eradicate a fear acquired through conditioning) in non-rational animals, is to repeatedly expose the animal to the conditioned stimulus in the absence of the unconditioned stimulus. The same approach is used in the first phase of treatment for humans suffering from post-traumatic stress disorder (Catherall, 2003, pp. 84-87). Eventually, this leads to "extinction" of the fear response. In reality, however, the fear is merely dormant, and may be re-awakened simply by exposure to some stressful or traumatic event (LeDoux, 1998, p. 145). Conditioned fears prove to be equally indelible in people who develop phobias. Phobias show only limited "amenability to reason": psychotherapy allows the fear of the phobic stimulus to be kept under control for several years, but after some stress or trauma, the fear returns in full force, just as it does in animals. Therapy, like extinction, cannot erase the memory (LeDoux, 1998, p. 146). The differences here between humans and other animals hardly deserve to be called a new "language game".

The study of fear disorders in human beings sheds further light on why they are not amenable to reason. Catherall (2003) points out that while anxiety disorders can be treated on a cognitive level (as in insight therapy), fear disorders cannot, because exposure to a traumatic stimulus triggers changes in the brain which prevent reasoning and language processing. PET scans show that in a subject exposed to a traumatic stimulus (e.g. a snake), the speech area of the brain (Broca's area) may be deactivated, making it impossible for the subject to access his explicit or declarative memory and deal with the fearful stimulus by recalling facts ("It's a green tree python, so it can't poison me") that would alleviate his fear. This kind of cognitive processing is only possible while the individual's fear state remains below a certain threshold (Catherall, 2003, pp. 79-80).

Thus Leahy's contention (1994, p. 137) that people (unlike animals) can be argued out of their fears if they can be shown to be groundless, is true of anxiety rather than fear as a whole. Only low-level human fears are "amenable to reasons" (Leahy, 1994, p. 135).

It should not be thought that fear behaviour in non-human animals is totally inflexible. In fact, we find a suite of adaptive behaviours to fear in animals. First, animals can lose their fear of an object through a process of habituation.

Second, animals lose their fear of an object if a change in perception reveals that it was not the danger they thought it to be.

Third, animals living in hierarchical societies are capable of learning from experience, that they need no longer fear formerly dominant conspecifics that lose status.

Fourth, juvenile animals can learn specific skills for avoiding things they fear - such as predators.

Is there, then, any minimum level of adaptability that we should expect in an animal capable of intentional agency, which is undergoing the emotion of fear? Habituation does not seem particularly impressive: as we saw in chapter two, it is not one of the behavioural conditions for intentional agency. By contrast, an animal's ability to use incoming sensory data to self-correct its behaviour is an integral part of the mechanism of intentional agency, be it operant conditioning, spatial navigation, tool use or social learning (see DF. 1 to DF. 4), and should therefore feature in even a minimal definition of fear as a mental state:

Conclusion E.10 Before it can be credited with fear at its lowest cognitive level, an animal must not only be capable of intentional acts enabling it to avoid some dangerous stimulus, but also be capable of adjusting its behaviour when new sensory data reveal the stimulus to be harmless.

Regan (1988, p. 42) cites the case of a deranged man, bereft of reason, who shows signs of terror and jumps back when confronted by a rubber snake, apparently trying to avoid it. Is he really afraid? We cannot alter the man's behaviour by telling him that the snake is not real, but we can still change his perception - either by removing the rubber snake or by chopping it into pieces. If the man then calms down, then we can say he was afraid.

It is certainly true that humans have a unique capacity to control their fears - even innate ones. Most of us have sufficient control over our fear of snakes that we can pick up a green tree python, which we know to be harmless, and a few individuals can even conquer their fears enough to handle poisonous snakes. But what these examples show is the uniqueness of human cognition, rather than human emotion, as we re-define what is and is not dangerous ("That green snake won't hurt you; it's harmless.") Humans, unlike other animals, are capable of modifying their behaviour on the basis of information conveyed through the medium of language: we have been taught that despite appearances, green tree pythons will not hurt us, and even cobras can be handled safely. A non-human animal knows that snakes are dangerous, but because it lacks language, it cannot explain precisely why they are, and thus it is incapable of understanding why its innate fear response to green tree pythons is inappropriate.

Language also allows human beings to rationalise stimuli that might otherwise frighten them. According to Grandin (1997), prey animals (including horses and cattle) have an innate tendency to acquire fears of things that look out of place (even a piece of paper blowing in the wind), sudden movements (which resemble the movements of predators) and high-pitched noises. But even though these animals cannot tell themselves that there's nothing to be afraid of, there is no reason to doubt the reality of their fear.

Cognitive requirements of anger

The difficulty in ascribing anger to animals arises from the advanced mental states that anger appears to require. Thus Aristotle defined anger as "a longing, accompanied by pain, for a real or apparent revenge for a real or apparent slight... when such a slight is undeserved" (1959, The Art of Rhetoric, 1378a, J. H. Freese (tr.), London: William Heinemann). Leahy (1994, p. 80) argues that these cognitive requirements are beyond the competence of non-human animals, and that it is only their enraged behaviour that resembles that of angry human beings.

Two comments are pertinent here. First, instead of saying that only undeserved slights can elicit anger, it would be better to say that the knowledge that a slight is deserved can reduce (and perhaps obviate) anger. Of course, most animals, like very young human infants, lack the concept of "just deserts", so their feelings of rage cannot be assuaged in this way.

Second, Aristotle's definition of anger is too narrow: he focuses on "slights", or insults, but in human life, one may feel anger at another individual for a variety of reasons: a slight, an action that gave offence without being intended to do so, some sensory irritation ("He has terrible body odour") or a physical obstruction ("I wish he'd get out of my way").

Third, we can speak of various "cognitive grades" of anger. A human longing for revenge against a mortal enemy is obviously of a much higher grade than the momentary surge of anger an infant may feel towards an "offending" individual who is thwarting her wishes (e.g. by denying her something she wants), but I would argue that if the infant attempts to strike back at the individual, acting on certain strategic beliefs about appropriate ways of doing so (e.g. "Pushing didn't work? OK, try punching or kicking"), then her behaviour qualifies as intentional agency and hence bona fide anger of a low-level variety: rage. I propose the following tentative conclusion:

Conclusion E.11 In order to be capable of anger at its lowest cognitive level, an animal must be capable of intentional acts directed against some offending object, individual or bodily irritation, which are accompanied by certain strategic beliefs about an appropriate way to strike at the offending stimulus.

This ability is likely to be found in all vertebrates, as well as many insects and cephalopods.

We could define a higher grade of anger as the desire to strike out at an offending individual. The cognitive requirements of anger are likely to be satisfied by many species of vertebrates and possibly even insects. The desire to strike out at an offending individual requires nothing more than (a) an ability to recognise other individuals and (b) an ability for "book-keeping", or keeping track of other individuals' behaviour during past interactions. The former ability has been documented even in wasps (Tibbetts, 2001) and is widespread across fish families (Bshary, Wickler and Fricke, 2001). The latter ability has been identified in at least two kinds of fish - sticklebacks and guppies are capable of "book-keeping" with several partners simultaneously, and there is tentative evidence that they tend to adopt a "tit-for-tat" strategy in their dealings with one another: a player starts co-operatively and does in all further rounds what his/her partner did in the previous round (Bshary, Wickler and Fricke, 2001). Non-human primates are renowned for their ability to keep track of their own and others' misdeeds over a longer period, as described by van der Waal in his book "Good Natured" (1996).

Cognitive requirements of desire

Frey (1980) offers an ingenious argument against the occurrence of desires in animals. He considers the straightforward case of a dog that desires a bone. "Suppose", he argues, "my dog simply desires the bone: is it aware that it has this simple desire? It is either not so aware or it is" (1980, p. 104). He then argues against both possibilities. If the dog is unaware of its simple desires, then it has unconscious desires. While it might make sense to say that some of a creature's desires are unconscious, it makes no sense to say that all of them are, for then the creature's conduct would be no different from that of a creature with no desires at all.

If one the other hand, the dog is aware of its desires, then we have to answer the epistemological question: how do we know that it is aware of them? Nothing in the dog's behaviour could tell us this.

However, Frey's argument hinges on the questionable assumption that to have a conscious desire, one must be aware of one's desire. An alternative position (Lurz, 2003) is that to have a conscious desire, one must simply be aware of its object. What this awareness might consist in will be discussed below.

Additionally, there may be ways of ascertaining whether a dog is aware of its desires. For instance, I argue below that the phenomenon of hedonic behaviour suggests an awareness in an animal of its internal affective states.

Conclusion E.12 The basic emotions of fear, anger and desire do not presuppose the use of language.

2. What are animal emotions "about", and what is each basic kind of animal emotion about?

We still need to address the question of what emotions are "about". If emotions cannot be "cashed out" in the language of subjective feelings, behavioural dispositions, or cognitive states, can we understand them better in "physicalist" terms? I examine two popular physicalist accounts of emotions, which envisage them either as brain states (Panksepp's (1998) neurophysiological account), or internal bodily feelings (e.g. LeDoux's (1998) "feedback" theory). I argue that Panksepp's neurophysiological approach to the emotions offers the most promising avenue for investigating emotions in animals. By situating animals' brain states within their historical, environmental and physical contexts, it explains (as bodily feedback theories cannot) how physical states can possess the property of intentionality. Panksepp's account also sheds light on the question of how intentional agency could have arisen in animals.

(a) Are emotions best understood in neurophysiological terms?

Jaak Panksepp. Photo courtesy of Bowling Green State University, Ohio.

At first glance, the notion that the "aboutness" of animal emotions can be explained in terms of their underlying brain states looks distinctly unpromising. Brain states do not seem to be "about" anything, so a neurophysiological account of intentionality appears doomed from the start. The prospects of a brain-based account explaining the other features of animal emotions look equally dim. It is by no means clear how these states can explain the mentalistic quality of animal emotions; nor does it appear that brain states alone can account for the bodily manifestations of animal emotions, let alone their social role.

To attempt to reduce the meaning of an emotion or feeling to its neurophysiology would indeed be tantamount to a Rylean "category mistake", but that is not what I am proposing. It is now widely acknowledged that animals' emotional responses are regulated by their brains. I shall argue that there are good reasons to support the view that the study of neural processes, in their physical and environmental context, enables us to explain all of the essential features of animal emotions, identify and distinguish between each kind of emotion in animals, and determine what each kind of emotion is for.

I suggested above that each kind of emotion has its own generic intentional object that explains its "aboutness" and makes it what it is. One merit of Panksepp's (1998) neurological approach to the emotions is that it allows us to identify the generic object of each emotion in a straightforward way. On a generic level, at least, this offers a promising solution to the philosophical problem of how a physical state of affairs such as a brain state can be "about" something.

Panksepp (1998) argues that the basic kinds of emotions in animals arose in response to different kinds of environmental challenges their ancestors encountered. Each emotion arose in response to a unique environmental challenge in the lives of certain groups of animals, which their ancestors' brains evolved to meet. Meeting that challenge is what each kind of emotion is "about". An animal's response to each of these challenges is mediated by several "emotion systems" within its brain, easily identifiable to specialists, that give each kind of emotion its characteristic neurological "key signature".

A neurophysiological account of the intentionality of the basic kinds of animal emotions

How does this help us with "aboutness"? I propose that there are two robust senses - one mind-neutral and the other mentalistic - in which the various kinds of emotions are "about" the environmental challenges they evolved to meet. First, each environmental challenge has caused the evolution of a distinctive suite of emotional responses which are directed at it. This has been accomplished through natural selection over millions of years: an animal's emotional response to a challenge (e.g. jumping back at the sight of a snake) promotes its survival. Second, as emotions are capable of motivating intentional actions as well as reactions, we can say that each environmental challenge has caused the evolution of a kind of mental capacity which is specifically directed at it. (For instance, an animal's emotion of fear can motivates it to not only react, but also act intentionally, in a way that saves its life.) Thus the environmental challenges facing animals are not only the ultimate causes, but also the objects, of the different kinds of emotions in animals.

Of course, bacteria and plants also have many organismic traits that have evolved to meet environmental challenges. These traits possess "aboutness" in the first but not the second of the two senses described here. Their "aboutness" is thus non-mentalistic.

Searle (1999) criticises causal accounts of intentionality, on the grounds that "[y]ou can always get the causal relations without the intentionality" (1999, p. 91). It is certainly possible to contruct "wayward causal chains" in which a mental state is caused by one physical event but is about something else entirely. However, I believe that Searle's criticism does not undermine the account defended here, for two reasons.

First, the causal relationships described here are very strong ones, as (i) the features of animals' emotion systems have been shaped directly, by the very properties of the environmental challenges they have evolved to meet, and (ii) animals' emotion systems promote survival in virtue of their intrinsic properties. There is no room for wayward causal chains here, as natural selection (unlike variation) is non-random: certain organismic traits are selected for they are inherently survival-promoting in the animal's current environment.

Second, the environmental challenges that shaped the evolution of animals' emotion systems are not only their causes, but also their objects, insofar as animals' emotion systems are directed at them.

Searle elucidates the intentionality of individual mental states (including emotions) in terms of their role in intentional agency. For Searle, intentions are causally self-referential: "If I want to drink water, and then I drink water by way of satisfying my desire to drink water, then my mental state, the desire (that I drink water) causes it to be the case that I drink water" (1999, p. 105).

