Published in Behavioral and Brain Sciences
Volume 25, Number 4 (August 2004)

Facial expression of pain: An evolutionary account

Acknowledgements

I am grateful for encouragement, discussion and comments on drafts of this paper to Gerhard Andersson, Kenneth Craig, Geert Crombez, Helena Cronin, Chris Eccleston, Paul Ekman, Nicholas Humphrey, Maria Jannert, Judith Kappesser, Stephen Morley, Kenneth Prkachin, Tony Roth, Aubrey Sheiham, Colin Tudge, and Patrick Wall; to members of the Darwin@LSE Work in Progress group at the Centre for the Philosophy of Natural and Social Sciences, London School of Economics, London UK; to all the BBS referees; also to James Kirkwood of UFAW for sources on non-human animals in pain.

Abstract

This paper proposes that human expression of pain in the presence or absence of caregivers, and the detection of pain by observers, arise from evolved propensities. The function of pain is to demand attention and prioritise escape, recovery and healing; where others can help achieve these goals, effective communication of pain is required. Evidence is reviewed of a distinct and specific facial expression of pain from infancy to old age, consistent across stimuli, and recognizable as pain by observers. Voluntary control over amplitude is incomplete, and observers better detect pain which the individual attempts to suppress than to amplify or to simulate it. In many clinical and experimental settings, facial expression of pain is incorporated with verbal and nonverbal-vocal activity, posture and movement in an overall category of pain behaviour. This is assumed by clinicians to be under operant control of social contingencies such as sympathy, caregiving, and practical help; thus strong facial expression is presumed to constitute an attempt to manipulate these contingencies by amplification of the normal expression. Operant formulations support skepticism about the presence or extent of pain, judgements of malingering, and sometimes the withholding of caregiving and help. However, to the extent that pain expression is influenced by environmental contingencies, "amplification" could equally plausibly constitute release of suppression according to evolved contingent propensities which guide behaviour. Pain has been largely neglected in the evolutionary literature and that on pain expression, but an evolutionary account can generate improved assessment of pain and reactions to it.

Pain motivates the individual to promote safety and recovery; the function of facial expression of pain is examined in relation to evolved propensities for specific behaviours. A distinct and specific facial expression of pain is recognized as pain by observers, even under attempts to suppress it. However, pain is largely disregarded in evolutionary literature on facial expression and help-giving, and facial expression of pain neglected in clinical pain settings, where operant formulations of global expression and social contingencies dominate. An evolutionary account of pain facial expression can generate improved assessment of pain and reactions to it.

Keywords: pain, facial expression, adaptation, evolutionary psychology

The insights provided by the application of evolutionary psychology to established fields of evidence has clinical relevance in areas such as fears, anxiety and depression (Dimberg & Ohman 1996; Gilbert 1992; Marks & Nesse 1994). Certain evolved behaviours, such as the attachment behaviour of human and other primate infants, are accepted as orthodoxy. But broader notions of an evolved ‘human nature’, that is, of specialised propensities ¹ which generate functional behaviours effective in the social and material world, are still viewed with skepticism or indifference by many in the psychological and medical communities (Andersson 1993; Schaffner 1998). Among the more important reasons for this may be concern about genetic determinism or, at least, a distaste for invoking genetic factors (Daly & Wilson 1988; Crawford 1989; Plomin 1989; Tooby & Cosmides 1992), and suspicion of "Just-So Stories", post hoc explanations from available evidence (Gould & Lewontin 1979). Far from endorsing the oversimple gene explanations which are sporadically revisited in psychiatry (Plomin 1989), evolutionary psychology aims to broaden our understanding of behaviour, in all its flexibility and contingency in relation to the immediate environment and the individual’s learning history, by addressing the cognitive adaptations which are part of our heritage (Cronin 1991).

Pain is the "final mediator" of a wide range of selection pressures (Walters, 1994): sub-lethal injuries which may still threaten survival or reproduction are incurred during predation, intraspecific combat, and competition with conspecifics. By virtue of its aversiveness, pain serves to promote the organism’s health and integrity, to the extent that congenital absence of pain on injury significantly shortens human life (Damasio 1999; Wall 1999): "Suffering offers us the best protection for survival" (Damasio 1994, p.264). The adaptive value of experiencing pain at first seems self-evident: distinguishing harmful from harmless situations, prompting avoidance of harm and its associated cues, giving a high priority to escape from danger, and promoting healing by inhibiting other activities which might cause further tissue damage (Bateson 1991). Research in animals has focussed mainly on immediate escape from pain (e.g. Rachlin 1985), but according to Wall (1979; 1999), this is less important than making recovery the overriding priority after escape. The presence and intensity of pain are often poorly related to the degree of tissue damage, making it too late for prevention of injury if not for future avoidance; neither escape nor avoidance would require that pain continued well into the recovery phase, demanding attention and not habituating to any appreciable degree in humans (Eccleston & Crombez 1999). The affective dimension of pain appears to share mechanisms with vigilance to threat (Chapman 1995; Crombez et al. 1998), and threat itself facilitates attention to pain (Eccleston & Crombez 1999).

The gate control theory (Melzack & Wall 1965) brought about a paradigm shift in the study and understanding of pain. It proposed that the pain signal following tissue damage is modulated at each synapse, thus throughout its transmission, by the balance of signals from the periphery and from descending pathways originating in multiple sites in the brain. This allowed for the influence on the signals and their transmission of memories and prior learning; beliefs, fears and expectations; and emotional state. However, these affective-motivational aspects have been sidelined in subsequent research which has established much more about the sensory-discriminative dimension of pain, aspects such as its quality, location and intensity which are largely determined by peripheral input (Chapman & Nakamura, 1999; Craig, 1999). While there is debate on the extent of anatomical separation of sensory-discriminative and affective-motivational processing of pain in the brain, there is consensus on the importance of recognising the separate processes (Price 1999; Wall 1999). Clinical and scientific focus, however, remains on pain sensation and sensory discriminative processing, for a number of reasons. Almost all experimental work is performed on animals, with most attention to quantification of nociceptive stimuli and their local effects; some attention to a restricted range of behaviours (largely escape and avoidance); and none to emotion and cognition. Experimental work is largely restricted to acute pain and to peripheral and spinal mechanisms; while brain imaging techniques offer rich data, its interpretation lacks adequate models (Wall 1999). "The careful sensory neurophysiologist who strays from the spinothalamic pathway quickly becomes lost in a huge and complex maze of reciprocal connections." (Chapman & Nakamura 1999, p.114).

While the neurophysiological model performs far better than its predecessors in building an understanding of pain and of methods of analgesia, it cast little light on the evolutionary function of pain and related behaviour. Pain undoubtedly motivates to action (Hinde 1985; Frijda 1994; Damasio 1994; Wall 1999). serving as a "lever" for decision making and for action based on drives and instincts (Damasio 1999). Behaviour following injury shows remarkable consistency across species (Walters 1994), modified by contextual variables such as the nature of the threat, the severity or imminence of injury, its location, and the costs of active defence. On the basis of accumulating evidence about the activity of areas of the brain concerned with motor function, Wall (1999) proposes analysis of pain by synthesis: that sensory inputs are analysed, classified and identified by premotor systems in terms of motor actions relevant to the input, with the priorities of first escaping the stimulus, then limiting further damage and prioritising recovery, and then seeking safety and relief. However, it is harder to adduce evidence for this from laboratory studies in which possible behaviours are constrained, often conditioned rather than unconditioned, and the widespread use of electric shock as the noxious (but not tissue-damaging) stimulus in research with laboratory animals raises questions of generalisability to injury-related pain (Walters 1994; Panskepp et al. 1997). Outside the laboratory, there is a dearth of observations of the behaviour of wounded mammals (Walters 1994; Mench & Mason 1997; Fleckness & Molony 1997), and what observations exist rely on the interpretation of behaviour or changes in behaviour whose function is not fully understood (Mench & Mason 1997). Assessment methods for pain in domesticated and farm animals are unstandardised (Fleckness & Molony 1997). The extent of pain in animals soon after injury, in the escape or active defence phase, is uncertain compared to its presence later on when the animal is resting and protecting the injured area (Wall 1979, 1999). This has led to models, such as those of Bolles and Fanselow (1980) which locate the warning function in the emotional experience of threat and fear in the early postinjury phase.

The emotion (affect) dimension of pain is therefore largely absent from much pain research, but since pain does not fit the paradigm of emotion (Izard 1991; Ekman 1992; Frijda 1994) it falls outside investigation of emotional expression. By contrast, the definition of pain by the International Association for the Study of Pain as "an unpleasant sensory and emotional experience associated with actual or potential tissue damage, or described in terms of such damage" (IASP 1979) provides a central role for emotion. Outside the pain field it is rare to find pain described other than as an aversive sensation associated with avoidance and escape, even in evolutionary writing on adaptations: for instance, Nesse and Williams (1994) describe pain phenomena in humans entirely in terms of design compromises for defence. Imaging of pain processing in humans and clinical lesion studies indicate distinct locations for encoding unpleasantness aspects (in the anterior cingulate) compared to sensory aspects (in the somatosensory cortex) (Rainville et al. 1997; Damasio 1994), and for learned anticipation of pain compared to actual pain (in different parts of the anterior cingulate cortex, the insular cortex and the cerebellum: Ploghaus et al. 1999). It is noted that all these areas where pain is processed are close to areas involved in motor responses (Ploghaus et al. 1999; Rainville et al. 1997), recalling Wall’s model of analysis of pain by synthesis with possible motor responses.

In humans, emotional aspects have largely been investigated by self-report, relying on consciously represented experiences (such as fear) or consciously initiated action (such as coping). The rich literature on cognitive content and processes, including emotion, associated with pain, consists of a bewildering array of associated concepts (such as sense of control, beliefs about illness, coping attempts), few of which bear any reliable relationship either with identifiable cognitive processes or with specific behaviours. By contrast, experimental work on attention and pain (Eccleston & Crombez 1999), which includes methodologies which sample nonconscious processes, complements the motivational model: pain grabs attention, interrupts associated behaviour and urges action towards mitigating it; the more intense and threatening the pain, the more disruptive of attention to anything else. Considerations of adaptive mechanisms and function of pain are, however, rare in the pain field, with some notable exceptions such as the work of Craig and colleagues on facial expression, and of Crombez and Eccleston and colleagues (Crombez et al. 1996; 1998; Eccleston et al. 1997) and Chapman (1995) on cognitive mechanisms and pain. In a broader cognitive context, theoretical models of consciousness which incorporate somatic experience (Melzack 1999; Damasio 1999; Chapman & Nakamura 1999) provide the basis for integration of supraspinal processing of pain, but still tend to underrepresent the interaction of thoughts, beliefs and emotional state with pain (Keefe et al. 1996). An ideal model of pain would extend the ecological validity of the current detailed understanding of neurophysiology and integrate it with behavioural, cognitive and emotional dimensions. Models developed in human research are diverse, with minimal theoretical integration between them or with sensory experimental work (although cognitive and behavioural models are often combined in treatment), and very little reference to function in evolutionary terms. Some of those who have attempted integrated theories note that they struggle against the bias in language (Sullivan 1995) and pervasive dualism (Chapman 1995; Morris 1998).

A major psychological model which has dominated pain psychology in clinical settings for several decades is the operant model proposed by Fordyce et al. (1968) described more fully below. While it is a theory of pain behaviour rather than of pain, and explicitly builds on the gate control model of pain, it is important because, consistent with Skinnerian behaviourist principles, it asserted that pain was unknowable, and that the proper focus of theory and empirical work was observable and quantifiable behaviours associated with pain (henceforth "pain behaviours"). Over repeated episodes of pain, or with persistent pain, unconditioned verbal, paraverbal, and motor responses come under the control of external contingencies. These behaviours contribute to or constitute aspects of disability. Support for the concept was largely drawn from its behavioural origins, specifically from positive associations of frequencies of pain behaviours with availability of social reinforcement, and from accounts of treatment (Fordyce et al. 1968; Sternbach 1968; Fordyce et al. 1981). However, there are relatively few tests of the covariation of pain behaviours with social and instrumental reinforcement (Keefe & Dunsmore 1992; Turk & Matyas 1992), and clinical practice rarely establishes by observation the contingent relationships of behaviours targeted in treatment. Instead, the availability of potential social reinforcement is taken to support the likelihood that pain behaviours observed in clinical settings are operantly controlled. It seems to be widely overlooked that many pain behaviours such as vocalisation of distress, guarding, limping, and avoiding activity, are shared with many other species in acute and chronic pain where there is no suggestion that they are actually under operant control, although in some circumstances certain behaviours can be so manipulated.

Behaviourism aims to deduce the working of the mind from behaviour, but in the field of pain, the external contingencies whose balance and likelihood determine the frequency of behaviour become, by oversimplification and in the absence of a theory of motivation, the reasons for that behaviour. It has, however, contributed to the development of an interactional and socially contextual formulation of pain behaviour. Cognitive theory in pain represents behaviour as driven by beliefs and their emotional tone (such as the fear-laden certainty that the processes causing pain threaten physical integrity); expressive behaviour is conceptualised as an epiphenomenon of internal experience, or more often left to behavioural explanations. By contrast with behaviourism, evolutionary psychology aims to understand behaviour through a theory of mind, incorporating cognitive, emotional, and motivational dimensions. Its focus is "behavioral control mechanisms [which] are organized to process adaptively significant information and respond thereto" (Daly & Wilson 1991, p. 163). But the two perspectives do not necessarily provide contrasting predictions of observed behaviour. Environmental influences, from cultural norms to interactions with particular individuals, act on behavioural tendencies which are, already, the outcome of contextual propensities to action, with reference to the balance of costs and benefits (for the future of the individual’s genes, not the individual) of possible behaviours in a particular situation. Where possible, descriptions generated by behavioural and evolutionary theories are contrasted, but at this stage of development of the theory, no opposing hypotheses are put up for test. The challenge is not to all application of behavioural theory to the field of pain behaviour, but to its weakness in explaining such behaviour particularly in clinical setting, and to its shortcomings in relation to the interactive nature of pain expression, as in the systematic biases evident in certain observer judgements of pain.

This paper first addresses in greater detail the unsatisfactory aspects of the global pain behaviour construct to describe clinically relevant behaviours, particularly those which affect clinical decisions and pain sufferers’ wellbeing. It then sketches the tenets of evolutionary psychology applied to propensities for behaviour and the function of that behaviour. Germane areas of work include that of facial expression of emotion, as proposed by Prkachin (1997), and particularly the work of Ekman and colleagues and of Fridlund. These are then applied to the evidence on facial expression of pain, studied systematically and with close reference both to theory and to clinical dilemmas by Craig and LeResche and their colleagues. Although they make reference, particularly in work on infants’ expression, in consistency of expression (Prkachin 1992a), and in reviews of the field (LeResche & Dworkin 1984; Craig et al. 1992), to a communicative facial expression as the product of evolution and as a cue for help-giving (Prkachin et al. 1983; Anand et al. 1999), their research questions have not primarily been informed by evolutionary psychology. Their studies constitute an invaluable canon in the field largely overlooked by those studying facial expression of emotion, and this paper relies substantially on their findings, reinterpreted in a broader context. The major issues are the extent to which facial pain expression, wittingly or unwittingly, constitutes a reliable communication of pain; the factors influencing suppression and amplification; the issue of simulation of pain expression; and detection and the biases affecting observer judgement. The paper concludes with areas of particular difficulty for the approach such as application in chronic pain, ecological validity of methodology, and some clinical issues which urgently require addressing.

However, this paper comes with a warning. Of major concern to the author and to colleagues in the pain field is the risk that description of an evolved pain expression in acute pain might be mistaken as offering a tool for distinguishing "real" from "faked" or "imagined" pain across all circumstances. The patient as an unreliable reporter of verifiable internal events is blamed when the two do not correspond, often resulting in treatment failure (Armstrong 1984). Psychoanalytic, personality, psychopathological and humanistic schools of psychology provide models which share with psychiatry certain dualistic assumptions about deviant expression of unmet needs through the unverifiable complaint of pain. Similar misconstructions, to the considerable detriment of patients, of behavioural signs of distress as pathognomic of malingering arose following the publication of work describing physical manoeuvres which indicated patient distress rather than pain intensity (Waddell et al. 1980). This paper emphatically cannot provide a means of identifying intentional simulation of pain.

Variants of the operant model are barely represented in the pain field, which has developed with considerable consistency from the original propositions by Fordyce (1976). Pain behaviours are by definition observable, and many of the contingencies for pain behaviours consist of social reinforcement such as sympathetic attention or help, but can also be the socially sanctioned avoidance of unpleasant events such as burdensome duties (Fordyce 1976). This model identified as targets for treatment both the pain behaviours and their governing contingencies. For instance, the patient whose pain elicited the sympathy of family members only when he drew attention to it found that attention withdrawn and given instead when he communicated other than about his pain. Several experimental and clinical studies provide evidence of change in the frequency of pain report (Block et al. 1980; Linton & Gotestam 1985; White & Sanders 1986; Flor et al. 1995) and in observed activity (Ritchie 1976; Flor et al. 1987) in relation to actual reinforcement or the presence compared to the absence of a source of social reinforcement. While these studies may explain substantial variance in the maintenance of one or more pain behaviours, and possibly the production of those behaviours in a social setting, we can deduce little about etiology or function of these behaviours. By contrast, two studies found a substantial proportion of variance in pain behaviour during medical examination to be attributable to medical findings such as number of previous surgeries or limitation of joint movement (Keefe et al. 1984; Novy et al. 1996), and several attempted replications or part-replications of the study by Block et al. (1980) have produced results inconsistent with operant predictions (Lousberg et al. 1992; Paulsen & Altmaier 1995; Williamson et al. 1997).

More broadly, by design there was no place for emotion within the model nor for cognitive content or process in generating or moderating behaviour, although learning history was identified as an influence on current behaviour. With the growth of cognitive psychology, within and without the pain field, the associations between cognitive variables such as beliefs and expectations, emotions such as fear and distress, and behaviours including pain behaviours, are under investigation. The importance of escape and avoidance when threatened with increasing pain are now revisited in a cognitive-behavioural formulation (Vlaeyen & Linton 2000). Of particular importance are vigilance to threatened injury or reinjury, sustained through the recovery period (Walters 1994), and fear-laden beliefs about threat and risk which bias interpretation of ambiguous stimuli (such as increased pain or associated symptoms) and maintain vigilant attention to pain-related cues (Craig 1999; Vlaeyen and Linton 2000).

The concept of pain behaviour and its use have been subject to criticism (see, for instance, Melzack 1985; Wall 1985; Keefe & Dunsmore 1992; Turk & Matyas 1992), including the neglect of possible adaptive properties of various pain behaviours (Turk and Matyas, 1992) but this has not had the impact it merits on the theory and investigation of pain behaviour. Despite the diversity of pain behaviours (verbal complaint; vocal, facial and postural behaviour; and particular activities or their avoidance) suggesting differences in function (Prkachin 1986, 1997), treatment and correlational studies generally represented pain behaviour as a unidimensional construct. Facial expression is the exception to this tendency, and excellent and detailed studies of facial expression of pain allow re-examination of the data using an evolutionary perspective. The relationship of facial expression to other behaviours remains unexplored in pain. Other behaviours, while they may augment or modify the message in the face (Argyle 1988), may have other functions: cancer patients who viewed their videotaped behaviour in the presence of a spouse (Wilkie & Keefe 1991; Wilkie et al. 1992) mainly explained it as attempts to reduce pain (e.g. by rubbing or support), or to control the display of pain to attain superordinate goals such as remaining active or not embarrassing others.

For observers, pain as a symptom presents particular difficulties, the more acute when, like clinicians, the observers must make a judgement and decision based on their observation, since the problem presented may consist entirely of self-report. There may be no consistent or proportional evidence for a pain problem from other sources such as clinical signs or investigations, since the changes are in neurophysiological function rather than anatomical structure, as described above. Many injury-related pains outlast healing; others are not attributable to injury, as in much headache, visceral pain, and neuropathic pain. One of the strengths of research on pain behaviour is that it addresses the difficult issue of skepticism of observers about the credibility of pain, almost completely neglected by other psychological perspectives. Observers generally trust nonverbal expression more than self-report (Craig et al. 1992; Poole & Craig 1992). All behaviour is influenced by social context: the important distinctions for the complaint of pain are between subjective pain experience and its observable expression, and between influences on behaviour arising from learned contingencies of response but not represented in conscious experience and the conscious and deliberate intention to manipulate those responses by particular behaviours. For instance, an influential paper by Waddell and colleagues (1980) which identified five signs on physical examination any three of which indicated distress worthy of psychological assessment, described one of the signs as "overreaction during examination which "may take the form of disproportionate verbalization, facial expression, muscle tension" (p. 119). Unless the physical examination is always conducted below the threshold of pain, which rather vitiates its purpose, some facial expression of pain is likely to be observed. It is interesting that of the five signs, this was the most influential in surgeons' judgements, even though it was the most subjective; and that contrary to the authors' expectations, these signs failed to distinguish patients with medicolegal or compensation cases (that is, with the prospects of gain from the judgement that they had severe pain) from those without.

While operant formulations did not generally equate pain experience with pain behaviour (Fordyce et al. 1985), they directed attention towards presumed social and material benefits of being observed to be in pain (such as family and medical attention, compensation and disability payments, and suspension of onerous duties) and away from the costs and losses of being in pain, of more interest to those who studied subjectivity and emotion. The explanation of patients’ behaviour – the complaint of pain and unwillingness to resume normal activities – by reference to assumed benefits is adopted widely in clinical medicine, as an apparently scientific way to express skepticism, and as a way to explain the lack of association between the patient’s complaint and the lack of confirming medical evidence which is common in chronic pain.

The fact that facial expression conveys information does not necessarily imply either conscious control before or during the expression, or intentional signaling. Unfortunately, definitions of pain behaviour, including facial expression, tend to use ambiguous terminology which may appear to describe deliberate communication: "The observable phenomena of pain consist essentially of some kind of signal sent by the patient that he is in pain" (Fordyce et al. 1968, p. 179: italics added); the "observable means of communicating pain and suffering" (Turk et al. 1985); the expressive individual may or may not be aware of this communication. In fact, the only time we can be certain that expression is intentional is when it is posed, with or without the associated experience, under instruction.