The point I wish to make, however, is that while individual emotions have the property of "aboutness", the various kinds of emotions possess this property too. I find nothing to disagree with in Searle's account of the intentionality of individual emotions, but it is not designed to answer the question of what each kind of emotion is about. The chief merit of a neurophysiological account of emotion, as opposed to rival cognitive or "body-based" accounts, is that it tells us what the generic object of each kind of emotion is (a distinctive kind of environmental challenge), and where to look for it (in animals' brains, where the emotion systems that regulate their emotional responses are located).

Conclusion E.13 The generic intentional object of each kind of emotion felt by an animal is not a property or a "thing", but an environmental challenge that the animal's ancestors evolved to meet.

Brains also have a history. Each neurologically distinct "emotion system" in an animal's brain which regulates its response to an environmental challenge reflects the way in which its brain evolved to meet this challenge. The variety of emotions we see in an animal constitutes a repertoire of challenges, to which its brain and body are adapted.

Studies of animals' brains can thus reveal (i) the original motivational context of the basic animal emotions, (ii) the "generic intentional object" of each basic emotion, (iii) the proper taxonomy of the basic animal emotions, (iv) the evolutionary history of these emotions, and hence (v) which animals possess these emotions.

The adequacy of a neurophysiological account of emotions

A neurophysiological account which sitautes the emotions in their environmental and physical context, can also explain the five features of animal emotions we identified above.

As we saw in chapter two, intentional agency, which is what makes an emotions a mental state, is mediated through an animal's brain, which stores its internal representations of the animal's goals, as well as the means to obtain them, and moves the animal's body accordingly. Feedback from the body allows the animal to continually update its neural representations. Insofar as a neurophysiological account sheds light on the evolution of intentional agency, it can help explain how emotions originally arose as mental states. Comparisons between different phyla of animals, of varying degrees of complexity, are relevant here. Prescott's (2000) outline of the evolution of action selection, which allows an animal to "select" rapidly between competing behavioural alternatives, lays the groundwork for a possible future investigation into the evolution of intentional agency.

The possibility of a neurophysiological explanation of phenomenal consciousness remains controversial. Panksepp (2004) regards affective experience as an emergent product of the animal brain's neurobiological complexity; for McGinn (1999), consciousness is elusive and may always remain so. Since I have argued above that consciousness may not be an essential feature of animal emotions, I shall return to this issue in section 4 below.

It might be thought that a brain-centred approach is too narrow to encompass the second feature of animals' emotions: their bodily manifestations in animals' physical responses and intentional behaviour. However, even proponents of a rival "bodily feedback" theory of the emotions (discussed below) acknowledge that it is the brain which generates the body's response to emotions (LeDoux, 1998, pp. 292-295).

Neurophysiology accounts for the third feature of animal emotions very well, as it explains the different kinds of feelings animals have, the range in their intensities, and the variety of their intentional objects, in terms of the environmental context in which the "emotion systems" in animals' brains originally evolved.

Neurophysiology also sheds light on the fourth and fifth features. Since animals' emotions arose in response to different kinds of environmental challenges their ancestors encountered, we can say that these challenges motivate animals to seek, fight or avoid them. Feelings also play a role in animals' social lives, because they can motivate them to seek the company of others, to play and to care for their young.

A final objection to a neurophysiogical account of emotion is that the same emotion could have multiple physical realisations in different groups of animals, but if we identify emotions with brain states we will have to say that this is impossible. This objection rests on a misunderstanding. It is entirely possible that two lineages of animals (say, mammals and birds) evolved very different neurological systems (call them A and B) for coping with the same kind of environmental challenge. On the neurophysiological account being defended here, systems A and B mediate the same emotion, as they arose in response to the same generic challenge.

Conclusion E.14 Panksepp's (1998)neurophysiological approach to animal emotions, which situates them in their environmental and physical contexts, can account for all of the essential features of animal emotions.

Before we can conclude that a neurophysiological account of the emotions is true, however, we have to show that rival theories of the emotions fail to explain them adequately. We have already examined cognitivist accounts and found them wanting, but our discussion would be incomplete without mention of another group of accounts in the physicalist mould which warrant serious consideration: body-centred accounts, which reduce emotions to internal feelings.

(b) Can emotions be explained as bodily states?

William James. Photo courtesy of Center for Cognitive Liberty and Ethics.

One popular scientific theory of the emotions, known as the James-Lange theory because it was originated independently by Williams James and Carl Lange in 1884, holds that emotions are internal feelings that are generated by the body's internal physiological reactions to events. These reactions centre on the body's autonomic and motor functions:

[W]hen we see [a] ... bear, we run away. During this act of escape, the body goes through a physiological upheaval: blood pressure rises, heart rate increases, pupils dilate, palms sweat, muscles contract in certain ways... Fear feels different from anger and love because it has a different physiological signature. The mental aspect of emotion, the feeling, is a slave to its physiology, not vice versa: we do not tremble because we are afraid or cry because we are sad; we are afraid because we tremble and sad because we cry (LeDoux, 1998, pp. 44-45).

LeDoux (1998, pp. 292-295) has recently proposed a more sophisticated version of the James-Lange theory, which he calls a "feedback theory" of the emotions. LeDoux holds that emotions are generated in the brain's amygdala, but acquire their distinctive feeling as a result of bodily feedback.

Theories which construe emotion in terms of bodily feedback do a good job of accounting for the second of the five features of animal emotions identified above - their bodily manifestation - as well as part of the third feature - their distinct kinds. Contrary to criticisms voiced by Cannon in 1929, we now know that the body's somatic feedback system has the requisite speed and specificity to account for the rapidity and diversity of our emotional responses (LeDoux, 1998, pp. 292-295). If we confine ourselves to emotions that do not presuppose language, each kind of emotion can be characterised in terms of its distinctive pattern of bodily response. For instance, fear can be characterised by an increase in heart rate and blood pressure, decreased salivation and increased perspiration, respiratory changes, scanning and vigilance, an increased startle reflex, defecation and either freezing (at low intensity) or flight (at high intensity) (LeDoux, 1998, pp. 144, 172; Panksepp, 1998, pp. 208, 213).

However, I would argue that no theory of emotions which describes them in terms of bodily feedback can explain why emotions are mental states and why they have intentional objects - the first and third of our five essential features of animal emotions.

Even though a "bodily feedback" theory allows us to construe emotions as unconscious mental states which are experienced bodily before being felt consciously by human beings (and possibly some other animals), the problem is that bodily reactions per se give us absolutely no warrant for regarding animal emotions as mental states of any kind, whether conscious or unconscious. It has been argued in Conclusion E.3 above that there is nothing about a bodily reaction as such that requires explanation in terms of mental states. A mind-neutral intentional stance can account for the behaviour observed. If we characterised emotions in terms of their bodily reactions, we could ascribe them to any organism capable of responding to an environmental stimulus, but at the cost of robbing them of their mental status.

Additionally, de Sousa (2003) argues that "feeling theories, by assimilating emotions to sensations, fail to take account of the fact that emotions are typically directed at intentional objects". Of course, bodily feelings have a cause, but that does not mean they have an intentional object. In any case, the cause of an emotion and its object may be quite different: if A, while drunk, gets annoyed at B over some trifling matter, drunkenness is the cause of A's annoyance, yet its object may be some innocent remark of B's, which merely occasioned (not caused) the annoyance (de Sousa, 2003).

Colin McGinn. Photo courtesy of Department of Philosophy, Rutgers, the State University of New Jersey.

In particular, bodily feelings, which are states of the body, cannot account for the fact that the intentional objects of emotions are generally states of affairs outside the body:

Suppose I am delighted that my son has become a doctor. I may have various sensations in my body that express this emotion - say, lightness in my limbs and a warm feeling in my viscera. But the object of my delight is not my body; it is my son's success. My bodily sensations are directed to my body and my emotion is directed to my son. Therefore my emotion cannot be identical to my bodily sensations - for the two have different objects (McGinn, 2003).

The James-Lange theory can be faulted on other philosophical grounds. First, the undeniable fact that bodily states can induce emotions does not show that they always or even typically do so. McGinn (2003) argues that "[t]here is causal interplay between feelings and their bodily expression, rather than a one-way dependence".

Second, the language we use to describe emotions does not translate into descriptions of bodily states. McGinn (2003) cites Wittgenstein's observation that the horribleness of my grief when someone I love dies cannot be explained as the horribleness of the sensations I feel in my body. The neurophysiological account of emotions defended above fares better here: on a generic level, grief is about the environmental challenge it arose to cope with - separation from a socially significant other (Panksepp, 1998, p. 50) - and in this particular case, it is about the death of the person I loved.

Antonio Damasio. Photo courtesy of University of Iowa Health Care.

Panksepp (1998) also criticises theories of emotion that are based on bodily reactions for their implicit cognitivism: emotions are supposed to arise from "our cognitive appraisal of the commotion that occurs in our inner organs during certain vigorous behaviors" (1998, p. 56). Thus Damasio (1994) differentiates emotions according to the entire pattern of somatic and visceral feedback from the body, arguing that feelings are "mental sensors of the organism's interior" (2002).

However, this characterisation overlooks the motivational aspect of emotions. There is an impressive body of evidence (Panksepp, 1998) that animals possess several basic motivational systems, which Panksepp refers to as emotional "operating systems". Each emotional system has its own neural circuitry and evolved to meet a special kind of environmental challenge, which motivates emotional behaviour in animals. These emotional systems in animals are thus defined by their brains, rather than their bodies. I will discuss these systems below.

Conclusion E.15 Internal bodily feedback cannot account for all of the essential features of animal emotions.

3. How do we identify basic emotions in animals, and which animals can be said to have them?

(a) Which basic emotions do animals have?

Deficient methodologies for distinguishing the emotions

Before discussing how we should differentiate animal emotions, I would like to explain why certain popular methods of distinguishing human emotions are unsuitable when applied to animals.

Linguistic analysis supports the existence of a short list of basic emotions. Most people, when asked to name the emotions they feel, mention love, anger, fear, sorrow and joy. Other emotions figure in less than 20% of responses (Panksepp, 1998, p. 46). The chief problem with a linguistic approach is that it cannot be applied to non-human animals. Another reason why language may be an unreliable guide to animal emotions is that some of the the emotions that have been proposed for animals by scientists (e.g. eagerness, play and nurturance) are not generally described as emotions.

Taxonomies of emotions based on facial analysis, starting with the work of Paul Ekman, have shown that six emotions - surprise, happiness (or joy), anger, fear, disgust and sadness (or distress) - have facial expressions that are universal across all cultures and can be clearly recognised in photos (Le Doux, 1998, p. 113; Evans, 2001, p. 7). However, most people would not consider surprise or disgust to be "proper" emotions. Additionally, Ekman's research focuses on only one species of animal: Homo sapiens. Panksepp (1998) argues that facial expression is a poor indicator of emotion in other animals:

Although in humans and some related primates the face is an exquisitely flexible communicative device, this is not the case for most other mammals, which exhibit clear emotional behaviors but less impressive facial dynamics. Although most animals exhibit open-mouthed, hissing-growling expressions of rage, and some show an openmouthed play/laughter display, they tend to show little else on their faces (1998, pp. 46-47).

Bodily indicators might seem to be a good general criterion for distinguishing animal emotions, as they could be reliably measured by sensors attached to an animal's body. However, as we saw above, this method of differentiating emotions fails to take account of their intentional objects.

How a neurophysiological approach to the emotions allows us to distinguish them according to their generic intentional objects

Since we have argued that each kind of emotion has its own generic intentional object which makes it what it is, and explains what it is about as well as what it is for, it would make sense to distinguish between animal emotions by identifying their respective objects. However, such a proposal faces three problems. Fortunately, a neurophysiological account resolves all of them.

First, it is possible for an animal to have different emotions towards the same object on separate occasions. For instance, an animal may fear a certain individual on one occasion and happily play with the same individual on another occasion.

A neurophysiological approach to the emotions resolves this paradox. An animal may indeed experience two different emotions towards the same physical object on separate occasions, because it may instantiate two different kinds of environmental challenge, and thus two distinct generic intentional objects. Thus a primate may fear an individual it knows, while (s)he is exhibiting a sudden burst of rage, but on other occasions enjoy playing with the same animal. (Of course, the complex human phenomenon of feeling mixed emotions - e.g. love and hate - towards the same individual simultaneously cannot be explained in this fashion.)

Second, we still have to account for emotions that have no intentional object, such as an ill-defined feelings of depression or the undirected emotional states that can be induced by removing parts of animals' brains. Surgical removal of cortical regions makes animals exhibit symptoms of rage which is not always directed at an appropriate target - sometimes, it appears to lack an intentional object - although Panksepp (1998, p. 79) suggests that the animals' lack of directed behaviour simply reflects their disorientation. Of course, one could argue that even in these situations, the kind of emotion experienced still has a generic intentional object. But this begs the question: how do we distinguish emotions according to their various kinds?

Finally, an emotion may have an intentional object which is real but inappropriate, such as when one fears mice or the number 13. What makes these inappropriate emotions of the same kind as their appropriate counterparts?