As described in the first section of this paper, operant mechanisms act on unconditioned pain behaviours. But where the implicit assumption is of a tabula rasa at birth, operant mechanisms become the exclusive explanation for pain, invoked with minimal or no evidence. Prkachin et al. (2001) comment critically on the wealth of speculation about family members' potential influence on the development of pain compared to the dearth of hypotheses or studies of how family members acually process pain information. The lack of research, even in animals, on unconditioned pain responses, contributes to this bias in the field. It remains to be seen whether definitive tests can be designed where the processes can be observed working separately or together. None of the studies reviewed in this paper were undertaken with the intention of distinguishing between contrasting predictions. Nor do any do so coincidentally, other than the experimental finding that subjects show more facial expression of pain when alone than when they believe themselves observed (Kleck et al. 1976) or are observed by neutral strangers (Badali 2000): in the latter study, self report showed the same bias, but tolerance to pain was the same across situations. At times, predictions by different models coincide, for instance, in anticipating that the person in pain will show more vivid expression of pain, whether through the release of suppression of pain expression in the presence of someone identified as a help-giver (evolutionary explanation), or on the basis of previous reinforcement by that person or class of persons of the expression (operant explanation).

Nevertheless, this paper reviews evidence collected in the investigation of facial expression of pain in the light of evolutionary theory. To varying extents in the different areas, evolutionary theory provides a superior explanation, or provides one where operant theory is silent. The theory of evolved pain expression predicts a distinct and specific expression of pain, across the lifespan, across pain stimuli, and across cultures. By contrast, operant theory would anticipate pain expression shaped by contingencies which might be specific to cultures and life stages. While evolved pain expression can only be proposed in the context of coevolution of its accurate detection by observers and its interpretation in terms of approximate severity, there is no literature on the operant processes by which accurate detection would be learned, and, again, it could be culture-specific. In evolutionary terms, the ability to suppress pain expression would potentially be of survival advantage (as when hiding pain from antagonists); but for the observer, whether ally or antagonist, the ability to detect pain expression despite attempts to suppress it would potentially have survival advantage. The position is not so clear for the ability to amplify or simulate pain expression: while it might bring advantages for the individual in pain but at a fairly trivial level of successful social cheating, for the onlooker, the ability to detect that pain was amplified, or much more so, simulated in its absence, could have important implications for spending of precious resources. Operant theory would make far stronger predictions about amplification and simulation: on the assumption that more often than not pain expression brings benefits, the ability to simulate and amplify would be rewarded, unlike the ability to suppress. Again, it is not clear what operant theory would predict for the observer detection of amplified and simulated pain expression.

3 Evolutionary psychology

An account is proposed which draws on evolutionary studies of social behaviours. Certain pain behaviours, or rather the propensities which underlie them, are likely to have been subject to natural selection insofar as they conferred advantage on the individuals who used them, across contexts from the presence of potential help-givers to that of known antagonists. Of necessity, vigilance to observed pain cues in others co-evolved. The individual expressing pain would derive benefit if expression of pain were reliably followed by actions by observers which promoted recovery and survival; protection from danger; and aid in obtaining basic requirements (Prkachin et al. 1983; Prkachin 1997). If the person in pain might survive rather than not, and the cost to helpers is low, selection advantage follows as with other help and exchange of information - the currency of kin or reciprocal altruism. While the exploitation of others' help achieves maximum payoff in a single encounter, individuals' sensitivity to possible "social cheating" (to the other taking a benefit without paying the costs) protects them from repeated exploitation over a series of interactions (Cosmides & Tooby 1992, 1997). The following section expands this hypothesis and reviews supporting evidence.

Recent development of evolutionary theory in the light of genetics (neo-Darwinism) combines Darwin’s understanding of selection pressures with their application at a genetic level, not on the individual or group (Dawkins 1976). Selection operates at the level of function, not at the level of physical structures or behaviours which subserve the function. 'Fitness is not a psychological goal. Psychological mechanisms have evolved as a means to the end of fitness in the social and material environment of evolutionary adaptation" (Daly & Wilson 1991, p. 164). One or more genes may be expressed in complex contingent behaviours dependent on contextual variables, including the frequencies of alternative behaviours, or alternatively as universal ("fixed") phenotypes, in which case no variation is evident now. Applied to psychology, this has led to the study of cognitive and behavioural routines as "innate, specialized learning mechanisms that organize experience into adaptively meaningful schema or frames" (Crawford 1989, p.16). Such mechanisms are likely to be specific rather than general (Tooby & Cosmides 1990), and related to environmental demands, threats and challenges in that they "focus attention, organize perception and memory, and call up specialized knowledge required for domain-appropriate inferences, judgments, and choices. In the context of life history theory, they are mechanisms that mediate environmentally contingent tactics by processing information" (Crawford 1989, p.16). Skinner likened evolution by natural selection to learning by trial and error (Maynard Smith 1995): success is defined as an increase in the frequency of the target outcome. Behaviours which promote survival and/or reproductive advantage thereby maximise the likelihood of passage of their owner’s genes to the next generation: selection shapes bodies and minds to the conditions within which they exist (Daly & Wilson 1988). So proposed rules must be tested not against an apparent fit between current behaviour and environment (Tooby & Cosmides 1992), but to the presumed ancestral environment(s) and probable variations in behaviour, perceptions, or propensities.

What criteria can be used to address the suggestion that a behavioural routine results from natural selection of an underlying propensity? Selection advantage relies on benefits outweighing the cost incurred by any action, so a convincing case must be made for those benefits, as in the detection of fear. That in turn requires that the behaviour itself is specific, and that it is detected and interpreted by others. Tooby and Cosmides (1990; 1992) proposed further criteria required to support the claim that a behaviour, or rather the propensity governing it, was an evolved adaptation: the specification of process and mechanisms by which a change in the frequency of the characteristic in the population could occur; a task analysis of problems in the ancestral environment which were solved by the adaptation; a description of the adaptation in terms of mechanisms or algorithms, rather than behaviour; and where possible an analysis of how the adaptation is manifest under current conditions where they differ from the ancestral environment.

These criteria are addressed here with reference to facial expression in general terms, with further detail on pain facial expression later. The same selective pressures towards better communication between group-living pre-humans and humans which brought about the development of language (Pinker 1994) would have operated for conveying information by nonverbal means, including facial expression (Fridlund 1994). Altruism to genetic kin in proportion to their kinship (Williams 1966) and reciprocal altruism, the exchange of help over repeated encounters to the benefit of both parties (Trivers 1971; Hamilton 1971/1996), explain help giving in terms of adaptation. The latter has been developed as social contract theory (Cosmides & Tooby, 1992) of psychological adaptations for mutually beneficial and reciprocal exchanges. This can only be sustained if both parties in the exchange are alert to defaulting (social cheating): to taking a benefit without paying the cost (reciprocation or exchange), or without being entitled to it (Cosmides & Tooby, 1992; 1997), in which case nonreciprocators are recognised and excluded from help-giving (Hamilton 1971/1996; Frank 1988; Cosmides & Tooby 1992; Tooby & Cosmides 1996). By contrast with exchange-based mechanisms, a further adaptation concerned with the giving and receipt of help is the cultivation of friendship and irreplaceability within close social groups (Frank 1988; Tooby & Cosmides 1996) such that when a member of that group is in dire need (and for the same reason unlikely to be able to reciprocate), others’ stake in the needy individual’s survival ensures their attempts to help. Reliable communication clearly provides the means for such adaptations to evolve.

The "problem" to which facial expression of pain is suggested as a solution is twofold, concerning benefits to the signaller and to the onlooker. In certain cases, these differ for immediate response to threat of pain or pain on injury and for the responses during the recovery phase, but it would be mistaken to imply that these are entirely separate since the recovery phase is punctuated by acute pains, related to necessary movement, body functions, or pathological processes at the injury site.

The benefit to allies who receive information about potential danger is exemplified by the countless examples across species of alarm calls, and in humans and some primates, vicarious learning from fear signals of conspecifics has been well demonstrated (Mineka & Cook 1993; Vaughan & Lanzetta 1980). Information is of value for safety and survival, since in any set of circumstances there are many possible courses of action, but few are beneficial (Tooby & Cosmides 1990). Although there are exceptions, where information about painful danger is not specific enough for subsequent avoidance (e.g. slow-developing inflammatory responses), attention to cues associated with pain would generally be advantageous. There are also likely benefits to an antagonist competitor or predator in detecting pain in the other, in order to press advantage by a variety of means. Over the longer time frame of recovery, the benefits and costs to the friendly onlooker are best described within the framework of help relationships described; there is no separate case to be made for the antagonist onlookers in the recovery phase since this situation returns the that of immediate threat.

Benefits to the signaller in pain are more varied and uncertain, and must be set against the resource costs of signalling and the possible risk of increasing vulnerability. For the individual in pain and under threat or during recovery who expressed pain in the presence of observant allies, help and protection rendered could be crucial for survival. The skeletons of Neanderthals (two hundred thousand years ago) and those of Homo erectus and early Homo sapiens (around one hundred thousand years ago) show healed major bone fractures (Pinker 1994; Mithen 1996; Hamilton 1971/1996), compatible with receiving help from other humans while the injured party could not fend for him or herself. For the individual in pain in the presence of antagonists, the acute expression of pain might disclose weakness and increase vulnerability, to the cost of the individual in pain. Such a disclosure could also initiate a negotiated withdrawal to safety (Gilbert 1992), with obvious benefits for the individual's survival.

Various selection pressures on the individual's capacity both to exert voluntary control over the amplitude of facial expression (suppression and amplification), and on the capacities of observers to identify pain and its approximate severity despite those attempts, can be posited for the injured individual and for his or her allies and antagonists. Although facial expression does not require consciousness on the part of the signaller to communicate a message to onlookers, consideration below focuses mainly on the conscious modulation of the facial expression of pain, since that is where empirical evidence is concentrated. It extends as far as is possible towards formulation of behavioural patterns and information processing considerations. The interesting field of reflexive aspects of facial expression (facial feedback), that is, of change of subjective and physiological measures in relation to voluntary modification of pain facial expression (Lanzetta et al. 1976; Ekman 1993), is not covered due to the lack of research in pain.

At least partial control for the individual in pain over the amplitude of his or her pain expression allows strategic use of the signal; the ability of the observer to make at least crude distinctions between degrees of pain, without any specific information being added, is also advantageous. The individual in pain might gain both by amplifying pain expression better to convey need to allies and by suppressing pain expression to hide vulnerability in the presence of antagonists. Simulation of pain in its absence may represent a different signal and require other explanations than those that apply to amplification, or it may be an extension of it. Unfortunately, few studies of simulation have been carried out. For their part, group-living allies need to detect pain in any one of their members when survival of all depends on the effectiveness of each. Antagonists benefit from information about another's pain despite attempts to suppress it, whether they press the advantage or withdraw. Construction of the likely advantages and disadvantages of simulation of pain are of necessity more speculative. The benefit for the individual in being able to simulate pain convincingly may lie in the receipt of help, in the chance to withdraw from the threat of physical damage from an antagonist, or in the possibility of relief from aversive physical tasks. However, humans are alert to social cheating, as described above, of the sort represented by feigning of pain to gain material or tactical advantage, so that in any stable group the simulation of pain would yield rapidly diminishing returns and the withdrawal of help. Costs to the help-giver who believes simulated pain to be real may be wasted energy expenditure or loss of face; it is less likely to have implications for survival.

The major work in the area of facial expression has been that of Ekman and colleagues, active for over 20 years. Ekman proposed distinct faces expressing distinct "basic" emotions, each of which has a particular adaptive function and whose expression is modified by socially learned "display rules" ("culture-specific prescription about who can show which emotions, to whom, and when": Ekman 1993). Subsequent research has largely supported this proposal, with the strongest evidence for facial expression of happiness, surprise, fear, anger, disgust, and sadness, which show distinctiveness, rapid pre-awareness onset, brief duration, involuntariness, and automatic appraisal by observers (Ekman 1992). Facial expression is under both cortical and subcortical control (Rinn 1991), although it is mistaken to suggest that "spontaneous" or nonvolitional facial expression is necessarily unaffected by overlearned display rules, and a distinction has been proposed between the degree of modification by display rules and the degree of learning involved in the modification (Matsumoto & Lee 1993). Connections run from the limbic system and neocortex to the facial nucleus, in which musculotopic organisation (over upper and lower facial movement) has been identified (Morecraft et al. 2001). While the limbic system is involved in facial nerve activity in spontaneous facial expression, its involvement is less well established for posed expressions (Argyle 1988). Further support comes from the evidence that humans perceive emotional expression categorically, although muscular changes in the face occur along continuous dimensions (Etcoff and Magee 1992; Young et al. 1997), like colour perception, where recognition appears to proceed by fit to cognitive categories while not excluding distinctions made within categories (Young et al. 1997).

Fridlund’s behavioural ecology model (Fridlund 1994) builds on Ekman's work but is more explicit about the role of expression in social negotiation with kin, allies or antagonists (see also Hinde 1985), and thus on the coevolution of observers’ attention and interpretation. "The form and intensity of a display, and the alignment of the display with personal resources or intentions, would depend on contextual factors such as common interests, availability of resources, and presence of kin, intruders, or predators" (Fridlund 1997, p. 113). Thus a "fear" face signals readiness to submit; a "felt" smile readiness to affiliate; a "false" smile readiness to appease. For the observer, both hypervigilance and hypovigilance have potential costs. Vigilance for displays requires sensitivity to them (as in electrophysiological evidence for speech and face recognition), selectivity about particular display components, and skepticism about the meaning (might it be deception, or random noise?). Despite apparent automatic processing of the emotional valence of faces (Stenberg et al. 1998), disparity between the relative ease of identification of posed expressions compared to spontaneous expressions when observers have no context (Motley & Camden 1988) gives some support to the importance of context. Debates on the correct model continue: for instance, Russell (1997) emphasises context rather more than Fridlund, proposing that dimensions of pleasure and arousal are used to parse the face, and then specific emotion attributed according to the context and knowledge of the person; whereas Buck (1994) envisages a model combining notions of emotional "readout" and of contextual interpretation.

To what extent can work on facial expression of emotion be extended to pain? Fundamental questions concern the extent to which it is a specific expression, whether it is detected by observers, and what are the effects on those observers. Most of the empirical evidence comes from Craig and colleagues, whose research questions only partially coincide with those which would be posed from an evolutionary psychology perspective, so that even robust findings may lead only to tentative conclusions. Further sources of evidence are harder to interpret. Fridlund (1994) uses morphological comparison across species in his study of social smiling, but studies of primate pain expression and its morphology have not been done (Universities Federation for Animal Welfare, 1989; National Research Council 1992), and discontinuity in morphology would not disconfirm the hypothesis of an evolved pain expression.

The extent of difference between the acute pain face and chronic pain face is an unanswered empirical question, but the two functions are related and are likely at times to coincide. LeResche et al. (1992) found the frequency of chronic pain patients' facial action units (FAUs) was higher for those with pain of longer duration at baseline and during experimental and clinical pain, and interpreted this as evidence that "pain facial expressions are subject to reinforcement and become more frequent over the course of this chronic pain condition: that is, they are indeed chronic pain behaviors." More systematic observation and ecologically valid studies are urgently needed, although the latter pose considerable ethical challenges. Since the study of facial expression in enduring emotional states is generally neglected (Ekman 1992), no hypotheses can be drawn from this quarter. In this paper I have chosen to describe pain expression without making a sharp distinction; it remains to be seen whether it would be better described as related expressions which may often occur as a blend. While the description above is of the acute (rapidly habituating) pain facial expression conveying threat and alarm to onlookers, it might also serve to elicit immediate help from allies and those with an investment in the person in pain. The chronic pain face is relatively undescribed, and it is difficult to do so with no painfree baseline for comparison (since the pain is by definition intractable), and also because it is more difficult to isolate from expression of associated emotions. Moreover, even without experimental or clinical manipulation, chronic pain is punctuated by acute exacerbations and it is these which have been investigated experimentally. However, in these studies chronic pain patients at baseline show some facial activity associated with acute facial expression of pain (using mean data or their own expression on experimentally or clinically produced acute pain: LeResche et al. 1992), but the experimental setting does not represent a situation which is likely to cue soliciting of help, and instead may produce some apprehension about expected pain.

Universality of an expression, across cultures, sex and contexts, is a requirement. But it presents considerable methodological difficulties (Russell 1994; Ekman 1994; Izard 1994), and neither universality of expression, nor cultural difference, constitutes decisive evidence for or against an evolved response rather than one learned in infancy. Low variance across cultures is most apparent for disgust and fear (Ekman et al. 1987), but represents relatively weak evidence of universality. What evidence there is on universality of pain facial expression is discussed below; it is strongest in relation to infantile pain expression. From a related field, congenitally blind children show full facial expression of emotion (Matsumoto & Lee 1993, Fridlund 1994), although this cannot be used definitively to support an innate model of expression since their learning of expression may employ nonvisual channels. Methodological and ethical problems preclude definitive tests. Another possible source of evidence is clinical cases where lesions show the dissociation or association of particular functions or behaviours (Fridlund 1994), but availability of such evidence is largely serendipitous, and interpretation of single cases is fraught with difficulty.

Evidence is reviewed under the headings of distinctiveness, consistency, and the degree of voluntary control. For distinctiveness and voluntary control, both the behaviour of the individual expressing pain, usually using the Facial Action Coding System ² (FACS: Ekman & Friesen 1978), and the judgement of the observer are explored. A further distinction is made on the grounds of ecological validity between clinical participants and events, and experimental ones. Data are described first from painful stimuli (such as a pain-exacerbating movement) applied to chronic pain patients, or painful clinical procedures applied to infants, both of which might be expected to demand attention and to have meaning, and second from experimental paradigms in which volunteers are subjected to a variety of painful stimuli in the context of negligible personal meaning and considerable control over the situation. Nevertheless, only pain-free volunteers can provide data on simulation where the stimulus for pain intensity is controllable.

Darwin (1872/1965) noted specific mouth and eye movements and expression in man in pain: "the mouth may be closely compressed, or more commonly the lips are retracted, with the teeth clenched or ground together... The eyes stare wildly as in horrified astonishment, or the brows are heavily contracted". In fact, the morphology of facial expression in relation to pain does not correspond well to this description. Several studies using FACS (Craig 1980; Craig et al. 1992; Prkachin 1992a) identify a combination of facial actions which appears to be specific to pain, and which they have largely used in subsequent work, including as a single index (e.g. Prkachin et al. 1994). Core action units (AUs) in adults are brow lowering (corrugator: AU4), cheek raise and lid tighten (both parts of orbicularis oculi: AU6 & 7), nose wrinkle and upper lip raise (both parts of levator labii: AU9 & 10), and eye closing (AU43) (Prkachin 1992a; Craig et al. 1992). It is better described as a fuzzy set than a prototype (Prkachin & Craig 1995); but the formulation of prototypes and variants has been retained here for consistency with more extensively investigated facial expressions of sadness, fear and anger (Ekman & Friesen 1978), although there is no "pure emotion" referent for any expression. These prototypes and variants are shown in Table 1: as can be seen, some action units are shared within the distinct composite (LeResche 1982; LeResche & Dworkin 1988). This, of course, describes the acute pain face, against baselines of no pain and of chronic pain; like the facial expression of negative emotion, but unlike other measures of pain, the acute pain face habituates rapidly if there is no further pain stimulus (Craig & Patrick 1985).

Table 1 about here

Facial expression appears to carry unique variance in pain (Patrick et al. 1986; LeResche & Dworkin 1988; Prkachin & Mercer 1989) but there is relatively little data on the extent to which the face encodes a particular component of the pain experience, such as intensity, meaning, or affect. Available data suggest that while facial expression is more sensitive to stimulus intensity than is self-report (Patrick et al. 1986), in chronic pain patients it encodes cognitive and emotional dimensions of pain rather than sensory quality (by comparison with self-report: LeResche & Dworkin, 1988). In an investigation of baseline, clinically produced and experimentally produced pain in patients with chronic pain of recent and longer term onset, LeResche et al. (1992) found that facial expression of those with longer term pain shared a small amount of variance with anxiety and with a tendency to catastrophise.

More complex associations have emerged with several negative emotions. In a FACS analysis of frequency and intensity of pain and emotion facial expressions before and during the acute pain of venipuncture performed on chronic pain patients, Hale and Hadjistavropoulos (1997) found that the intensity of disgust expression increased on venipuncture with pain expression, while happiness expression decreased in anticipation (on swabbing before venipuncture). Overall, pain facial action unit (FAU) frequency and intensity was associated with those of disgust, anger, fear, and a decrease in happiness. The facial expressions of disgust, fear, and anger, plus sadness also occurred with pain expression on FACS analysis of chronic pain patients subjected to painful clinical examination (LeResche & Dworkin 1988), but there was no correlation between pain FAUs and selfreported fear and sadness. Multidimensional scaling of similarity judgements between pain and negative emotion prototype facial expressions showed pain to fall close to sadness and anger, at the opposite end of the axis from fear, surprise and disgust (Kappesser & Williams 2001). In a study of social influence, facial expression appeared insensitive to tolerance or intolerance to pain modelled by stooges. This contrasted with self-report, behavioural and some physiological measures which were modified in the direction of the model’s behaviour (Craig & Prkachin 1978; Prkachin & Craig 1985).

Infants from 25 weeks’ gestation, despite considerable individual variation, show a characteristic pain face (Grunau & Craig 1987; Craig et al. 1993; Stevens et al. 1994; Lilley et al. 1996) consisting of eyes squeezed shut, brows lowered/bulging, deepening of the nasolabial furrow, and open lips, occurring together, with mouth stretched vertically and horizontally and taut tongue. This description applies to infants who are awake at pain onset, with a proportionally lesser response for infants who are lightly or deeply asleep at pain onset (Grunau & Craig 1987); it is more than a spinal reflex (Craig et al. 1988). Facial expression provides the most reliable indicator of pain in the first few months (Johnson & Strada 1986; Hadjistavropoulos et al. 1994) and unlike other behavioural and physiological indicators at that age can be distinguished from responses to other stresses (Izard et al. 1995; Lindh et al. 1997). From 3 years, when children are able to indicate pain intensity by verbal and pictorial means, facial expression accounts for up to 65% of variance in their ratings (Goodenough et al. 1998; Breau et al. 2001). In children and adolescents whose cognitive impairment precludes the reliable use of verbal or pictorial pain measures, facial expression appears consistent with painful procedures (van Dongen et al. 1999; LaChapelle et al. 1999).