On a neurophysiological account, particular cases where an animal feels an emotion that is about nothing at all (an object-less emotion that may be naturally or articially induced), or nothing that merits the kind of response shown (inappropriate emotions) still qualify as bona fide instances of their kind, because they are accompanied by the same internal brain states as appropriate expressions of these emotions, and generate similar bodily responses. In these anomalous cases, the emotions felt by the animal are genuine, but the challenge they originally evolved to meet is absent.

What kinds of emotions do animals have?

The brain of every species of mammal contains various basic emotional systems. Panksepp (1998, pp. 48-49)defines these emotion systems in terms of the following features:

The first, second and fifth features are particularly relevant to my proposed method for classifying emotions according to their kinds. A suite of sensory stimuli that trigger identical unconditional responses in one of the emotion systems in an animal's brain all instantiate the same kind of environmental challenge that the animal has to deal with - i.e. its generic intentional object. The instinctive motor outputs triggered by the inputs are the animal's first line of defence. The fact that they can be subsequently modulated by cognitive inputs, means that provided the system satisfies the cognitive requirements for intentional agency (specified in chapter two), the system's response can be described as genuinely emotional and not merely reflexive.

Conclusion E.16 The emotion systems described by Panksepp (1998) correspond to basic kinds of animal emotions. The environmental challenge each system evolved to meet is the generic intentional object of the corresponding emotion.

Panksepp (1998) summarises the research to date on these systems:

There is good biological evidence for at least seven innate emotional systems ingrained within the mammalian brain (1998, p. 47).

"In the vernacular", writes Panksepp, the seven emotional systems include "fear, anger, sorrow, anticipatory eagerness, play, sexual lust, and maternal nurturance" (1998, p. 47). The first four systems are "evident in all mammals soon after birth" (1998, p. 54, italics mine), because the relevant neural circuits lie below the cortex, at a deeper level where the similarities between mammals are most profound. The neural circuits for our emotional systems are actually very ancient, since they arose in response to persistent external environmental challenges. The three special purpose systems - sexual LUST, maternal CARE and PLAY - are engaged only at certain times in mammals' life-cycles, and are not as clearly understood.

The classification of eagerness, play and nurturance as emotional states will strike many readers as odd. However, it makes good sense to do so if they motivate us to act as other emotions do. Panksepp also acknowledges the existence of many more affective feelings (e.g. pain, hunger), but argues that they do not qualify as basic emotional systems because they do not meet all the above criteria (1998, p. 47).

The seven emotion systems are summarised in the table below. (The capitals are Panksepp's.) More detailed information relating to accompanying bodily states and brain circuitry can be found in an Appendix.

Table 3.1 - Emotion Systems identified to date in Mammals (based on Panksepp, 1998)
Emotion System: FEAR system.

Animal Emotions corresponding to this system: Fear, anxiety, alarm and foreboding

Corresponding Environmental Challenge (generic intentional object): Pain and the threat of destruction

Motivation: Avoiding the threat of bodily harm

Characteristic behaviour: Freezing (mild intensity); flight (high intensity); scanning and vigilance


  • Contrary to earlier ideas that fear was a learned, conditioned response to cues that predict pain., FEAR is now known to be affectively distinct from pain: electrical stimulation of the brain's pain systems does not produce fear in humans, and electrical activation of the FEAR system in the brain does not appear to evoke the sensation of pain in either humans or animals (Panksepp, 1998, p. 215). Although animals fear painful stimuli, fear and pain are neurologically distinct.
  • FEAR is also distinct from PANIC. PANIC is thought to be linked with an animal's distress when separated from a socially significant other (e.g. its mother).

Emotion System: RAGE system

Animal Emotions corresponding to this system: Rage, anger

Corresponding Environmental Challenge (generic intentional object): Events that restrict the animal's freedom (physical restraint or irritation of the animal's body surface) or access to resources (e.g. an invasion of the animal's territory)

Motivation: The need to compete effectively for environmental resources.

Characteristic behaviour: Tendency to strike out at, attack, bite or fight the offending agent (a living creature).


  • Not all aggressive behaviour can be classified as RAGE. Of the three kinds of aggression circuits identified in the brain - corresponding to predatory aggression, intermale aggression and affective attack - only the last has the distinctive pattern of the RAGE system described above.
  • The SEEKING system mediates wanting as opposed to liking. It is not a pleasure system but an appetitive system.

Emotion System: PANIC system

Animal Emotions corresponding to this system: Loneliness, panic, grief

Corresponding Environmental Challenge (generic intentional object): Social loss

Motivation: The urge to be reunited with companions after separation, which helps to create social bonding

Characteristic behaviour: Cries of distress when separated from caregiver

Notes: PANIC is distinct from FEAR. PANIC is thought to be linked with an animal's distress when separated from a socially significant other (e.g. its mother).

Emotion System: Exploratory, appetitive SEEKING system

Animal Emotions corresponding to this system: Anticipatory eagerness

Corresponding Environmental Challenge (generic intentional object): Positive environmental incentives such as food, water, sex and warmth.

Motivation: The need for food, water, sex and warmth.

Characteristic behaviour: Stimulus-bound appetitive behaviour: forward locomotion, sniffing, and investigating, mouthing and manipulating objects in the animal's environment. Also self-stimulation - a tendency to engage in events that increase the animal's arousal.


  • Predatory aggression belongs to the mammalian brain's SEEKING system, not the RAGE system.
  • The SEEKING system mediates wanting as opposed to liking. It is not a pleasure system but an appetitive system.
  • The SEEKING system is a single emotion system, despite the diversity of targets (e.g. food, sex). Guided by regulatory imbalances in the animal's body, the SEEKING system triggers non-specific search behaviour that helps the animal obtain what it wants.

Emotion System: PLAY (special purpose system)

Animal Emotions corresponding to this system: Play

Corresponding Environmental Challenge (generic intentional object): Opportunity for rough-and-tumble play with conspecifics

Motivation: The need for social interaction

Characteristic behaviour: Rough-and-tumble (RAT) play between juveniles or between parent (usually the mother) and offspring. RAT play includes pinning and dorsal contacts, but varies widely among mammals. Solitary running, jumping, prancing and rolling in herbivores may also represent a form of play. Also laughter (humans) or very high-frequency chirping (rats).

Notes: The mammalian brain also contains one or more pleasure systems that correspond to the alleviation of bodily imbalances and a return to an optimal level of functioning. The location and number of the brain's pleasure systems remains unknown.

  • PLAY is quite distinct from aggression as well as SEEKING. It roughly corresponds to "joy". It may not represent a single system.
  • Emotion System: LUST (special purpose system)

    Animal Emotions corresponding to this system: Sexual desire

    Corresponding Environmental Challenge (generic intentional object): Opportunity to procreate

    Motivation: The need to procreate

    Characteristic behaviour: Males: courting, territorial marking and mounting.
    Females: decrease in aggression towards aroused males. Active tendency to solicit male attention. Receptive posture indicating readiness to be mounted (lordosis reflex).

    Emotion System: Maternal CARE (special purpose system)

    Animal Emotions corresponding to this system: Nurturance (parental love)

    Corresponding Environmental Challenge (generic intentional object): Offspring requiring maternal care

    Motivation: The need to care for one's offspring

    Characteristic behaviour: Responsiveness to distress signals by offspring. Nursing offspring and providing them with warmth and shelter (e.g. a nest). Gathering offspring together.

    The mammalian brain also contains one or more pleasure systems that correspond to the alleviation of bodily imbalances and a return to an optimal level of functioning. However, the location and number of the brain's pleasure systems remains unknown (Panksepp, 1998, pp. 181-185). Because the list of basic emotions described above is not exhaustive, the supposition that only these emotions correspond to natural kinds is unwarranted.

    It should also be stressed that neurological criteria alone cannot establish the presence of mental states such as emotions. To show this, it has to be demonstrated that each of the emotion systems described above can motivate intentional agency (Conclusion E.2). However, since mammals are certainly capable of intentional agency (as argued in chapter two) we can provisionally assume that they possess the seven-plus kinds of emotions described by Panksepp. Arguments that some of these emotions presuppose "higher" mental states unique to human beings are addressed below.

    Conclusion E.17 All mammals possess at least seven distinct kinds of emotions, whose neural circuitry is well-defined. These seven emotions can motivate intentional acts which promote the telos of a mammal in different ways. Each kind of emotion has a different generic intentional object.

    While Panksepp's approach to the emotions avoids anthropocentrism, it remains very focused on the most studied animals: mammals. I propose to use it as a starting point, bearing in mind the dangers of generalising to other animals.

    (b) Which Animals Have Emotions?

    Some of the brain's emotional architecture appears to be very old. A neurophysiological approach to emotions suggests that some are common to all vertebrates (and possibly other animals), while others may be unique to mammals and birds.

    Panksepp (1998, pp. 42-43, 48-51, 70-79) uses a refined version of Maclean's model of the triune brain for illustrative purposes, describing it as an "informative perspective" (1998, p. 43) which "provides a useful overview" (1998, p. 70), while conceding that it is a "didactic simplification from a neuroanatomical point of view" (1998, p. 43). The deepest layer of the forebrain, known as the basal ganglia, is where behavioural responses related to seeking, fear, anger and sexual lust originate. This region is well-defined in all vertebrates.

    Conclusion E.18 It is likely that all vertebrates share the (conscious or unconscious) emotions of fear, anger and sexual desire.

    The next, loosely defined layer is commonly called the limbic system - a term attacked as outdated by LeDoux (1998, pp. 98-103) but defended as a useful heuristic concept by Panksepp (1998, pp. 57, 71, 353). (Both authors agree that certain specific structures in the limbic system serve important functions relating to the emotions.) This region of the brain contains neural programs relating to social emotions such as maternal care, social bonding (companionship), separation distress, and playfulness (Panksepp, 1998, p. 71). The limbic system is of similar relative size across all mammals, but is much smaller in reptiles.

    Conclusion E.19 The social emotions, such as parental care, social bonding (companionship), separation distress, and playfulness, are likely to be absent in fish and amphibians.

    Surrounding the limbic system is the neocortex, which Panksepp describes as the "storehouse of our cognitive skills". This region is most developed in human beings, but is not where feelings originate: "We cannot precipitate emotional feelings by artificially activating the neocortex either electrically or neurochemically" (1998, p. 43), whereas subcortical stimulation of specific areas can induce anger, fear, curiosity, hunger and nausea in mammals (1998, p. 79).

    Conclusion E.20 There are no basic kinds of emotions that are unique to human beings. The only emotions that are specific to human beings are those whose cognitive requirements are beyond the capacities of other animals.

    It is hard to draw conclusions for invertebrates, as their brains have a very different layout. For instance, the amygdala, which which is thought to be responsible for the emotional evaluation of stimuli (Moren and Balkenius, 2000), is confined to vertebrates (Panksepp, personal communication, 11 April 2004). However, LeDoux claims that invertebrates "do the same thing with other circuits" (personal communication, 13 April 2004). Below, I propose criteria for identifying each kind of emotion, which are applicable to all animals.

    (c) A general strategy for identifying occurrences of basic emotions in animals

    Earlier, we argued that any positive or negative internal state that is capable of motivating animals acting intentionally is a genuine emotion (Conclusion E.2). It was argued in chapter two that operant conditioning, spatial learning, tool-making or social learning are all forms of intentional agency. This suggests one method of identifying genuine emotional states in animals.

    Conclusion E.21 An animal which can be motivated to undergo operant conditioning, spatial learning, tool-making or social learning in pursuit or avoidance of one of the motivators described in Table 3.1 is experiencing a mental state: the basic emotion corresponding to that motivator.

    However, since each basic emotion is mediated by an "emotion system" in the animal's brain that responds specifically to the relevant motivator, the emotion cannot be identified in an animal using behavioural criteria alone. An underlying brain system, specific to that motivator, must also be identified:

    Conclusion E.22 In order to conclusively establish that an animal is indeed undergoing a basic emotion, an emotion system in the animal's brain and/or nervous system has to be identified, which responds specifically to the relevant motivator.

    Scientists customarily describe these motivators as "rewards" and "punishments", which suggests that there is a single motivational pleasure-pain axis. I propose that there are in fact several axes - a fear axis, an anger axis, and a desire axis, among others - and thus several different ways to motivate animals undergoing operant conditioning and other forms of agent-centred learning.

    For example, an animal's ability to undergo operant conditioning in order to avoid a noxious or dangerous stimulus, strongly suggests that it is being motivated by fear. (The discovery of a circuit in the animal's brain or nervous system that responds specifically to danger would confirm this.) We saw in a case study in chapter two that the fruit fly Drosophila is capable of undergoing operant conditioning, to avoid being fried by a heat beam. The term "fear" could be appropriately applied here, even if phenomenological consciousness is absent.

    Conclusion E.23 The emotion of fear is likely to occur in insects capable of operant conditioning (e.g. fruit flies).

    Spatial learning, tool-making and social learning provide other settings in which fear can be identified behaviourally (and subsequently confirmed neurologically). An animal that learns by its own efforts how to navigate its way around a dangerous location, or how to use a tool in order to repel something dangerous, or how to manipulate the behaviour of others in its group, in order to reduce the risk of danger, can properly be said to be motivated by fear. (Of course, unlearned fears are no less mentalistic than those felt towards objects we have learned to fear. From an epistemological perspective, however, we can only identify unlearned fears as mental states in animals that are capable of undergoing operant conditioning or some other kind of learning that requires mental states.)