In summary, the acute pain face encodes intensity and emotional and cognitive dimensions; since pain often occurs with other negative emotions, comparison with facial expression of those emotions requires further convergent findings and a technology to identify blends of facial expression. Despite these methodological limitations, facial action units are generally satisfactory to describe facial expression of pain, with a fuzzy set rather than a prototype as the aim. FACS methods have described a reliably produced expression, across ages and a range of cognitive capacities, but probably with some specifics related to the particular experimental paradigm and stimulus characteristics, a relevant issue given that these are somewhat narrow by comparison with everyday experience.

Craig (1980) noted strong consistencies in the morphology of facial pain expression from birth through to old age, but sensitive to sociocultural norms and to immediate social context (see also Craig et al. 1992; LeResche et al. 1992). The integrated pain response of the newborn shows facial activity in common with child and adult versions (Craig et al. 1988), although the infant and young child may show less differentiation between pain and fear (Bush & Harkins 1991). Development of the expression with age has not been described in detail, but in comparison of infant and adult pain FAUs, it is important to recognise that what matters from an evolutionary point of view is that the infant’s expression communicates effectively, rather than that it approximates the adult expression (Anand & Craig 1996). Facial expression of pain appears to be largely spared in cognitively impaired elderly people no longer able to report pain verbally or diagrammatically (Porter et al. 1996; Hadjistavropoulos et al. 1996; 1998; 1999; LaChapelle et al. 1999; Benedetti et al. 1999). The issue of communication of pain by nonverbal subjects has been neglected until recently but is clearly of considerable clinical importance (Anand & Craig 1996).

Table 2 about here

The expression appears largely constant across the different experimental pain stimuli of electric shock, cold, pressure, and ischaemic pain (Prkachin 1992a), as shown in Table 2. Intensity and duration of orbital tightening varied with pain intensity across all stimuli, and of levator contraction and brow lowering tended nonsignificantly in that direction. Other frequently produced facial action units (FAUs) are mouth opening (AU25), jaw drop and horizontal mouth stretch (AUs 26 and 27) often coded together as a unit, and oblique lip pull (AU12). All showed change from baseline for all pain stimuli, but the direction of change was not consistent across stimuli (Prkachin 1992a), and it may be that the properties of certain stimuli evoke particular expressions, as in startle to electric shock (Prkachin 1992a). Over a matter of seconds, even with increasing self-report of pain, the frequency of FAUs diminishes (Craig & Patrick 1985). Acute exacerbation of chronic musculoskeletal pain shows a sufficiently characteristic pain expression for consistent data to emerge (LeResche & Dworkin 1988; Prkachin & Mercer 1989; Craig et al. 1991; Prkachin et al. 1994); under these conditions, moderate relationships have been demonstrated between facial expression measures and verbal pain ratings (Craig et al. 1991) and self-reported disability (Prkachin & Mercer 1989). However, whether these would hold for other chronic pains, such as visceral and neuropathic pains, remains unexplored, so it cannot be said with confidence that there is a pain expression characteristic of the ongoing pain in chronic pain, insofar as that can be treated as a stimulus. The question presents serious methodological problems, in particular the lack of comparison condition unless an adequate analgesic is available which, by definition, is not the case in chronic pain.

Cross-cultural studies of facial expression of pain have not been published (Ekman 1989; 1993; Prkachin 1992a) as they have for facial expression of emotions by Ekman and colleagues. LeResche and Dworkin (1984) found a common description of pain facial expression across Western cultures but concluded that "research has not yet even begun to address the issues of universality or cultural differences in facial expressions of pain". Many studies of cultural differences in verbal and behavioural expression of pain have been criticised for disregarding within-group variation to reproduce stereotypes (Prkachin & Craig 1995), with little attention to facial expression of pain or its detection. Methodological problems in this area are considerable (Russell 1994; Ekman 1994; Fridlund 1994; Russell 1995; Haidt & Keltner 1999). Not only is it extremely difficult to find cultural groups which have not been exposed to Western faces and expression, by television and film as much as by direct contact, but the need for translation in both directions undermines experimental rigour. The design of studies is bedevilled by the unhelpful counterposition of innate and cultural explanations, such that only the most extreme hypotheses can be disconfirmed. One good quality study (Rosmus et al. 2000) used a well-developed neonatal version of FACS, and standard cry analysis, during routine immunization at two months of healthy Canadian born Chinese and non-Chinese babies whose mothers differed as expected in acculturation to a Western lifestyle. Differences emerged in one of seven facial characteristics and two of six cry characteristics on needlestick, with no baseline differences. Despite the careful design, interpretation remains difficult. There are differences in the rate of motor development across populations, and although all babies were alert and awake before immunization the Chinese babies (who reacted more) may have been more aroused due to setting differences. Also, the role of culture in infant pain response in general remains unclear.

Overall, the evidence is strongest for a reliable pain expression across the lifespan. Within the limits of experimental conditions, the evidence is good for consistency of facial expression across painful stimuli, in otherwise pain-free or chronic pain subjects. For both, facial action coding systems have proved a very useful tool for analysis of data. However, the question about similarities and differences across cultures has not been satisfactorily attempted, and there are very considerable difficulties in designing a study which is not open to multiple explanations.

The counterpart of a distinctive and specific expression in the individual is its detection and correct interpretation by observers. Judgement of pain in another person relies heavily on facial cues: brow lowering, eye blinking, cheek raise, and upper lip raise account for over half the variance in ratings. These are used consistently by observers to judge pain in adults and in children (Watt-Watson et al. 1990; Craig et al. 1991), but with systematic underuse of some cues, even the most common (Prkachin et al. 1994; Hadjistavropoulos et al. 1996). There is evidence of reasonably accurate identification of pain expression in adults and infants using the specific facial movements described above (Prkachin et al. 1994; Goodenough et al. 1997; Lilley et al. 1996; Lindh et al. 1997; Breau et al. 2001), but the data on distinction of pain in adults and children from other facial expressions, such as those of fear or anger, are scant (Boucher 1969; Huebner & Izard 1988; Keltner & Buswell 1996; Hale and Hadjistavropoulos 1999; Kappesser & Williams 2001). The face in pain is highly salient for observers (von Baeyer et al. 1984; Craig et al. 1992; Goodenough et al. 1997), who ranked eyes the most important cue, followed by brows, eyelids, mouth, head, forehead, and then other body parts (Prkachin et al. 1983). Facial expression has been shown to contribute significantly to observer ratings of adults in pain (Ahles et al. 1990; Hale and Hadjistavropoulos 1999), even when contradicted by a false verbal report of no pain (Poole & Craig 1992). Carers of severely cognitively impaired children (van Dongen et al. 1999) and adults (LaChapelle et al. 1999) report that they use facial expression as an indicator of pain, and children too young to use verbal and numerical pain scales are able reliably to use cartoon faces or photographed faces to indicate their own severity of pain, where severity is indicated by eye and mouth shape and depth of nasolabial furrow (Chambers & McGrath 1998). The capacity of humans, from a young age, to interpret the pain face seems adequately established.

As in other facial expression studies, observers of emotional pain faces show some activity in the relevant facial musculature (Vaughan & Lanzetta 1980; Dimberg & Ohman 1996) which appears to indicate empathy via mimicry (Bavelas et al. 1986; Hess et al. 1998), although inattention to the emotional quality of the expression, as when doubting its genuineness, may impair this response (Hess et al. 1998). The emotional concomitants of recognition of facial expression, and mimicry, have not been systematically investigated (McIntosh 1996) on emotion or in pain.

The involuntary nature of facial pain expression is suggested by the data above, its occurrence in infancy, and by findings reviewed below on attempts to suppress it. The questions addressed in research studies are the degree of voluntary control over amplitude - by instructions to suppress or exaggerate pain expression - and the degree to which pain facial expression can be simulated in the absence of pain. Only one study of the latter was found (Galin & Thorn 1993) and its interpretation is complicated by methodological shortcomings. Although often investigated and discussed together, the task of suppression differs in function from that of amplification and of simulation, so these are likely to differ in the evolutionary pressures to which they were subject. Production and detection of voluntarily modified pain expression are reviewed together, since where the interests of the injured individual and the observers conflict, the eventual balance reflects the significance of the payoffs in evolutionary terms. Studies are shown in Table 3.

Table 3 about here

The extent to which countervailing evolutionary pressures have resulted in the ability successfully to suppress, amplify, or simulate pain expression presents methodological difficulties for the design and interpretation of studies. First, FACS methods do not quantify aspects such as timing, speed, complexity and asymmetry of expression which are important in deception (Prkachin 1992b; Ekman 1993), and which are likely to affect observer judgements. Second, emotional expressions posed by actors (often for the production of prototype photographs) appear to mix spontaneous and deliberate elements, given some "method acting" component (Hess & Kleck 1990; 1998), and this complicates the use of facial action prototypes to analyse data. Deceptive intent on the part of posing subjects may introduce elements of facial expression related to deception (embarrassment, shame, guilt) rather than to the emotion of interest (Hess & Kleck 1998). This can affect both FACS and judgement studies. Third, Fridlund (1994) has made the telling criticism that these distinctions (in/voluntary, non/emotional, posed or deliberate/spontaneous) may bear no relationship to neural organisation: an evolutionary programme of research would use more naturalistic settings and address evolutionarily stable strategies and cooperation, in addition to those on manipulation and deception which have so far dominated.

Several studies have used pain patients undergoing clinical examination or executing movements likely to exacerbate pain. In these circumstances, facial expression of pain is detectable despite the instruction to subjects to try to conceal it (Poole & Craig 1992; Hadjistavropoulos et al. 1996), although overall activity may be subtle and differ only slightly from no pain and unmodified pain expression (Hadjistavropoulos & Craig 1994), or may appear only in specific action units (mouth opening: Hadjistavropoulos & Craig 1994; eye narrowing: Craig et al. 1991; Prkachin & Craig 1995). An extra and uncharacteristic facial action of outer brow raise has also been reported (Craig et al. 1991). These studies are summarised in table 3. In the only FACS study of healthy volunteers instructed to conceal pain expression (Galin & Thorn 1993), suppressed pain expression shared the frequency of particular FAs both with pain-free and genuine pain expressions (see Table 3), but interpretation must be cautious as the selection of sections for coding may have provided habituated rather than immediate facial expression of pain.

Two judgement studies on observer identification of suppressed pain, using chronic pain patients instructed to suppress pain expression during painful manoeuvres, produced mixed findings. In one, suppressed pain was more often identified as real than as no pain, when these options were provided (Hadjistavropoulos et al. 1996), but judges found the discrimination difficult and rated their confidence in their judgements relatively low. In a second study, observers identified "leakage" of expression around the eyes (Poole & Craig 1992). The studies using induced pain in healthy volunteers differ somewhat: in one, judges distinguished levels of pain stimulus intensity using facial cues (Patrick et al. 1986); in another (Galin & Thorn 1993), observers appeared to distinguish genuine from manipulated pain expression, although they were not necessarily accurate about whether the manipulation was suppression or amplification. It is probable in this study that some FAUs represented emotion related to deception. In a third (Lanzetta et al. 1976), levels of pain intensity were apparently not distinguished by observers but a uniformly low pain rating given across levels.