    Similar considerations apply to the other basic emotions, as shown in the table below.

    Table 3.2 Behavioural criteria for identifying intentional agency associated with each of the basic animal emotions

    Kind of Emotion How manifested in Intentional Agency (general description) How manifested in Operant conditioning How manifested in Spatial Navigation How manifested in Tool Use How manifested in Social learning
    FEAR Controlling and modifying one's behaviour in order to avoid danger Ability to modify one's bodily movements in order to avoid danger Ability to modify one's route in order to avoid danger Ability to manipulate objects in order to avoid danger Ability to model one's behaviour on others in order to avoid danger
    General experimental strategy for identifying the occurrence of FEAR Would the experiment be ethical? No (see below). Place an animal in a situation where it has to modify its bodily movements in order to avoid a predator or a noxious stimulus Place an animal in a situation where it has to modify its route in order to avoid a predator or a noxious stimulus Place an animal in a situation where it has to manipulate some object in order to avoid a predator or a noxious stimulus Place an animal in a situation where it has to model its behaviour on another animal in order to avoid a predator or a noxious stimulus
    RAGE or anger Controlling and modifying one's behaviour in order to get an opportunity to attack an offending agent/object Ability to modify one's bodily movements in order to get an opportunity to attack an offending agent/object Ability to modify one's route in order to get an opportunity to attack an offending agent/object Ability to manipulate objects in order to get an opportunity to attack an offending agent/object Ability to model one's behaviour on others in order to get an opportunity to attack an offending agent/object
    General experimental strategy for identifying the occurrence of RAGE Would the experiment be ethical? No (see below). Place an animal in a situation where it has to modify its bodily movements in order to get the opportunity to attack an offending agent/object Place an animal in a situation where it has to modify its route in order to get the opportunity to attack an offending agent/object Place an animal in a situation where it has to manipulate some object in order to get the opportunity to attack an offending agent/object Place an animal in a situation where it has to model its behaviour on another animal in order to get the opportunity to attack an offending agent/object
    Anticipatory Eagerness (SEEKING) Controlling and modifying one's behaviour in order to attain an attractive object Ability to modify one's bodily movements in order to attain an attractive object Ability to modify one's route in order to attain an attractive object Ability to manipulate objects in order to attain an attractive object Ability to model one's behaviour on others in order to attain an attractive object
    General experimental strategy for identifying the occurrence of SEEKING Would the experiment be ethical? Yes. Place an animal in a situation where it has to modify its bodily movements in order to attain an attractive object Place an animal in a situation where it has to modify its route in order to attain an attractive object Place an animal in a situation where it has to manipulate some object in order to attain an attractive object Place an animal in a situation where it has to model its behaviour on another animal in order to attain an attractive object
    PANIC Controlling and modifying one's behaviour in order to obtain comforting social contact Ability to modify one's bodily movements in order to obtain comforting social contact Ability to modify one's route in order to obtain comforting social contact Ability to manipulate objects in order to obtain comforting social contact Ability to model one's behaviour on others in order to obtain comforting social contact
    General experimental strategy for identifying the occurrence of PANIC Would the experiment be ethical? No (see below). Place an animal in a situation where it has to modify its bodily movements in order to obtain comforting social contact Place an animal in a situation where it has to modify its route in order to obtain comforting social contact Place an animal in a situation where it has to manipulate some object in order to obtain comforting social contact Place an animal in a situation where it has to model its behaviour on another animal in order to obtain comforting social contact
    LUST Controlling and modifying one's behaviour in order to get an opportunity to satisfy sexual urges Ability to modify one's bodily movements in order to get an opportunity to satisfy sexual urges Ability to modify one's route in order to get an opportunity to satisfy sexual urges Ability to manipulate objects in order to get an opportunity to satisfy sexual urges Ability to model one's behaviour on others in order to get an opportunity to satisfy sexual urges
    General experimental strategy for identifying the occurrence of LUST Would the experiment be ethical? Yes. Place an animal in a situation where it has to modify its bodily movements in order to get an opportunity to satisfy its sexual urges Place an animal in a situation where it has to modify its route in order to get an opportunity to satisfy its sexual urges Place an animal in a situation where it has to manipulate some object in order to get an opportunity to satisfy its sexual urges Place an animal in a situation where it has to model its behaviour on another animal in order to get an opportunity to satisfy its sexual urges
    Maternal CARE Controlling and modifying one's behaviour in order to get an opportunity to nurture offspring Ability to modify one's bodily movements in order to get an opportunity to nurture offspring Ability to modify one's route in order to get an opportunity to nurture offspring Ability to manipulate objects in order to get an opportunity to nurture offspring Ability to model one's behaviour on others in order to get an opportunity to nurture offspring
    General experimental strategy for identifying the occurrence of CARE Would the experiment be ethical? No (see below). Place an animal in a situation where it has to modify its bodily movements in order to get an opportunity to nurture its offspring Place an animal in a situation where it has to modify its route in order to get an opportunity to nurture its offspring Place an animal in a situation where it has to manipulate some object in order to get an opportunity to nurture its offspring Place an animal in a situation where it has to model its behaviour on another animal in order to get an opportunity to nurture its offspring
    PLAY Controlling and modifying one's behaviour in order to engage in rough-and-tumble play Ability to modify one's bodily movements in order to get an opportunity to engage in rough-and-tumble play Ability to modify one's route in order to get an opportunity to engage in rough-and-tumble play Ability to manipulate objects in order to get an opportunity to engage in rough-and-tumble play Ability to model one's behaviour on others in order to get an opportunity to engage in rough-and-tumble play
    General experimental strategy for identifying the occurrence of PLAY Would the experiment be ethical? Yes. Place an animal in a situation where it has to modify its bodily movements in order to get an opportunity to engage in rough-and-tumble play Place an animal in a situation where it has to modify its route in order to get an opportunity to engage in rough-and-tumble play Place an animal in a situation where it has to manipulate some object in order to get an opportunity to engage in rough-and-tumble play Place an animal in a situation where it has to model its behaviour on another animal in order to get an opportunity to engage in rough-and-tumble play

    It should be stressed that all of the above experimental identifications of basic emotions are tentative, and subject to the identification of a dedicated neurological system in the animal's brain (see Conclusion E. 19).

    The reader will have noticed that I have described four of my seven proposed strategies for experimentally identifying of basic emotions in animals as unethical, if carried out. It is not often that a researcher counsels against an experiment he/she has proposed, but there are good reasons to believe that animals dislike being made to feel afraid, angry, lonely or worried about their offspring. According to Panksepp, most animals "readily learn to turn off" electrical stimulation of the brain which artificially induces affective attack, or RAGE, so "we can conclude that most animals do have unpleasant affective experiences during such stimulation" (1998, p. 194). In other words, to induce anger in an animal is to do it emotional harm. I shall explain in chapter five why I believe such conduct to be wrong.

    It would of course be easy to limit the harm done. For instance, one can design non-invasive experimental techniques to identify anger, such as the following:

    Put the animal in a box. Place another animal (actually a lifelike toy) inside its territory and give the animal a chance to attack the toy, but only as long as it moves in a specific way (e.g. run at a certain speed in a specific direction). If it fails to do so, a clear plastic partition comes down and separates the animal from the intruder, which it can still see occupying its territory.

    Such a set-up would certainly prevent physical cruelty. However, the experimental animal would still be subjected to unpleasant stress. For this reason, I suggest that research would be better directed at identifying the three emotion systems in animals (seeking, lust and play) where ethical research can be performed.

    Seeking: the animal has to learn a specific strategy (which varies from day to day) for obtaining its food. (Already under way.) Variants: spatial learning; requirement for tool use; requirement for co-operation.

    In experiments for identifying sexual desire, the animal would have to has to learn a strategy in order to obtain access to and gain the opportunity to mate with a suitable partner.

    In play experiments, the animal would have to learn a strategy for obtaining access to its playmates. For instance, the animal could be placed on the other side of a clear plastic partition, giving it a full view of them while they were engaged in play. In order to access its playmates, the animal might have to learn to perform some complicated manoeuvre (e.g. press a lever in a particular way), or successfully navigate a maze, or use some object as a tool to open the partition, or let itself be guided to the other side by a video featuring another animal.

    4. Emotions and the First Person Intentional Stance: Which Animals are Conscious, and When?

    In what follows, I review the major arguments for and against phenomenal consciousness in non-human animals, and briefly discuss the significance of animal consciousness in everyday human life.

    In my discussion of methodology in chapter two, I proposed that we should adopt a mind-neutral intentional stance except where the adoption of a mentalistic stance enables us to make better scientific predictions than we otherwise would. The implications for our search for conscious emotions are that we do not need to look for events that can only be understood from a first-person standpoint; rather, we need to identify events that are easier to understand and predict scientifically from this standpoint.

    The key points I wish to make here, after reviewing the literature on animal emotions, are as follows:

    Non-Inferential Arguments

    Searle (1998) rejects attempts to infer the existence of conscious states in animals as vestiges of a dualistic mindset, according to which we must infer hidden mental causes from an individual's outward behaviour. Our belief in animal consciousness is a properly basic belief:

    [I]t doesn't matter really how I know whether my dog is conscious, or even whether or not I do know that he is conscious. The fact is, he is conscious and epistemology in this area has to start with this fact (Searle, 1998, p. 50).
    Allen (2003) criticises this position on the grounds that it fails to settle in practical terms the question of which non-human animals possess consciousness and which ones lack it (the Distribution Question). Where does one draw the line? As a biological naturalist, Searle is confident that science will one day tell us how lower level neural phenomena cause consciousness, and that when we know the "electrochemical" formula for the necessary and sufficient conditions for consciousness, we shall be able to apply it to difficult cases like snails (1998, pp. 45-47). Allen (2003, pp. 7-8) thinks it unlikely that such a formula will yield clearcut answers, and in the meantime, we have to make practical moral decisions as to what kind of legal protection we should accord different kinds of animals. Searle's non-inferential approach is incapable of telling us how to proceed here.

    Nevertheless, I would argue that for a small number of companion animals, Searle is right. What I am proposing here is a transcendental argument, which takes as its starting point a fact of everyday life: the fact that we can befriend cats and dogs. To affirm this is to affirm the conditions that make it possible - including the fact that they are phenomenally conscious. (There can be no friendship without the possibility of sharing mutually enjoyed experiences.) But it does not follow from this that most other animals (e.g. hippopotamuses) are conscious.

    Conclusion E.20 While the belief that selected animals (e.g. cats and dogs) are conscious is a psychological "given", the same does not apply to most other animals.

    Inferential arguments

    Look at arguments FOR animal phenomenal consciousness (see Allen's 2002 and 2003 articles)

    (1) Similarity arguments - neurophysiological & behavioural
    (2) Evolutionary arguments. Pain and other feelings evolved because of their survival value.
    (3) Inference to the best exlanation
    (4) Non-inferential awareness (Searle). I "just know" my dog is conscious. It doesn't matter how I know. Problem: which animals are conscious? Where do you draw the line?
    (5) Argument from illusion - animals are susceptible to the same perceptual illusions as we are. Extension: some of them can be trained to make corrections for their illusions.
    (6) Argument from efficacy of conditioning (Bermudez, 2000). It is impossible to divorce something's being a positive (negative) reinforcer from its feeling good (bad).
    (7) "Affective distortions" - extreme behaviours such as phobias, intense rage, grief, addiction - which make no sense except in the light of animal feelings.

    (8) Attention in fruit flies - Report in New Scientist, 14/2/2004.

    If we do not, after all, "just know" that most animals possess phenomenal consciousness, then it seems we must infer its presence or absence. Similarity arguments invoke some human or mechanical analogue to explain the behaviour of non-human animals, and use the similarities between the analogue's features and those of non-human animals to justify their conclusions about animal awareness, while dissimilarity arguments do the opposite and argue from disanalogies.

    Similarity arguments for animal awareness in non-human animals use human beings as the analogue, and trade on the numerous resemblances between our anatomy, behaviour, neurobiology and pharmacology and theirs. Similarity arguments against animal consciousness point to similarities between animal behaviour and that of machines or human individuals that lack awareness. Dissimilarity arguments are usually cited to prove that other animals do not possess awareness as we do. These arguments may be based on disanalogies between the behaviour of conscious human beings and that of certain animals, who are then said to lack awareness; or on the absence of some mental attribute in non-human animals, which is then alleged to be a pre-requisite for consciousness.

    Both kinds of arguments have been brought to bear on the subject of animal minds for at least the past 2,500 years, so it may help if we first examine the limitations of these arguments. Similarity arguments that make use of some human or mechanical analogue certainly work if all of the similarities between animals and the analogue are relevant to the capacity for consciousness, and none of the dissimilarities are relevant. In reality, life is not that simple, and we find both relevant similarities and relevant dissimilarities. In that case, one can attempt to "weigh up" the similarities and dissimilarities, to see which predominates.