In summary, the FACS data are reasonably consistent on the appearance of pain FAUs when subjects attempt to suppress pain, with reduced frequency of FAUs in the upper and lower parts of the face, and of reduced intensity mainly in the lower part. The data from judgement studies is more mixed, although there are indications that observers detect more than they can confidently report. It may be that in real situations, this is sufficient to instigate a check on the context (and the face) for further information.

In chronic pain patients subjected to extra painful movement, the instruction to exaggerate led to the use of several FAUs not prototypical in spontaneous pain expression (inner and outer brow raise, lip corner pull, chin raise). Lip stretch and eye closing occurred more frequently than in genuine pain, and blink less frequently, with overall higher frequency and intensity of FAUs characteristic of genuine pain (see Table 3) (Craig et al. 1991; Hadjistavropoulos & Craig 1994; Hadjistavropoulos et al. 1996). There are no published FACS studies of amplification in healthy volunteers subjected to painful stimuli, only of simulation (below).

Although two studies showed detection of amplification from genuine pain at a better than chance level, in a clinical population (Hadjistavropoulos et al. 1996) and in a volunteer population instructed to exaggerate pain (Lanzetta et al. 1976), amplification is not necessarily well discriminated by judges (Poole & Craig 1992). Nor is it clear what judgement heuristics they use: they do not consistently correspond to atypical facial movements identified by the use of the FACS (Hadjistavropoulos et al. 1996). Reference to table 3 suggests that amplified and simulated expression show most discrepant activity in the upper face, with little change in the lower face. Experimental conditions and instructions may have an important influence: when primed to expect deception (suppression and amplification of pain expression) untrained observers were no better at accurate detection of the actual condition, but gave more conservative judgements across all conditions: suppressed, genuine and amplified pain expression (Poole & Craig 1992). Observer judgements can also be influenced by priming with information suggesting subject hypersensitivity or analgesia (Prkachin et al. 1983). This suggests not so much detection of simulation but rather modification of judgements based on a good-enough reading of cues.

It is hard to summarise these rather diverse data, and the likelihood is higher in amplification than in suppression of contamination from two sources. Facial movements associated with deception (timing, overlap and asymmetry) are not identified by FACS methodology but may influence observer judgements. This could explain the discrepancy noted in the study which used both (Hadjistavropoulos et al. 1996). Second, emotions related to deception (shame, guilt, embarrassment) can be described and identified using FACS methodology but are not investigated in any of the pain studies reviewed above which use FACS; again, they are likely to affect observer judgements.

Simulation of pain in its absence may be a version of amplification rather than a separate function: there are no data with which to address the question at this stage. While in general simulation of emotion produces an impoverished expression (Gosselin et al. 1997), the only study of simulation by pain-free subjects suggested, on FAU analysis, more intense expression than genuine or baseline pain, consisting of more brow lowering, cheekraise, tightening of eyelids, closing eyes, and tightening lips (Galin & Thorn 1993). While identification was not quantified, feedback on accuracy of judgements appeared to improve it, but it is not clear what cues judge adopted (Galin & Thorn 1993).

Having reviewed what properties of the facial expression influence observer judgements, we turn to properties of the judge. Relatively little research has addressed the context of the judgement: the relationship of the observer to the person expressing pain, the information and beliefs which the observer consciously or unconsciously brings to the judgement, and the potential costs and benefits for the observer of erring in either direction. Such issues are crucial to an evolutionary analysis (Tooby & Cosmides 1990). In empirical studies, observer accuracy is usually judged by the degree to which the observer’s rating matches the subject's rating of pain. In fact, while subjective rating is important if idiosyncratic (Williams et al. 2000), there is no gold standard for pain measurement, nor can pain be represented unidimensionally. Further, this method makes the assumption that the face and the subjective rating express the same dimension/s of pain experience. Exploration of other facial expression, such as sadness, fear, and anger, has not focussed on quantification: it appears to be a peculiarity of pain, related to clinical concerns about judging authenticity.

With a few exceptions there is little variability among observers within a particular group in their accuracy (Hadjistavropoulos et al. 1996). Most judgement studies show "underestimation" of pain in relation to FAUs and the subject's own pain estimate and this is particularly marked where the judges are clinicians (Prkachin 1997). Underestimation of pain by reference to the sufferer’s rating is widespread among health professionals and staff who care for elderly people (Forrest et al. 1989; Grossman et al. 1991; Marzinski 1991; Keefe & Dunsmore 1992; Madjar & Higgins 1996), with some contrasting findings of close approximation (Grossman et al. 1991) or slight overestimation of low levels of pain (Zalon 1993). Inadequate analgesia, across ages and illness severity (Marks & Sachar 1973; Schechter 1989; Wilson & Pendleton 1989; Marzinski 1991; Finley et al. 1996; Bartfield et al. 1997; Demyttenaere et al. 2001) constitutes an imajor clinical problem (Keefe & Dunsmore 1992). A major cancer researcher has even suggested training patients to express pain in ways which doctors and nurses currently recognize (Keefe & Dunsmore 1992).

The common explanation for the phenomenon of underestimation by health professionals is that repeated exposure to pain and professional training develop a relative insensitivity to pain and/or a scale with a higher upper limit. This is used, for instance, to explain greater underestimation among more experienced staff (Prkachin et al. 2001; Mason 1981; von Baeyer et al. 1984). However, while observation of patients with high levels of pain may extend the clinician's reference scale for judging pain intensity, the notion that repeated exposure in itself results in lower judgements of pain is contradicted by the findings concerning spouses and caregiving family members as judges. While accuracy is very variable (Madison & Wilkie 1995), overestimation is at least as common as underestimation, although all studies involved cancer pain (Ferrell et al. 1991; Clipp & George 1992; Dar et al. 1992; Madison & Wilkie 1995). In a study by Prkachin et al. (2001), students with family experience of chronic pain showed a closer approximation to stranger-patients' estimates of their own affective rating of pain than did students with no such experience. This would seem to suggest that exposure to pain improved accuracy, undermining the hypothesis of developed insensitivity. Neither does it support the exposure hypothesis, since in the same study, health professionals estimated patients' pain less accurately than did the students with no family experience of pain (Prkachin et al. 2001). Differences appeared to lie in weighting of what was observed, rather than in sensitivity to variation in facial expression per se: all groups were more accurate at higher levels of pain. The role of feedback (so that the observer can compare his or her estimate with the sufferer's) is unclear: one study found that providing patients' own ratings resulted in more attention to facial expression than when they were unavailable (Poole & Craig 1992), but accuracy was not estimated; another found that training had relatively little effect (Solomon et al. 1997).

In normal infants and children, nonspecific signs such as respiration, activity and vocal cues are used as well as facial behaviour by nurses (Pigeon et al. 1989; Vetter & Heiner 1996; Craig et al. 1996; Hamers et al. 1996; Goodenough et al. 1997; Howard & Thurber 1998; Hudson-Barr et al. 1998). The caregivers of non-verbal cognitively impaired children and young adults attended most to vocalization; facial changes of some sort were noted by 80%, with changes in the eyes the most frequent at 50% (McGrath et al. 1998), but it is possible that there was a lower frequency of face to face interaction than between verbal children and their parents or caregivers. Underestimation of pain in infants and children occurs both in health professionals (Selbst & Clark 1990; Vetter & Heiner 1996; Hudson-Barr et al. 1998) and less consistently in parents (Craig et al. 1996; Chambers & McGrath 1998; Chambers et al. 1998): both nurses’ and parents’ pain ratings may be unrelated to those of the children rated, through overestimation and underestimation (Demyttenaere et al. 2001). Actual detection of pain from facial expression in photographs of neonates undergoing painful and nonpainful stimuli was poorer in health professionals than in a control group who were not health professionals (Balda et al. 2000), the association remaining even when the number of children in health professionals’ families was controlled. An elegant study using child patients’ own ratings of current pain and threshold at which they desired analgesic indicated serious undermedication in the first few days following major surgery, a very common clinical event (Demyttenaere et al. 2001). While inadequate analgesia is likely to arise from underestimation of pain, a study by Hamers et al. (1996) found that more experienced nurses were less willing than less experienced nurses to give analgesics to children, although their estimates of pain did not differ. While exposure through professional experience is a possible explanation for these findings, it cannot account for the inconsistent estimations by parents.

If experience and training do not improve accuracy, but instead professionals with responsibility for providing comfort develop increasing conservatism in pain estimates with exposure to strangers in pain, what biases are affecting judgement? There is some experimental work which casts light on this, albeit not necessarily using facial expression.

Contextual variables can affect judgements of pain and the decisions which arise from them. In studies using written vignettes, both students and experienced physicians who estimated pain in cases with differing medical evidence showed an interaction between the pain level provided in the vignette and the existence of medical evidence, with discounting of high pain levels where medical findings were absent (Chibnall & Tait 1995; Chibnall et al. 1997; Tait & Chibnall 1997). Although an invalid inference, expected severity of pain from an identifiable injury may influence observer judgement of pain in adult patients (Todd et al. 1994; Wilson & Pendleton 1989; Hadjistavropoulos & Craig 1994), and consistent with this, the two studies which showed reasonable approximation of clinician to patient estimate involved an authenticated basis for pain, postoperative (Zalon 1993) or cancer (Grossman et al. 1991). Some biases are less contextual than characteristic of the observer. Observer beliefs concerning employment and risks of opioid addiction may also affect judgements of pain and analgesic requirement by physicians (Carey et al. 1988; Marks & Sachar 1973), who also rate less pain for more physically attractive pain patients (to whom better health is consistently attributed) than for less physically attractive pain patients on videotape (Hadjistavropoulos et al. 1995). By contrast, spouses who believed their spouse-patient with cancer to minimise his or her pain tended to overestimate pain severity.

In children, contextual variables are also used: parents and nurses rated pain on venipuncture in sick children lower for children who were more experienced with venipuncture and higher where venous access was difficult (that is, more damage was inflicted), although neither was represented in the young patients’ own pain ratings (Manne et al. 1992). The same study showed parents’ pain ratings to be significantly predicted by their own anxiety and expectations of painfulness taken before the venipuncture, but a similar examination of parents’ ratings on immunization of healthy children found that parents’ anxiety had no influence on pain ratings (Breau et al. 2001).

These findings suggest that observers use context and beliefs or judgement biases to modify information from the facial (and other) expression of pain. Where facial expression goes unnoticed or uninterpreted and patients cannot express pain verbally, even unambiguous medical information may not lead to the suspicion of pain. For instance, Marzinski (1991) found that while only one quarter of the residents of an Alzheimer’s unit with typically painful conditions, including metastatic cancer, routinely received analgesia; Sengstaken and King (1993) found that chronic pain problems went unrecorded and unmedicated by the attending physician in over one third of communicative residents of an intermediate care geriatric nursing home, and given that few noncommunicative residents were identified as suffering from chronic pain, the authors estimated a high level of underdetection. Experimental studies represent ideal conditions for judgement of facial expression: judges attend under instruction to selected videotapes of subjects. In clinical settings, detection also depends on the frequency of spontaneous expression, which in cancer patients moving in ways which exacerbated their pain approximated one expression every two minutes detectable on videotape (Wilkie 1995). The clinician divides his or her attention between the patient's face, body, notes and other sources of information and may miss brief facial expressions. Clinicians may judge chronic pain patients at rest by reference to an expected acute pain face and therefore discount the patient's report of pain; they may also observe an expression which blends pain with fear, embarrassment, anger, disgust and sadness: confusion of one emotional expression with another can be investigated (e.g. Keltner & Buswell 1996; Stenberg et al. 1998; Kappesser & Williams 2001). In addition, observers' attention to the question of veracity may suppress an affective or empathic response (Hess et al. 1998), degrading the quality of judgement made, a condition also accessible to empirical investigation.