    A problem arises when the similarities and dissimilarities are incommensurable, as they relate to different sets of features. This is what I found in my search of the literature relating to other animals' capacity to experience pleasure and pain: on the one hand, there are strong behavioural similarities between human beings and other animals (especially mammals, birds and reptiles), suggesting that consciousness may be widely distributed in the animal kingdom; and on the other hand, massive neurological dissimilarities between the brains of mammals and other animals, which seem to preclude these other animals from having conscious experiences of any kind.

    A review of the behavioural evidence relating to conscious pleasure and pain

    When our chosen indicators for animal consciousness conflict, we have no choice but to reject one set of attributes as an unreliable indicator of animal awareness, and make the other set of features a definitive yardstick for consciousness. In so doing, we have to reject some of our basic intuitions about consciousness.

    The question then arises: which set of features do we select as our yardstick? The fact that arguments have raged on this matter for centuries suggests that conflicting intuitions are at work here, so it may help if we ask why. The problem, I suggest, is that there is no generally agreed answer to the question of what consciousness is for. If we knew that consciousness is for X-ing, we could resolve the Distribution Question simply by determining which animals can and cannot X, but there is still no agreed answer as to what X might be.

    Since consciousness is said to be a first-person phenomenon, we might try to find X by looking for a kind of behaviour that could only be explained on a first-person account. In practice, however, what we find is that an alternative, third-person account is possible for all of the various kinds of behaviour that are usually cited as evidence of animal awareness. It is difficult, perhaps impossible, to envisage a genre of behaviour that requires subjectivity.

    At this point, we can react in one of two ways: (a) decide that we have set the bar too high and plump for some more modest criterion, or (b) decide that non-human animals are not conscious after all. Both choices appear unpalatable: the former is an intellectually soft option, while the latter appears too sceptical.

    I suggest that philosophers should swallow their pride and admit that their conceptual tools have failed to resolve the deadlock. A new tack is needed. The most promising course of action, I would argue, is to find out where consciousness is located in the human brain, and why those regions are so special. This might help us understand what consciousness is for. After that, we can decide which animals possess it by ascertaining whether they can perform the requisite feats.

    Indicators of Pleasure and Pain

    Simple behaviour patterns

    Stress responses

    All cellular organisms exhibit some form of stress response. For instance, the bacterium Bacillus subtilis exhibits a stress response to heat shock, salt stress, ethanol, starvation for oxygen or nutrients, and so on, which is mediated by the same set of general stress proteins, as well as specific stress proteins (Hecker, Schumann and Volker, 1996). Plants respond to stress by releasing ethylene all over their surfaces, which promotes cell growth and other restorative responses. However, these responses can be described in third-person terminology, and there appears to be no scientific advantage in re-describing all of these stress responses as episodes of conscious pain.

    Conclusion E.21 A stress response is not a sufficient criterion for the occurrence of conscious pain.


    A variety of animals exhibit a trait called nociception. Nociception includes the sensory detection of potentially injurious stimuli by specific receptors known as nociceptors, the transmission of information within the nervous system, and the resulting response. (The protozoan Paramecia exhibits an avoidance response when poked with a needle, but because it is triggered by changes in electrical activity at the cell surface membrane rather than a nervous system, this response is not considered to be nociceptive.) Among invertebrate animals, only sponges are thought to lack nociception; even Cnidaria such as sea anemones (which lack a brain) respond to aversive mechanical, electrical and chemical stimuli. The roundworm C. elegans exhibits a nociceptive heat response to an acute heat stimulus (Wittenburg and Baumeister, 1999), as do earthworms, leeches, insects, snalils and octopuses (Smith, 1991). Surprisingly, one group of vertebrates, the elasmobranchs (sharks and rays) "lack the neural structures for processing nociceptive information, much less sensing pain" (Rose, 2002, p. 22). It is thought that sensitivity would be a disadvantage for these fish, as they often feed on prey with embedded barbs, such as stingrays.

    However, the biological function of nociception can be explained using third-person terminology, without recourse to conscious states such as pain.

    Confirming evidence from human studies indicates that nociception is processed at a subconscious level which can be described using to third-person terminology. In human beings, spinal cord neurons send axons to a cluster of neurons in the brainstem, known as the reticular formation. This network processes the nociceptive information, sends it to various subcortical brain structures, and also generates the suite of complex but innate behavioural responses to nociceptive stimuli (Rose, 2002, p. 17). It is important to understand that none of this occurs at the conscious level, as human beings are never aware of the neural activity taking place below the level of the cortex - whether it be in the spinal cord, brainstem or cerebral regions beneath the neocortex (Rose, 2002, p. 6). Brain monitoring techniques have shown researchers that human beings consciously perceive only information that is processed in special regions of the cerebral cortex, known as the associative regions, and that activities processed outside the cortex cannot be accessed by consciousness (Roth, 2003, p. 36).

    Conclusion E.22 Nociception per se is not a sufficient criterion for consciousness.

    Complex nociceptive responses in human beings lacking consciousness

    Studies of brain-damaged human beings show that complex nociceptive responses to stimuli, which are mediated by the spinal cord and brain stem, can occur in the absence of consciousness. People in a persistent vegetative state can make organised responses to nociceptive stimuli and appear to be wakeful. People born without cerebral hemispheres still exhibit nociceptive behavioral responses: when a limb is stimulated, they will withdraw the limb, vocalise, exhibit facial grimaces, and release hormones and neurotransmitters associated with pain. In other words, they exhibit wakefulness without consciousness (Rose, 2002, pp. 14, 17).

    Conclusion E.27 Being awake is not a sufficient criterion for consciousness.

    Conclusion E.28 A complex nociceptive response to noxious stimuli is not a sufficient condition for the occurrence of conscious pain in animals.

    Presence of Opiate Receptors

    Many animals, including earthworms, molluscs and insects, manufacture opioid substances such as enkephalins and beta-endorphins - which are known to deaden pain in humans - in their bodies. These creatures also have nerve receptors that respond to opiates. However, the mere presence of opiates in an animal's body is insufficient to establish that it can suffer pain: the substances in question serve other biological functions as well as pain relief, and it is believed that in evolutionary terms, opioids originated in order to attack bacteria and send signals to the immune system (Stefano, Salzet and Fricchione, 1998).

    Pharmacological criteria cannot yield decisive evidence of pleasure or pain, because it is possible that the molecules which modify our responses to pleasure and pain originally evolved for another purpose. In fact, this appears to be the case. Stefano, Salzet and Fricchione (1998) note that the association between pain-killing opioids and anti-bacterial compounds goes back at least 500 million years, as both are found in invertebrates as well as vertebrates. In addition, there are chemical affinities between opioids and bacteriocides: pro-enkephalin, a naturally occurring analgesic molecule, contains an anti-bacterial peptide named enkelytin. The authors remark:

    In the process of immune defense or neural activation, a bacteriocidal compound, enkelytin, may be released along with the opioid peptides (1998, p. 267).

    As it turns out, opioid peptides do more than alleviate pain: they also serve to stimulate immunocytes, which stage an immune reponse in the body.

    Stefano, Salzet and Fricchione (1998) argue that bacteria and viruses are and always have been persistent threats to animals, which had to develop means of combating these threats. The authors hypothesise that the association of enkelytin and opioids is part of such a strategy, and that the function of enkelytin is a dual one: to attack bacteria and allow time for other substances (opioid peptides) to stimulate the immune system, while an animal is orienting itself to an invasion by bacteria. Pain may have evolved later, as a means of alerting the animal to the presence of a noxious stimulus such as bacteria. They conclude:

    The reason for the evolving relationship between opioid neural and immune processes now appears quite simple, that is analgesic priority-setting activities associated with an anti-infectious / anti-inflammatory process. This combination would provide a high degree of survival benefit to any organism since it would ensure appropriate behavior to meet not only these non-cognitive challenges but also cognitive ones (1998, p. 267).

    Additionally, the administration of man-made opioid substances such as morphine or nalaxone produces analgesic effects in these animals and can reduce or abolish their responses to noxious stimuli (Smith, 1991). But here again, we need to ask: is the animal's response a true pain response, or should it be described in more neutral terms? It is easy to envision how opiates might serve a useful role in modulating animal behaviour, even in the absence of consciousness. For example, an animal with an injured leg benefits from nociceptive reflexes that encourage keeping weight off the leg, but the animal is best served by suppressing this reflex with opiate-like neurotransmitters when being chased by a predator.

    Conclusion E.23 The presence of opiate receptors in animals' brainstems is an insufficient criterion for the occurrence of conscious pain.


    The phenomenon of pain-guarding, in which an animal shows protective behaviour towards an injured part, has been cited as evidence that animals feel pain (Grandin and Deesing, 2002).

    I would argue that the complete absence of pain-guarding in certain animals can reasonably be taken as evidence that they lack the capacity for pain. It is hard to see how we can still meaningfully speak of a creature as being in pain if it shows no inclination to protect an injured body part.

    Conclusion E.24 Absence of pain-guarding in certain kinds of animals is strong evidence that they are incapable of feeling conscious pain.

    Smith (1991) cites a review of the biological evidence concerning pain in insects:

    No example is known to us of an insect showing protective behavior towards injured parts, such as by limping after leg injury or declining to feed or mate because of general abdominal injuries. On the contrary, our experience has been that insects will continue with normal activities even after severe injury or removal of body parts.

    Conclusion E.25 Insects (and by extension, worms, whose nervous systems are simpler) are incapable of feeling conscious pain.

    The phenomenon of pain-guarding is well-documented among mammals and birds, and there is tentative but conflicting evidence of its occurrence in reptiles (Grandin and Deesing, 2002). A recent, well-publicised report by Sneddon, Braithwaite and Gentle (2003), claiming to have identified pain-gaurding in fish, has been subjected to a devastating critique by Rose (2003a) (see Appendix). I have not been able to locate any reliable accounts of pain-guarding in amphibians, fish or cephalopods, although Grandin and Deesing (2002) discuss a few cases of pain-guarding in fish that may alternatively be due to physical illness or fear. It would be unwise, given our present lack of knowledge, to be dogmatic regarding these animals, as "absence of evidence is not evidence of absence".

    Pain-guarding would seem to constitute strong prima facie evidence of pain in animals. On the other hand, the biological function of pain-guarding can be explained without recourse to a first-person account, let alone conscious states. For example, an animal with an injured leg benefits from nociceptive reflexes that encourage keeping its weight off the leg.

    Conclusion E.26 Pain-guarding per se is not a sufficient condition for the occurrence of conscious pain.


    A recent well-publicised report by Sneddon, Braithwaite and Gentle (2003) claims to have identified evidence of pain guarding in fish. Administration of bee venom to the lips of trout affected both their physiology and behaviour. Fish injected with venom exhibited significantly increased respiration, rubbed their lips against gravel and performed a characteristic sideways "rocking" behaviour. In response, Rose (2003a) has written a devastating critique of the report's findings. Briefly, Rose:

    (a) acknowledges the occurrence of nociception in bony fish;
    (b) argues forcefully that the behaviour exhibited by the trout injected with bee venom is inconsistent with pain guarding, and if anything indicates oral insensitivity on their part;
    (c) argues that Sneddon et al. (2003) used a faulty definition of pain of pain in their article - instead of relying on the definition used by the International Association for the Study of Pain ("pain is a conscious experience, with a sensory component and a component of emotional feeling (suffering)"), they considered any form of nociception which is more complex than a reflex to be evidence of pain. Rose argues that this way of distinguishing pain from nociception is invalid because there are clearly complex, non-reflexive behaviors (exhibited by decorticate human beings) that can be purely nociceptive and unconscious.

    Pain-guarding revisited

    The fact that mammals may exhibit pain guarding of a limb even when it is structurally sound and capable of bearing weight (Grandin and Deesing, 2002) is not readily explicable in biological terms. A first-person account seems more appropriate here: we could say that the behaviour is indicative of pain.

    Conclusion E.29 Pain-guarding of a structurally sound limb is prima facie evidence of conscious pain.

    Satiety in animals

    Mammals also exhibit the phenomenon of satiety: both rats and humans find sweet liquids less appealing just after a meal than when hungry, and make the same appetitive judgements as humans do, depending on how much sugared water they have drunk: their pattern of changing preferences is indistinguishable from that of people (Vines, 1994). However, satiety in rats and humans is likely to be explicable in terms of underlying chemical processes.

    Conclusion E.33 The phenomenon of satiety need not indicate conscious pleasure in animals.

    (ii) Actions said to be indicative of pleasure and pain

    Self-administration of analgesics

    Cases where injured animals will actively seek out analgesic drugs would seem to suggest that they are in pain. Grandin and Deesing (2002) describe such a case in rats:

    Colpaert et al. (1980, 1982) performed a series of very important experiments which showed that rats with chronic inflammation of the joints will drink water containing an analgesic instead of a sweet solution that control rats preferred. The rats' intake of fentanyl analgesic followed the time course of arthritis that was induced with an inoculation with Mycobacterium butycium (Colpaert et al. 2001). This study clearly shows that rats drank the medication to reduce pain and not for its rewarding effects. Because the rats choose water containing an analgesic which possibly tasted bad compared to the highly palatable sweet solution shows that self-administration of pain relief may be taken as evidence that rats experience pain and suffer in a way similar to humans.

    However, a more minimalistic interpretation is also possible: the rats' preference for opiates and willingness to tolerate a bitter taste in exchange for pharmacological relief is the result of some chemical "weighing-up" process in their bodies, rather than a mental evaluation of the pros and cons of imbibing bitter medicine. Before we adopt a first-person stance to explain these phenomena, we should ask whether it makes any useful prediction that a third-person stance would not.

    Weighing-up processes are also known to occur in bacteria: if E. coli's sensors detect an attractant (e.g. galactose), and later sense another compound (e.g. glucose) that is more attractive than the first one, a "weighing" of the relative quality of the nutrients occurs, and the chain of reactions resulting in directed motion is amplified. The co-presence of attractants and repellents in solution generates an integration of the "run" and "tumble" responses, at the chemical level (so-called "conflict resolution"). However, as Kilian and Muller (2001, p. 3) point out, the way in which bacteria react to a chemical is utterly inflexible, at the molecular level, and the apparently complex behaviour of bacteria in response to multiple simultaneous stimuli (positive and/or negative) is merely the resultant of two or more inflexible existing action patterns (built-in preferences). The behaviour of the bacteria can be perfectly well described using a third-person intentional stance.

    There is no good philosophical reason to adopt a first-person stance to account for injured animals' willingness to self-administer analgesics, unless we find evidence of behavioural flexibility in animals weighing up their options.

    Conclusion E.30 An injured animal's preference for a bitter solution containing analgesics over a sweet solution does not constitute conclusive evidence that it is consciously experiencing pain.

    Animals can learn to avoid noxious stimuli

    Bermudez (2000) has argued that the ability of many animals to undergo conditioning can only be explained by the fact that the unconditioned stimuli (primary reinforcers) with which new behaviour patterns are associated feel pleasant or unpleasant:

    [L]earning through conditioning works because primary reinforcers have qualitative aspects. It is impossible to divorce pain's being a negative reinforcer from its feeling the way it does (Bermudez, 2000, p. 194).

    But as we saw in chapter two, associative learning does not require a mind, let alone consciousness. We could explain the behaviour of conditioned animals using third-person terminology: reinforcers work because animals have innate drives to seek or avoid them.

    For this reason, I cannot agree with Cabanc (2003) when he claims that the ability of animals (including mammals and lizards but not frogs and toads) to learn after a single exposure to avoid foods whose taste they associate with subsequent illness, constitutes evidence that they consciously remember a painful experience. Nor does the inability of frogs and toads to form such associations prove they lack consciousness: other factors, such as physiology or a shorter memory span, may be responsible.

    Conclusion E.31 An animal's ability to undergo associative learning is not a sufficient criterion for the occurrence of conscious pleasure or pain.

    Tendency of animals to engage in self-stimulation

    The tendency of animals to engage in self-stimulation has been cited as evidence that they feel pleasure, as it seems to require an animal to pay attention to its own internal states, rather than external objects. The animal may have to perform a complicated action (press a lever) in order to prolong the arousal of its brain's SEEKING system - hence it needs to "pay attention" to both its arousal state and its current course of action (which it needs to modulate to prolong the arousal state).

    Self-stimulation has been identified in a wide variety of animals, including fish, crustaceans, and even snails (Panksepp, peronal communication, 30 May 2004). (In these cases, however, no complex motor behaviour, such as pressing a lever, was required on the animal's part.) We could say that the animals are seeking to prolong their arousal, but non-mentalistic explanations are possible: the animal is now in some automated "do loop", or the instinctual arousal of seeking and the consequent affect keeps the animal magnetised in a repetition compulsion (Panksepp, personal communication, 30 May 2004).

    Conclusion E.32 An animal's tendency to engage in self-stimulation is an insufficient criterion for the occurrence of conscious pleasure.

    Trade-offs and Relative Rankings of Goods by Animals

    More suggestive is the willingness of animals to make trade-offs whereby they expose themselves for a short time to an aversive stimulus in order to procure some attractive stimulus. It has been shown that lizards are willing to leave a warm refuge, where they were supplied with standard food, and venture out into a cold environment, in order to acquire a more palatable food (lettuce). The lizards sought out the lettuce, even though they did not need the food. Additionally, they appeared to weigh up the relative costs and benefits of their choices: when it got too cold, the lizards stayed in their warm enclosure and ate the nearby food, but as experimenters improved the quality of the food in the cold corner, the lizards proved willing to tolerate lower temperatures (Cabanac, 2003). The researchers concluded that the lizards were making decisions based on palatability, a form of pleasure.

    Cabanac (2003) reports that "all aspects of palatability reported by humans can be found in rats as well, including decision making in conflicts of motivation, palatability vs. cost".

    Researchers such as Marian Dawkins (1994) have also found ways of ranking animals' desires for different "goods", by measuring how much they are willing to work (e.g. peck a key) to obtain each good, or alternatively, how much discomfort they are willing to put themselves through in order to obtain various goods. For instance, hens are averse to squeezing through narrow gaps, and even a hungry hen will not squeeze through a 9 centimetre gap to get food, but will readily do so to obtain access to a floor that is suitable for scratching or dust bathing (Vines, 1994).

    From an economist's perspective, the behaviour described above probably meets the requirements for wanting, as the strength of animals' desires for different goods allows economists to construct utility curves. All that would be needed to complete the picture would be evidence to animals' willingness to exchange one combination of goods for an equally desirable combination.

    McKee, M. 2004. "Material Girls." In California Wild, the magazine of the California Academy of Sciences, spring 2004 edition. Web address: stories/materialgirls.html.
    (Maggie McKee is a science writer from Washington DC.)

    Since animals' short-term appetitive behaviour is so similar to our own, it would seem churlish to deny the overwhelming behavioural evidence that these animals experience conscious likes and dislikes. However, it has not been shown that a first-person account yields better scientific predictions than a third-person account that employs more neutral terminology. Cabanc himself employs such terminology: the lizards face "conflict between two motivations: a thermoregulatory drive (to avoid cold) and an attraction to palatable bait" (2003).

    Rational and irrational pursuit in animals

    Berridge (2001) suggests ways of distinguishing between rational and irrational choice in humans and animals:

    Let's grant at the outset that the rationality or irrationality of your choice has nothing to do with why you like it, or with whether anyone else likes it too. The question of rationality hinges only on whether your choice consistently follows your expectations of hedonic likes.

    Thus an animal is choosing rationally when it chooses what it expects to like, even if its expectations happen to be wrong. Rational pursuit may be manifested when animals are trained to work for real rewards, which come only sporadically, so the animals learn to persist in working for a reward. Under extinction conditions, when the rewards no longer come at all, the animals will keep working for quite some time because they still expect the reward: they have learned that perseverance pays off.

    Irrational pursuit, on the other hand, occurs when an animal desires something it neither likes nor expects to like:

    The notion of irrational choice may seem to be self-contradictory when viewed from the perspective that people always choose what has the most value or decision utility to them... However, as documented by a number of authors..., people may sometimes choose an outcome whose eventual hedonic value does not justify their choice...

    Irrational pursuit can be identified when an animal, under the influence of some drug (e.g. dopamine), is suddenly presented with the rewarding stimulus, which cues hyperactive pursuit of the stimulus.

    Conclusion E.29 Only mammals and birds feel pain.

    Affective distortions

    By affective distortions I mean extreme behaviours - such as addiction, phobias, intense rage and grief - which make no sense except in the light of animal feelings.

    A scientific test for subjectivity

    Consider the nineteenth century scenario of two healthy young bachelors who engage in a pistol duel in order to win the hand of a woman they both ardently wish to marry. From a purely biological standpoint, a fight to the death is absurdly irrational: each man faces a high probability of dying without leaving any children behind him if he fights, but is still likely to leave a large number of descendants behind if he refrains from duelling and marries someone else. Only if marrying the woman would double each suitor's expected number of descendants would it be biologically "rational" for him to expose himself to a 50% risk of sudden death. In reality, of course, the suitors do not weigh the risks in this way, because they have feelings. Each man mistakenly believes that he will be heartbroken forever if he does not marry the woman, and each confidently believes (buoyed by a false sense of optimism) that his prospect of victory is much greater than 50%.

    The foregoing example suggests one way of identifying kinds of behaviour which manifests feelings in animals:

    (1) The kind of behaviour in question should not be "hard-wired". In particular, it should satisfy the requirements for one of the four varieties of intentional agency, described in chapter two (see DF. 1 to DF. 4). Roughly speaking, the behaviour is voluntary in the sense that Aristotle would have allowed for animals.

    (2) The behaviour should be undertaken in pursuit of some goal that the animal might find rewarding (e.g. a dainty morsel of food, or a desirable mate).

    (3) The behaviour should be risky to the animal's biological prospects (of leaving descendants).

    (4) The biological risks should "outweigh" the biological benefits: the animal can expect to be biologically worse off (as measured by its long-term number of descendants) for engaging in the behaviour.

    (5) The behaviour should be systematic - i.e. typical of a species and not just an individual. An individual may behave erratically, but if a species which has evolved for millions of years engages in biologically illogical behaviour, then we have to ask why the behaviour has persisted over time.

    If the above conditions are satisfied, then the only sensible explanation for the behaviour is that it is feeling-driven. The reason why the behaviour has not been eradicated would then be that feelings confer benefits on their possessors.

    Condition (4) is elaborated in an Appendix, where I explain why I have chosen the long-term number of descendants as a measure of biological rationality of an organism's choice, and set forth the mathematical conditions for a biologically irrational choice.


    Let us consider a class of risky behaviour B, which is found in some species of organism, and which may adversely affect the organism's reproductive prospects. Suppose also that B satisfies the requirements for some form of intentional agency (see Conclusions DF. 1 to DF. 4). Let r be the probability of an adverse effect on the organism's reproductive success. Let E(not-B) be the organism's expected number of progeny if it abstains from behaviour B, E(fail) be the organism's expected number of progeny if it engages in behaviour B and its reproductive prospects are harmed, and E(success) be the organism's expected number of progeny if it engages in behaviour B and its reproductive prospects are enhanced. Then I propose that B manifests phenomenal consciousness if:

    E(not-B) > r.E(fail) + (1 - r).E(success).

    We can be more precise if we stipulate the number of descendants after N generations. (Allows for considerations of fitness - not all progeny are equal from an evolutionary viewpoint.)

    Which animals are CONSCIOUS?


    Silby, B. 1998. E-paper. "On A Distinction Between Access and Phenomenal Consciousness." Web address:

    Robert Lurz, "Neither HOT nor COLD: An Alternative Account of Consciousness", Psyche, 2003, Volume 9, No. 1, at

    Fox, D. 2004. "Do fruit flies dream of electric bananas?" In New Scientist, vol. 181, issue 2434, 14 February 2004, page 32. Web address:

    Carruthers, P. 2000. Phenomenal Consciousness. Cambridge University Press.

    Block, N. 1997. "On a Confusion about a Function of Consciousness". In The Nature of Consciousness, edited by Block. N., Flanagan. O., and Guzeldere. G. MIT Press.

    Rosenthal, D. 1986. "Two concepts of consciousness." In Philosophical Studies 49: 329-359.

    Gallup, G. 1998. "Animal Self-Awareness: A Debate - Can Animals Empathize? - Yes." In Scientific American, 1998.

    Different kinds of consciousness

    The term "consciousness" has various usages: it can be ascribed to both animals and their mental states. We may impute consciousness to a creature (e.g. a bird), or we might argue about whether its mental states (e.g. its perceptions of a worm) are conscious. Accordingly, philosophers, following Rosenthal (1986), draw a distinction between creature consciousness and state consciousness.

    Creature consciousness comes in two varieties: intransitive and transitive. We can say that a bird is conscious simpliciter if it is awake and not asleep or comatose, and we can also say that it is conscious of something - e.g. a wriggling worm that looks good to eat. Furthermore, an animal with transitive creature consciousness might be conscious of an object outside its mind (e.g. a worm) or of an experience inside its mind (e.g. an unpleasant sensation). In the former case, the creature is said to be outwardly conscious of the object; in the latter case, it is said to be inwardly conscious of its experience.

    State consciousness, by contrast, can only be intransitive. As Dretske (1995) puts it:

    States ... aren't conscious of anything. They are just conscious (or unconscious) full stop.

    Ned Block (1997) criticises the concept of state consciousness as a mongrel concept, and has proposed a distinction between two different types of state consciousness: access consciousness and phenomenal consciousness. A mental state is access-conscious if it is poised to be used for the direct (i.e. ready-to-go) rational control of thought and action. Phenomenally conscious states are states with a subjective feel or phenomenology, which, according to Block, we cannot define but we can immediately recognise in ourselves. However, some philosophers have queried the explanatory relevance of this distinction (e.g. Silby, 1998).

    Finally, certain scientists distinguish between primary consciousness (also called "core consciousness" or "feeling consciousness") - a moment-to-moment awareness of sensory experiences and some internal states - and higher-order consciousness, also known as "extended consciousness" or "self-awareness" (Rose, 2002).

    The debate about animal consciousness is not a debate about creature consciousness: wakeful and dormant states are known to occur in various phyla of animals, and all cellular organisms (including bacteria) are capable of responding to events occurring in their surroundings. Rather, what is at stake is state consciousness, and in particular, phenomenal consciousness, which roughly corresponds to what Rose calls primary consciousness. In the discussion that follows, I shall use the term "conscious" to mean "phenomenally conscious".