The information here is consistent with the evolutionary model elaborated in section 3. The observer who is a potential helper must be alert to the possibilities of social cheating, and give help (benefit) only to those who pay the cost, that is, who genuinely have pain. So they look for confirmation, such as evidence of injury or clinical signs of structural damage or pathology. However, where these are weak or absent, and where benefits are believed to be of considerable value to the person complaining of pain (such as sanctioned and paid absence from work, entitlement to welfare support proportional to severity of disability, or access to abusable drugs), suspicion is heightened and the judgement of pain modified. Even in the experimental setting, priming of observers by warning that they will see no pain, suppressed, genuine and faked pain expressions in chronic pain sufferers undergoing an acute exacerbation with movement results in underestimation of pain across conditions (Poole & Craig 1992). The experimental subject judging a stranger has nothing to lose by overestimating pain; the parent, spouse or family member may need to protect against repeated exploitation or may be concerned to provide fully for the patient's needs, and the data above offers fewest clues to the influences on their judgements; the health professional is gatekeeper for many benefits, risks losing face among his or her peers if s/he appears gullible to patients' complaints, and may be subject to professional or legal investigation if s/he prescribes opioids much more generously than his or her peers. While the operant model of pain cannot be held responsible for the requirement that health professionals and others must make credibility judgements, literature arising from it has contributed to the attribution of agency to a variety of contingent benefits in the complaint of pain. This is further developed in the next section.

In integrating the hypotheses and data presented, it is important to hold in mind that "Evolution does not create specific behaviours; it creates mental organisations and inference systems that make people behave in particular ways. The inference systems are activated whenever information in a certain format from the environment is presented" (Boyer 2001, p.268). This paper has attempted to examine what information presented by the environment activates or deactivates a hypothesised inferential system or evolved propensity. My aim is to make as explicit as possible the biases which operate in the production and judgement of facial expression of pain. It should be noted that neither operant nor evolutionary explanations should be taken to imply conscious intent in the modulation of pain expression or the judgements made by observers acting in accordance with evolved propensities (Daly & Wilson 1991), although cooperation with certain experimental conditions requires conscious intent in the signaller, and judgement of conscious intent is required to diagnose malingering.

To what extent does the evidence support contextual evolved propensities or inference rules for the production, detection and interpretation of facial expression of pain? For the strategic expression of pain in the presence of caregivers, there are suggestive findings on which it would be unwise to rely heavily: both Kleck et al. (1976) and Badali (2000) showed that pain expression was greater when subjects were alone than when they were observed or they believed themselves observed by neutral onlookers. This suggests an important thought experiment or prediction: that the increase in pain behaviour observed in the presence of solicitous others (Block et al. 1980; Flor et al. 1995) and health care staff could be the release of suppression of pain expression, rather than its amplification. A parallel could be found in the expression of grief, where temporary loss of an attachment figure in infancy and young childhood produces a consistent pattern of emotional responses including expression of grief and anger to the attachment figure when s/he reappears, but not reliably to other adults available in the interim (Weiss 1991). In adulthood, grief is expressed among kin and friends rather than among strangers or alone (Parkes 1972), although with some cultural variation; and recent research on shared and nonshared negative emotional experiences confirms the tendency for individuals to mask and conceal emotion which they choose not to share (Finkenauer & Rimé 1998). More systematic work in the facial expression area concerns smiling, which was better predicted by the interactive nature of the social situation than by happiness ratings across social and nonsocial situations (Fernandez-Dols & Ruiz-Belda 1997), and the apparently inhibitory effect of strangers on facial expression of emotional reactions to pleasant and unpleasant images (Buck et al. 1992). Better understanding of pain facial expression may offer insights beyond pain: a broader appreciation of facial expression would encompass need states as well as emotion.

Findings of greater facial expression of pain in the presence of potential caregivers than in their absence is compatible both with the evolutionary hypothesis that the presence of potential caregivers prompts the release of suppression of pain facial expression, and with the operant hypothesis that pain expression is amplified on the basis that amplification has been rewarded in the past. Nevertheless, distinguishing between the two is of considerable clinical importance. Until it is established, proponents of the operant model in pain would do well to express caution about the unscientific use of the model in the imputation of secondary gain when a patient shows more pain behaviour in clinical interview than when s/he believes him/herself unobserved: ambiguity over whether such communication is intentional allows the assumption that behaviours are deliberately produced, or not suppressed, an assumption that is to the serious disadvantage of the patient and contributes to the judgement of malingering.

A propensity characterised as "Don’t hide pain in the presence of a potential caregiver", would encourage free expression of pain in the presence of solicitous others and of those whose professional duty it is to attend and minister to pain, as well as under experimental demand. In the presence of others not expected to provide support of any sort, or alone, facial expression of pain would tend to be suppressed. There is some support for this in the finding that for children, while the presence of parents may disinhibit some pain expression, they rate the pain experience as less than when parents are absent but expression is reduced (Stoddard 1982 cited in Craig et al. 1996). Evolutionary pressures however will have acted on expressiveness in general, not only of pain, so that the demands of pain expression and detection will not be its only determinants.

What of the observer? Fridlund's (1994) framework of the requirements for the observer to show sensitivity to displays, selectivity about particular components of the display, and skepticism about its meaning may provide a heuristic for further investigation. There appears to be some capacity to detect pain despite attempts at suppression, and observer estimates of pain at low levels can approach or rarely exceed the sufferer’s estimate. However, observers may also tend to discount pain, more at higher pain levels, when there is not supportive evidence of injury, and when primed to expect deception. Professionals making judgements of pain tend to have more confidence in their judgement than is warranted (Lander 1990; Goodenough et al. 1997). On the basis of FAU findings, suppressed pain appears more likely to be identified by observers as pain present than simulated pain is identified as pain absent (Hadjistavropoulos et al. 1996), consistent with the costs and benefits in evolutionary terms. An anti-cheating device to protect the potential help-giver might take the form of specific detection of amplified or simulated pain, for which the experimental evidence is equivocal at best; or it might take the form of a general conservatism about estimating others’ pain which is influenced both by contextual information and by personal beliefs. As described above, the cost to health professionals of overestimating pain (and overprescribing treatment) is considerably higher, and therefore more warranting conservatism, than for neutral onlookers. Tellingly, when health professionals predict the extent of pain to be expected from medical procedures they do not necessarily underestimate it (Hodgkins et al. 1985): it is their reading of facial expression which results in underestimation. While training sensitivity (Solomon et al. 1997) may offer some possibility of improving accuracy, recognition and correction of systematic bias might offer more. For other caregivers, it should not be assumed that the needs of, for instance, children and their parents necessarily complement one another: parents and caregivers may tend towards conservatism in order to avoid exploitation. The evidence does support the generalisation that observers whose power to dispense benefits, such as health professionals and parents but not spouses or strangers, is far greater than that of the patient to obtain them by other means are more likely to underestimate pain and/or to underprovide related benefits such as analgesics. In the Dar et al. (1992) study, spouses who overestimated patients' cancer pain were also more liberal about use of opioids.

This paper constitutes an exploration and elaboration of ideas about pain and behaviour associated with pain in an evolutionary perspective, using available findings on facial expression of pain (albeit none arising directly from the framework of evolutionary psychology). Applying the criteria (described in section 3) for appraising the hypothesis that a specific behaviour or set of behaviours is the expression of an evolved propensity (Tooby & Cosmides 1990; 1992), there is good evidence that pain facial expression is specific and that it is distinguished from other facial expressions by onlookers. The balance of costs and benefits for the person in pain and the onlooker have been described for the acute threat phase and in the later recovery phase, as have their implications for survival, such that their advantage under natural selection requires no additional argument. Finally, a hypothesis has been proposed concerning the cognitive inferences which would govern the behaviour. The model draws on established work on mechanisms of altruism and help-giving, and in that context on the specific sensitivity to social cheating.

Many important questions remain. Some concern the content of the facial expression of pain: distinctions between acute and chronic pain states, and the extent to which context affects the encoding of different components. Pain varies in quantity, quality, temporal properties, and affective tone, while experimental work mainly addresses only pain intensity. These different dimensions are also differently represented in self-report, making convergence an empirical question rather than an expectation. Further questions concern other pain behaviours, such as limping, guarding, or rubbing, which while clearly more accessible than facial expression to voluntary control, like facial expression should not unquestioningly be assumed to be intentional and manipulative. Then there is the relationship between facial expression and other pain behaviours, since they generally occur together: it has been suggested that other motor behaviours add emphasis (Argyle 1988), even to the extent of a full display (writhing, thrashing) demonstrating "no cheating, no faking", given the high costs of activity to an already stressed organism (Bateson 1991). The emotional content of facial and other pain expression is important, and issues of differentiation from, and overlap and blending with, other emotional expressions require investigation (Fridlund & Duchaine 1996; Prkachin 1997). The distinction made by many clinicians in assigning all emotional expression and behaviour to a "psychological problem" category degrades assessment when it is intended to refine it.

Considerable methodological problems explain in part the lack of evidence on these questions. Ideally, dynamic recording and encoding methods could provide information on sequence, timing and duration of expression; spontaneous and deliberate expression differ (Hess & Kleck 1990) in ways which are lost in photographs. A more serious problem is that in laboratory studies of volunteer subjects, the pain has little meaning and high predictability and controllability, thus little emotional content or threat value. Additionally, in research on observer responses, posed and spontaneous expressions frequently differ in symmetry, timing, and content (Argyle 1988; Scherer 1992; Hess & Kleck 1997; Borod et al. 1997) and the instruction to pose may produce some signs of deception (such as blinking) which interfere with the posed expression and observers’ judgements (Hess & Kleck 1997).

Empirical and observational studies are needed both on the production of facial expression and on its detection and interpretation, and a major purpose of this paper is to stimulate empirical work as well as theoretical debate. On pain facial expression, evidence is thin on universality (where formulating disconfirmable hypotheses presents substantial difficulties and the study of cultural differences in pain expression lacks an adequate model of evolved and learned behaviours); on primate expression of pain and its social context; on everyday painful events in humans and onlookers' responses to them, preferably covertly observed; on variation in pain facial expression over time and across settings. More ecologically relevant refinement of current methodologies (Carroll & Russell 1996) could include chronic pain patients undergoing examination which exacerbates pain in a clinical setting where the results of the examination have important implications for the patient; and the manipulation of expected cues for suppression or expression of pain, such as the presence of strangers, sympathetic others, or potential professional caregivers. Observed behaviour, rather than self-report, constitutes the datum of interest, and avoids falling into unintended assumptions of consciousness and intentionality, which are not the currency of evolved propensities for behaviour (Daly & Wilson 1991).