    The contemporary philosophical debate is split into several camps, with conflicting intuitions regarding the following four inconsistent propositions (Lurz, 2003):

    1. Conscious mental states are mental states of which one is conscious.
    2. To be conscious of one's mental states is to be conscious that one has them.
    3. Animals have conscious mental states.
    4. Animals are not conscious that they have mental states.

    Proponents of so-called higher order representational (HOR) theories of consciousness accept propositions 1 and 2. HOR theorists argue that a mental state (such as a perception) only becomes conscious by virtue of its being an object of creature consciousness. Perceptions, on this account, are not intrinsically conscious; they require higher-order states to make them so. These higher-order states are variously conceived as thoughts (by HOT theorists) or as inner perceptions (by HOP theorists) (Wright, 2003).

    Exclusive HOR theorists like Carruthers also accept 4 but reject 3 - that is, they allow that human infants and non-human animals have beliefs, desires and perceptions, but insist (Carruthers, 2000, p. 199) that we can explain their behaviour perfectly well without attributing conscious beliefs, desires and perceptions to them.

    Inclusive HOR theorists, such as Rosenthal, accept 3 but reject 4. Rosenthal (1986) argues that animals can have very crude thoughts about their mental states - e.g. the thought that one is having a particular sensation.

    Defenders of first-order representational (FOR)accounts of consciousness, such as Dretske, accept 2, 3 and 4 but reject 1. Thus Dretske argues that a mental state becomes conscious simply by being an act of creature consciousness. An animal need not be aware of its states for them to be conscious. On this account, consciousness has a very practical function: to alert an animal to salient objects in its environment - e.g. potential mates, predators or prey. On the other hand, such an account is vulnerable to Chalmers' (1996) zombie argument, as one could conceive of an unconscious zombie with the same discriminatory abilities as a conscious animal.

    Lurz (2003) rejects Dretske's position, on the grounds that it seems counter-intuitive to say that an animal could have a conscious experience of which it was not conscious. According to Lurz's (2003) same-order (SO) account, it is the assumption (shared by HOR and FOR theorists) that to be conscious of one's mental states is to be conscious that one has them, that needs to be queried. Lurz suggests that a creature's experiences are conscious if it is conscious of what (not that) its experiences represent. By "what its experiences represent" he means their intentional object.

    I believe that if we are to resolve the current argumentative deadlock about animal consciousness, we might do well to set aside both our philosophical thought experiments (such as Chalmers' zombie) and our grammatical intuitions about the proper usage of the word "conscious", and focus instead on the empirical cases that underlie the arguments. In particular, recent research on attention mechanisms (discussed in Wright, 2003) sheds valuable light on animal consciousness.

    Wright (2003) considers the much-discussed case of the distracted driver, who is supposedly able to navigate his car for miles despite being oblivious to his visual states. FOR theorists happily grant that the distracted driver has conscious visual states of which he is not aware; HOR theorists deny this. Wright faults both camps for being too gullible, citing three driving studies which show that driving requires a certain minimum amount of attention to the road. What really happens in "distracted driving" is that the driver pays attention to the road for some of the time, but the other matter that he is thinking about demands a much greater share of his cognitive resources, with the result that the information about the visual scene is quickly bumped from working memory and never encoded in long-term memory. Hence the driver's shock when he comes to the end of his journey.

    Additionally, studies of inattentional blindness (IB) and change blindness (CB) refute the claim made by FOR theorists that subjects can have visual experiences that they are not attending to. Wright (2003) cites research on IB, showing that when subjects are engaged in visual tasks demanding a high degree of attention, they fail to notice unexpected objects in their field of vision, even when they occupy the same point on their visual scene as the objects they are attending to. During CB, subjects fail to notice large-scale changes in a scene directly before their eyes, because their attention is diverted to other features of the scene. The upshot is that "there seems to be no conscious perception without attention" (Wright, 2003). Dretske's (1995) assertion that "You may not pay much attention to what you see, smell, or hear, but if you see, smell or hear it, you are conscious of it" is therefore empirically wrong.

    The relevance of the above research to animals should be obvious:

    Conclusion E.20 Attending to an object is an necessary condition for being conscious of it.

    Conclusion E.21 Only those animals that are capable of attending to objects in their surroundings can be described as having phenomenally conscious states.
    Wright does not define exactly what he means by "attention". Scientific models of attention often include aspects of selectivity, selecting one item in favor of another. Either the selected one is enhanced, or the other one is suppressed. Other models discuss the assignment of resources to items. Due to the resulting reduction of data to be processed, models of computer vision sometimes use mechanisms of visual attention. Another aspect of attention is that it allows to select relevant items and suppress distractors for task specification.

    Which criteria should we use to identify phenomenal consciousness in animals?

    Use Carruthers' article to distinguish different kinds of consciousness.

    Suggest the following theses:

    Talk about: representationalism (presupposed here).

    (1) What we call phenomenal consciousness consists simply in paying attention to our experiences (esp. our internal bodily representations). Paying attention to external objects is creature consciousness, but need not be phenomenal. Third person account will do. (Aside: paying attention to internal states is not the same as looking inside your body. Seeing your body from the inside is not the same as seeing inside your body from outside.) Paying attention to internal states requires a first-person account.

    (2) Intentional activity, manifest as controlled behaviour, indicates paying attention.

    (3) Any creature capable of controlling its behaviour in order to stabilise or prolong its internal states (in the absence of any external object that mirrors or tracks those states) is conscious.

    Alternate argument: (2) Are there purely neural criteria for attention? See report on fruit flies in New Scientist 14/2/2004.

    Is Drosophila at the flight simulator conscious? Recall the case where it was flying blind without any cues. Was it exercising control here? Maybe not (see earlier remarks about whether intentional agency was required in this case). However if it had to perform complicated manoeuvres we might say yes. Certainly recent research shows that flies are capable of paying attention to external objects (see report in - New Scientist 14/2/2004).

    Rats engaged in self-stimulation. See Panksepp 1998.

    See Wright, W. 2003. "Attention and Phenomenal Consciousness." Washington University PNP Medical School Seminar Program (December 2003). Web address: Criticises Tye's theory in favour of HOR theory, but one which does not require thought-like or perception-like higher-order states, just higher-order attention states.

    Criticism of Wright. What exactly is attention? Wright spoils his definition by including consciousness in the definition, of a term which was supposed to elucidate the nature of consciousness. "The attentional resources I have in mind - and which seem to be at work in the earlier cited studies of the cognitive demands of driving - operate at a conscious level and have to do with access to and control over information by the perceiving subject." Later he complicates his case by suggesting that there are also "subconscious attentional mechanisms".

    However, his notion of control over information by the subject seems a promising one.

    Elsewhere he writes: "Attention marks certain elements of the scene as salient and the rest are lost to consciousness. Those elements of the before view that have been indexed are loaded into visual short-term memory (VSTM), persist through the distractor event, and are available for comparison to the after view." This is more detailed, but what about the other senses?

    Also refer to New Scientist report (14/2/2004) about attention in fruit flies.

    See Allen, C. 2003. Animal Pain. In Nous 2 April 2003.

    DON'T argue here: behaviour X can occur without any subjective feelings, so it can't indicate the presence of an emotion or pain. This ASSUMES that emotions and pain are always SUBJECTIVE, which has yet to be proven. Most of what goes on when we feel is SUB-CONSCIOUS (LeDoux, 1998). Maybe in some cases, ALL of what goes on is sub-conscious. (Sub-conscious fears? dislikes? desires? Maybe even pains?)

    Pain - aversive behaviour not a sufficient criterion; learned avoidance behaviour not a sufficient criterion; analogous neurophysiology not a sufficient criterion; presence of nociceptors not a sufficient criterion; insensitivity arising from a distributed nervous system not sufficient to rule out pain; pain guarding a sufficient criterion?
    Useful quotes:

    "Noiception is the physiological response to painful stimuli and it does not involve the highest parts of the brain and pain proper" (Grandin and Deesing, 2003).

    "In humans, the prefrontal cortex must be intact in order to experience the emotional sensation associated with pain (Freeman and Watts, 1950). However, neurobiologists long believed that the PFC is a recent evolutionary acquisition and is unusually large in the human brain. Recent advances in the study of prefrontal cortex find no justification for these beliefs. Jerison (1997) conducted a formal analysis of similarities and differences between species and provides evidence that the PFC is an ancient part of the mammalian brain, is put together in all mammals pretty much the same way, and its functions are basically similar. The percentages of frontal cortex in relation to the rest of the brain are 29% in humans, 17% in chimps, 7% in the dog, and 3% in the cat (Broadman, 1912, Fuster, 1980). Although cats have less PFC compared to dogs, we would argue against any suggestion that cats suffer less from pain than dogs, or that rats suffer less than cats. It is likely that the cat has sufficient frontal cortex circuitry to have the minimum required amount to fully suffer" (Grandin and Deesing, 2003).

    "On the question of size, the PFC in humans is very large, but not disproportionately large. In other words, as a brain becomes larger and more complex it requires more circuits which can associate and merge inputs from many different parts of the brain. A small brain requires a less complex 'control room' than a bigger brain" (Grandin and Deesing, 2003).

    There has been some controversy in the scientific literature on the evolution of the prefrontal cortex. Some of the controversy may be caused by differences in how the prefrontal cortex is defined. Wood and Grafman (2003) contains an excellent map of prefrontal cortex and its connections to other parts of the brain. The ventromedial prefrontal cortex is old from an evolutionary standpoint. It has direct connections to the amygdala (emotion center). The dorsal lateral frontal cortex developed later and it integrates inputs from many parts of the brain and it makes it possible for an animal to engage in more abstract behaviors. It receives emotional information via the more primitive ventromedial prefrontal cortex (Wood and Grafman 2003). Some scientists consider the dorsal lateral prefontal cortex to be the 'true' prefontal cortex" (Grandin and Deesing, 2003).

    "Although some fundamental uncertainty exists when it comes to assessing subjective experiences such as pain and suffering in mammals, certain criteria can provide insight. For example, when an animal shows protective behavior towards an injured part, such as limping after an injury to a leg, going off feed because of abdominal injury, or actively seeking relief from pain by ingesting both opiate and non-opiate analgesics, such responses can indicate that something more complex than a simple reflex is taking place. All mammals pain guard after an injury... Poultry also engage in pain guarding after beak trimming and will peck less (Duncan, et al 1989, Gentle, et al 1991). Pain guarding occurs even when a limb is structurally sound and capable of bearing weight... Research on de-beaked chickens shows they pain guard after the procedure and will reduce food intake. De-beaked chickens are reluctant to use their beaks. Sometimes a neuroma forms on the end of the beak after it heals. Neuromas can cause pain in man (Gentle, et al 1990). Chickens with neuromas reduce the number of pecks at feeding (Gentle, et al 1990; Duncan et al 1989).... Grandin (1997) and Bateson (1991) stress the importance of separating fear stress from physical stress such exertion from running or overheating. Fear has a powerful ability to override pain in the chicken. The work by Gentle and Corr (1995) shows that a chicken that was pain guarding by holding its leg up will stop pain guarding when it is placed in a scary novel place... However, it is likely that birds may experience pain differently. Recent work by Gentle (1997) show that decerebrate chickens will still pain guard legs injected with a substance that causes pain. The results suggest that in chickens, pain from chronic arthritis is organized in the brainstem. However, if the chicken's beak is trimmed and the frontal area of the brain is removed, pain guarding and other pain related behaviors are absent. But, if the beak is trimmed six days after the frontal area of the brain is removed, the chicken continues to pain guard (Gentle, et al 1997). It appears that chickens are unable to process two emotions simultaneously. Chickens may suffer from chronic pain when they are undisturbed, but when disturbed or frightened, the pain ceases and the chicken can only attend to the fear (Gentle and Corr, 1995). Prelaying behavior and feeding motivation can completely suppress pain coping behaviors in arthritic chickens (Gentle and Corr, 1995; Wylie and Gentle, 1998). Turkeys with degenerative hip disorders reduce spontaneous activity and sexual activity (Duncan et al 1991). The authors conclude that the different systems in a bird's brain may be less integrated than in higher mammals. A bird may be more mono channel and operate only one system at a time. The bird would probably be suffering if the pain or fear channel is operating" (Grandin & Deesing, 2003).

    Do reptiles or amphibians suffer from pain? Research shows that the nervous system of amphibians responds to analgesic drugs. Amphibians will respond to a painful stimulus applied to the skin. Many different types of analgesic drugs will reduce the response (Stevens et al., 1994; Stevens et al., 2001). Is this true suffering from pain or is it just a reflex like touching a hot stove? Do reptiles and amphibians pain guard or seek analgesics? Both these areas need to be researched. The antedotes below may provide some insights for guiding future research. Discussions with reptile specialists indicate that reptiles may or may not pain guard. Friend (1998 personal communication) indicates that iguanas will walk on a severely damaged leg and make no attempt to reduce weight on the damaged limb. Iguanas are physically capable of lifting a leg to favor it, but they do not. Lizards react to noxious stimuli which may cause acute pain, but may have little reaction to injuries that would cause long term pain... A tortoise with a sore mouth will not eat and if it has a sore toe it will not walk. This is likely to be true pain guarding. Snakes with a damaged mouth may refuse to eat or lie on their backs to avoid pain. A tortoise with an abscess in its head will refuse to eat. Eating resumes shortly after the abscess is drained" (Grandin & Deesing, 2003).

    "Research by Lynne Sneddon at the Roslin Institute indicates that fish engage in true pain related behavior. Fish that had acetic acid injected into their lips engaged in more pain related behaviors such as rubbing their lips on the gravel and rocking compared to saline injected controls. There were no differences in swimming activity. Administering morphine reduced the pain related behaviors... Fish injected with acetic acid took significantly longer to start feeding compared to saline injected controls. These studies indicate that to insure a reasonable level of welfare providing pain relief should be considered for fish. This excellent research study separated the variables of pain from fear by having a saline injected control" (Grandin & Deesing, 2003).

    Steve Kaufman's review (Vegan Outreach, of Rose's article, "The neurobehavioral nature of fishes and the question of awareness and pain", in Fisheries Science, Vol. 10, 2002, pp. 1-38:

    Many of us would claim that fishes' behavior clearly demonstrates that they can suffer. Yet, James D. Rose argues that scientific evidence strongly favors the conclusion that fishes do not feel pain. If valid, this has implications for animal advocacy.

    Rose's argument rests heavily on the observation that human consciousness depends on the neocortex, which is absent in fishes. While humans, fishes, and other vertebrates share more primitive spinal cord and brainstem structures, neuroanatomic studies show that the fishes' cortex is far less developed than that of humans and does not include a neocortex. Decorticate humans, who have no input from the brainstem to the cortex due to trauma, stroke, or other damage, can still exhibit behavioral responses to stimulating nociceptors (nerve receptors that are stimulated by injury). Such humans will withdraw a stimulated limb, vocalize, exhibit a facial grimace, and release hormones and neurotransmitters associated with pain, even though there is no evidence that they are conscious. Even people under general anesthesia receive analgesics to block the hormonal responses that unconsciously accompany nociception.

    The spinal cord and brainstem, which are much older structures than the cortex in terms of evolution, mediate the primitive withdrawal responses to nociception in all vertebrates. Spinal cord and brainstem activity occurs unconsciously in humans and, presumably, other animals. When you step on a tack, high velocity neurons to and from the spinal cord mediate rapid withdrawal of the foot, and slower impulses go the brain where, about a second later, you experience a decidedly unpleasant sensation.

    Humans and fishes have evolved independently for 400 million years. In humans, consciousness, including conscious perception of pain, requires multiple inputs from frontal and parietal neocortex structures that fishes lack. Further, frontal cortical structures that are absent in fish mediate the unpleasant aspect of pain perception in humans. People lacking input from the frontal cortex (due to disease, trauma, surgery, etc.) relate that they can feel painful stimuli, but it doesn't bother them.

    Fishes do have a cortex, which is anatomically far less complicated and much smaller in proportion to the brainstem than that of humans. The cortex in fishes is involved in sensory reception (particularly smell) and modulates responses to nociception, but fishes' cortex is not essential for their normal responses to noxious stimuli.

    Rose addresses several possible objections to his thesis:

    Fishes react to injurious or threatening stimuli. The ability to response to such stimuli is also seen in unicellular organisms and multicellular organisms without brains, and we do not attribute consciousness to them. And, humans have unconscious brainstem- and spinal cord-mediated responses to such stimuli.

    Fishes have nerve receptors that respond to opiates, which deaden pain in humans. Like humans, fishes have opiate receptors in their brainstems. Their presence in human brainstems demonstrates that opiate receptors have functions besides modulating conscious awareness of pain. It is easy to envision their use in modulating behavior without requiring consciousness. For example, an animal with an injured leg benefits from nociceptive reflexes that encourage keeping weight off the leg, but the animal is best served by suppressing this reflex (with opiate-like neurotransmitters) when being chased by a predator.

    Noxious stimuli alter fish behavior. While learning would appear to be compelling evidence for consciousness, it appears that humans gain considerable learning without conscious awareness. Perhaps humans require consciousness in order to learn about and make subtle discriminative choices in much more complex situations than fishes experience. Just as a computer can be programmed to unconsciously learn from experience, fishes may exhibit analogous abilities. Fishes can exhibit simple associative learning, and humans seem to have this capacity at an unconscious level as well. Furthermore, fish behavior is largely preserved after their cortex has been removed, with the principle deficit being loss of responses to smell, because their sense of smell is mediated by their cortex.

    While inconclusive, it appears that fishes have analogous structures to the human limbic system, which is involved in emotions and in life-sustaining behaviors in mammals, including reproduction, aggression, defense, feeding, and drinking. In humans, consciousness of emotions requires the neocortex that, again, is absent in fishes. Furthermore, a component of the limbic system known as the cingulate gyrus appears to be essential for the human emotional response to pain. This structure has been identified only in mammals, and researchers have done extensive neuroanatomic research on fish.

    Rose argues that, if fishes were conscious, their experiences would differ so fundamentally from those of humans that any attempts at empathetic understanding would be futile. Neuroanatomy research, including studies on injured humans and invasive experiments on animals, has shown that specific functions require specific neuroanatomic structures. Lacking the human structures involved in consciousness and emotional perception, Rose asserts that it is unreasonable to believe that fishes experience noxious stimuli in analogous ways to humans. Similarly, we can't prove that plants are insensate, but we strongly doubt that they experience damaging events as we do, because they lack the anatomic components we regard as essential to conscious experience.

    I think that the burden of proof is on those who would harm fishes, particularly if the damage to fishes' well-being were unnecessary. It seems most reasonable to avoid harming fishes because there is no way to know for certain whether or not they can have some kind of unpleasant sensation. Just as we can't empathize with a bat's echolocation, we may be unable to empathize with what a fish feels when hooked, but the possibility that fish find the experience subjectively unpleasant is good reason to refrain.

    -Steve Kaufman

    Different senses of "conscious" (Allen, 2002) - awake (not asleep); able to respond to stimuli; phenomenally conscious in a "subjective" sense; self-conscious (not relevant in this chapter).
    Distinguish different senses of "conscious":

    "Although consciousness has multiple dimensions and diverse definitions, use ofthe term here refers to two principal manifestations of consciousness that exist inhumans (Damasio, 1999; Edelman and Tononi, 2000; Macphail, 1998): (1) "primary consciousness" (also known as "core consciousness" or "feeling consciousness") and (2) "higher-order consciousness" (also called "extended consciousness" or "self-awareness"). Primary consciousness refers to the moment-to-moment awareness ofsensory experiences and some internal states, such as emotions. Higher-order con-sciousness includes awareness of one's self as an entity that exists separately fromother entities; it has an autobiographical dimension, including a memory of past life events; an awareness of facts, such as one's language vocabulary; and a capacity for planning and anticipation of the future. Most discussions about the possible existence of conscious awareness in non-human mammals have been concerned with primary consciousness, although strongly divided opinions and debate exist regarding thepresence of self-awareness in great apes (Macphail, 1998). The evidence that the neocortex is critical for conscious awareness applies to both types of consciousness. Evidence showing that neocortex is the foundation for consciousness also has led to an equally important conclusion: that we are unaware of the perpetual neural activity that is confined to subcortical regions of the central nervous system, including cerebral regions beneath the neocortex as well as the brainstem and spinal cord (Dolan, 2000; Guzeldere et al., 2000; Jouvet, 1969; Kihlstrom et al., 1999; Treede et al., 1999). Although consciousness has been notoriously difficult to define, it is quite possible to identify its presence or absence by objective indicators. This is particularly true for the indicators of consciousness assessed in clinical neurology, a point of special importance because clinical neurology has been a major source of information concerning the neural bases of consciousness. From the clinical perspective, primary consciousness is defined by: (1) sustained awareness of the environment in a way that is appropriate and meaningful, (2) ability to immediately follow commands to perform novel actions, and (3) exhibiting verbal or nonverbal communication indicating awareness of the ongoing interaction (Collins, 1997; Young et al., 1998). Thus, reflexive or other stereotyped responses to sensory stimuli are excluded by this definition. Primary consciousness appears to depend greatly on the functional integrity of several cortical regions of the cerebral hemispheres especially the 'association areas' of the frontal, temporal, and parietal lobes (Laureys et al., 1999, 2000a-c)... Wakefulness is not evidence of consciousness because it can exist in situationswhere consciousness is absent (Laureys et al., 2000a-c)" (Rose, 2002, p. 6).

    The reasons why neocortex is critical for consciousness have not been resolved fully, but the matter is under active investigation. It is becoming clear that the existence of consciousness requires widely distributed brain activity that is simulta-neously diverse, temporally coordinated, and of high informational complexity (Edelman and Tononi, 1999; Iacoboni, 2000; Koch and Crick, 1999; 2000; Libet,1999). Human neocortex satisfies these functional criteria because of its unique structural features: (1) exceptionally high interconnectivity within the neocortex and between the cortex and thalamus and (2) enough mass and local functional diversification to permit regionally specialized, differentiated activity patterns (Edelman and Tononi, 1999). These structural and functional features are not present in subcortical regions of the brain, which is probably the main reason that activity confined to subcortical brain systems can't support consciousness (Rose, 2002, p. 7).

    Our fundamental behavioral reactions to noxious stimuli, including vocalization, facial grimacing, and withdrawal, are mediated by subcortical brain and spinalsystems (Jouvet, 1969; Kandel et al., 2000; Laureys et al., 1999, 2000a,b; Young et al.,1998). Activation of these responses by noxious stimuli can occur without consciousness in people with extensive cortical damage (Figure 3) and in humans born without cerebral hemispheres (Kolb and Whishaw, 1996; Steiner, 1987). Thus, the behavioral displays related to noxious stimuli or emotion in humans, as in other animals, are stereotyped, automatic behavioral programs controlled by lower levels of the central nervous system, and these responses can be evoked without any corresponding awareness of noxious stimuli. Limb withdrawal and leg locomotor responses, of course, are produced directly at the spinal cord level (Rose, 2002, p. 17).

    Comments on Rose's definition: none of the criteria Rose lists for primary consciousness are necessarily confined to mammals. Other vertebrates, and even some invertebrates (e.g. bees and octopuses), are surely "aware of their environment". Many animals can follow commands to perform novel actions (birds? fish?). Nonverbal communication exhibiting awareness of the ongoing interaction is surely not excluded for non-mammals. Where is the evidence that these abilities require a neocortex?

    The indispensability of subjectivity (or: Which intentional stance is appropriate for discussing emotions and pain?)

    Is our knowledge of animals' mental states an inference to the best explanation? No, argues Allen (2003).

    Look at arguments AGAINST animal subjectivity (see Allen's article):
    (1) Dissimilarity argument: Animals lack key brain structures (especially a neocortex - see Rose's article on fish). Animals lack language (Descartes, Frey?).
    (2) Similarity argument: Animals behave like human beings on autopilot, or blindsight subjects.
    (3) Phenomenal consciousness requires ability to think about one's thoughts & distinguish appearance from reality (see Allen 2002, 2003).
    (4) Experimental arguments. It is experimentally possible to make human and animal subjects want things they are not conscious of (subliminal processing), and even to want things that they do not expect to make them feel better (irrational pursuit - addiction) (Berridge, 2003a, 2003b). Conclusion: wanting need not be conscious.

    Are humans and animals emotionally symbiotic? (Mary Midgley)

    The relevance of emotions to animals' interests (Having an interest vs. taking an interest. Addiction. Binged-out beetles.)


    Mary Midgley, M. "The Increasingly Official Mary Midgley Web page." Web address:

    PBS. "Inside The Animal Mind." Web address:

    University of Derby. "Animal Cognition Links." Web address:

    Berridge, K. 2003. "Irrational Pursuit: Hyper-Incentives from a Visceral Brain." In Brocas I. and Carrillo J. (eds.) The Psychology of Economic Decisions. Volume 1: rationality and well-being, pp. 17-40. Oxford University Press. Web address: Winter01/Berridge_Irrational_Pursuits.pdf.

    Berridge, K. 2003. "Pleasure, Unfelt Affect and Irrational Desire." In Manstead A. & Frijda N. (eds.), Feelings and Emotions: The Amsterdam Symposium. Cambridge University Press. Web address:

    Winkielman, P. & Berridge, K. 2003. "Irrational Wanting and Subrational Liking: How Rudimentary Motivational and Affective Processes Shape Preferences and Choices." In Political Psychology, 24 (4), pp. 657-680. Web address of abstract:

    Midgley, M. 1993. '"The Four-Leggeds, the Two-Leggeds, and the Wingeds": An Overview of Society and Animals.' In PSYETA, Vol. 1, No. 1. Web address:

    LeDoux, J. "The Emotional Brain, Fear and the Amygdala." In Cellular and Molecular Neurobiology, Vol. 23, Nos. 4/5, October 2003. Web address: Good Web site: Do fish feel pain? Plus: Fish behaviour relevant to pain Article on scientific meaning of attention Scientific models of attention often include aspects of selectivity, selecting one item in favor of another. Either the selected one is enhanced, or the other one is suppressed. Other models discuss the assignment of resources to items. Due to the resulting reduction of data to be processed, models of computer vision sometimes use mechanisms of visual attention. Another aspect of attention is that it allows to select relevant items and suppress distractors for task specification.

    Back to Main Page Previous chapter Next chapter References