Variability in observers' judgements have been described in relation to their beliefs and how those affect their attention, their interpretation, and weighting of various aspects of signals (Prkachin & Craig 1995). Systematic investigation of the contingencies affecting their judgements and responses could test predictions drawn from an evolutionary perspective of costs and benefits to those observers. There may be specific pain-related propensities, as exist for other cognitive processing of crucial information (Prkachin & Craig 1995), such as the apparently enhanced memory for the faces of cheaters (Mealey et al. 1996), and priority processing of angry faces (Hansen & Hansen 1988). The study of malingering at present consists of a critical evidence-based approach (e.g. Mendelson 1992; Chibnall & Tait 1995; Main & Spanswick 1995) and misguided attempts to devise malingering detection tools (reviewed by Fishbain et al. 1998). Despite consistent evidence of a low base rate (Mendelson 1992; Chibnall & Tait 1995), an antitherapeutic preoccupation with malingering and exaggeration persists in the pain literature and clinical and medicolegal practice. Understanding of the phenomenon and of preoccupation with it could be advanced by testing the hypotheses developed above, using methodologies of manipulation of contextual information and material relevant to biases, and adopting methods from studies on cooperation, help-giving, social cheating (Cosmides & Tooby 1992), and deception (Ekman & O'Sullivan 1991). It is important too to address the largely neglected problems of concealment of pain and observers' failure to recognise it or estimate its magnitude. The assessment and management of pain in infants and young children, and in cognitively impaired people of all ages, has only recently attracted interest, and proxy judgements by caregiving adults is still the norm in clinical practice. Facial expression offers the best means for accurate assessment in nonverbal subjects, and an important source of information in verbal ones.

This paper aims to encourage the development of evolutionary perspectives and the generation of relevant data. Of particular concern is the clinical arena in which I believe an evolutionary approach helpfully challenges the dominant operant approach whose overapplication has led to the assumption of benefits and disregard of costs to the person in pain who finds him or herself suspected of factitious complaint, and undertreated. The type of cost-benefit analysis generated by an evolutionary approach applied to the behaviour of the person in pain and of the person who observes him or her will bring better understanding of how conditioning processes affect or do not affect unconditioned behaviours associated with pain. An evolutionary perspective offers much to the pain field, beyond application to facial expression and pain behaviours: a focus on the functions of pain and an integrated rather than separate approach to its various components can only further understanding of pain.

Footnotes

1
These propensities are functionally specialised cognitive adaptations which guide thinking and behaviour in that domain of function, and therefore constitute the proper unit for describing and explaining adaptation. Like many cognitive processes, there is no implication of conscious intention (Daly & Wilson 1991). An example which makes the distinction very clear is given by Cronin (1991): the moth's flying into a candle flame is maladaptive behaviour, but the moth has evolved (long before candles appeared) to fly in a straight line by using effectively parallel rays from the moon, whereas light rays from a nearby source converge on the flame, as does the moth. Widely accepted examples in humans of functionally specialised cognitive adaptations include language learning and face perception. While adaptive in the environment in which they evolved, such propensities can have significant costs in the current environment, as they do for the moth.

2
The Facial Action Coding System (FACS) describes 44 discrete facial actions (of a single muscle or a combination) which can be distinguished by observers and which in particular combinations constitute prototypical facial expressions (see Table 1). They are analysed using trained raters who view sections of videotaped facial expression and record the occurrence, frequency, and/or intensity (on a five point scale) for those movements for which it is appropriate, e.g. intensity for brow lower but not blink. The main advantages of using FACS are that it is theoretically neutral and does not presume any emotion categories (Grunau & Craig 1990); it is well operationalised; and trained raters generally achieve high reliabilities. It has the disadvantages that it neglects duration of pain expression (LeResche & Dworkin 1988), and that it cannot represent temporal sequences, timing, or asymmetry, dimensions of facial movement which are particularly relevant in deliberately manipulated expression, and analysis is time-consuming so that use outside research settings is rare.

Ahles, T.A., Coombs, D.W., Jensen, L., Stukel, T., Maurer, L.H. & Keefe, F.J. (1990) Development of a behavioral observation technique for the assessment of pain behaviors in cancer patients. Behavior Therapy 21: 449-460.

Anand, K.J.S. Craig, K.D. (1996) Editorial: new perspectives on the definition of pain. Pain 67: 3-6.

Anand, K.J.S., Rovnaghi, C., Walden, M. & Churchill, J. (1999) Consciousness, behavior, and clinical impact of the definition of pain. Pain Forum 8: 64-73.

Andersson, G. (1993) Some implications of evolutionary psychology for behaviour therapy. Scandinavian Journal of Behavioral Therapy 22: 49-59.

Argyle, M. (1988) Bodily communication, 2nd edition. Methuen & Co.

Armstrong, D. (1984) The patient's view. Social Science & Medicine18: 737-744.

Badali, M.A. (2000) Effects of audience on self-report, behavioural and physiological responses to the cold pressor test. Unpublished MA thesis, University of British Columbia.

von Baeyer, C.L., Johnson, M.E. & McMillan, M.J. (1984) Consequences of nonverbal expression of pain: patient distress and observer concern. Social Science & Medicine 19: 1312-1324.

Balda, R. de C.X., Guinsburg, R., de Almeida, M.F.B., Peres, C. de A., Miyoshi, M.H. & Kopelman, B.I. (2000) The recognition of facial expression of pain in full-term newborns by parents and health professionals. Archives of Pediatric & Adolescent Medicine 154: 1009-1016.

Bartfield, J.M., Salluzzo, R.F., Raccio-Robak, N., Funk, D.L. & Verdile, V.P. (1997) Physician and patient factors influencing the treatment of low back pain. Pain 73: 209-211.

Bateson, P.P.G. (1991) Assessment of pain in animals. Animal Behaviour 42: 827-839.

Bavelas, J.B., Black, A., Lemery, C.R. & Mullett, J. (1986) "I show how you feel": motor mimicry as a communicative act. Journal of Personality & Social Psychology 50: 322-329.

Benedetti, F., Vighetti, S., Ricco, C., Lagna, E., Bergamasco, B., Pinessi, L.. & Rainero, I. (1999) Pain threshold and tolerance in Alzheimer's disease. Pain 80: 377-382.

Block, A.R., Kremer, E.F. & Gaylor, M. (1980). Behavioral treatment of chronic pain: the spouse as a discriminative cue for pain behavior. Pain 9: 243-252.

Bolles, R.C. & Fanselow, M.S. (1980) A perceptual-defensive-recuperative model of fear and pain. Behavioral & Brain Sciences 3: 291-323.

Borod, J.C., Haywood, C.S. & Koff, E. (1997) Neuropsychological aspects of facial asymmetry during emotional expression: a review of the normal adult literature. Neuropsychology Review 7: 41-60.

Boucher, J.D. (1969) Facial displays of fear, sadness and pain. Perceptual & Motor Skills 28: 239-242.

Boyer, P. (2001) Religion explained: The human instincts that fashion gods, spirits and ancestors, William Heinnemann.

Breau, L.M., McGrath, P.J., Craig, K.D., Santor, D., Cassidy, K-L. & Reid, G.J. (2001) Facial expression of children receiving immunizations: a principal components analysis of the child facial coding system. Clinical Journal of Pain 17: 178-186.

Buck, R. (1994) Social and emotional functions in facial expression and communication: the readout hypothesis. Biological Psychology 38: 95-115.

Buck, R., Losow, J.I., Murphy, M.M. & Costanzo, P. (1992) Social facilitation and inhibition of emotional expression and communication. Journal of Personality and Social Psychology 63: 962-968.

Bush, J.P. & Harkins, S.W. (1991) Conceptual foundations: pain and child development. In: Children in pain: clinical and research issues from a developmental perspective, eds. J.P. Bush & S.W. Harkins, Springer-Verlag.

Carey, T.S., Hadler, N.M., Gillings, D., Stinnett, S. & Wallsten, T. (1988) Medical disability assessment of the back pain patient for the Social Security Administration: the weighting of presenting clinical features. Journal of Clinical Epidemiology 41: 691-697.

Carroll, J.M. & Russell, J.A. (1996) Do facial expressions signal specific emotions? Judging emotion from the face in context. Journal of Personality and Social Psychology 70: 205-218.

Chambers, C.T. & McGrath, P.J. (1998). Pain measurement in children. In: The management of pain, eds. M.A. Ashburn & L.J. Rice, Churchill Livingstone.

Chambers, C.T., Reid, G.J., Craig, K.D., McGrath, P.J. & Finley, A. (1998) Agreement between child and parent reports of pain. Clinical Journal of Pain 14: 336-342.

Chapman, C.R. (1995) The affective dimension of pain: a model. Advances in Pain Research & Therapy 22: 283-301.

Chapman, C.R. & Nakamura, Y. 1999) Pain and consciousness: a constructivist approach. Pain Forum 8: 113-123

Chibnall, J.T., Tait, R.C. & Ross, L.R. (1997) The effects of medical evidence and pain intensity on medical student judgements of chronic pain patients. Journal of Behavioral Medicine 20: 257-271.

Chibnall, J.T. & Tait, R.C. (1995) Observer perceptions of low back pain: effects of pain report and other contextual factors. Journal of Applied Social Psychology 25: 418-439.

Clipp, E.C. & George, L.K. (1992) Patients with cancer and their spouse caregivers. Cancer 69: 1074-1079.

Cosmides, L. & Tooby, J. (1992) Cognitive adaptations for social exchange. In: The adapted mind, eds. J.H. Barkow, L. Cosmides & J. Tooby, Oxford University Press.

Cosmides, L., & Tooby, J. (1997) Dissecting the computational architecture of social inference mechanisms. In: Characterizing human psychological adaptations, Ciba Foundation Symposium 208, Wiley.

Craig, K.D. (1980) Ontogenetic and cultural influences on the expression of pain in man. In: Pain and Society, eds. H.W. Kosterlitz & L.Y. Terenius, Weinheim Verlag Chemie GmbH.

Craig, K.D. (1999) Emotions and psychobiology. In: Textbook of Pain, 4^th edition, eds. R. Melzack & P.D. Wall, Churchill Livingstone.

Craig, K.D., Grunau, R.V.E. & Branson, S.M. (1988) Age-related aspects of pain: pain in children. In: Proceedings of the Vth World Congress on Pain, eds. R. Dubner, G.F. Gebhart, R.F. Bond, Elsevier Science Publishers.

Craig, K.D., Hyde, S.A. & Patrick, C.J. (1991) Genuine, suppressed and faked facial behavior during exacerbation of chronic low back pain. Pain 46: 161-171.

Craig, K.D., Lilley, C.M. & Gilbert, C.A. (1996) Social barriers to optimal pain management in infants and children. Clinical Journal of Pain 12: 232-242.

Craig, K.D. & Patrick, C.J. (1985) Facial expression during induced pain. Journal of Personality & Social Psychology 48: 1080-1091.

Craig, K.D. & Prkachin, K.M. (1978) Social modeling influences on sensory decision theory and psychophysiological indexes of pain. Journal of Personality & Social Psychology 36: 805-815.

Craig, K.D. Prkachin, K.M. & Grunau R.V.E. (1992) The facial expression of pain. In: Handbook of pain assessment, eds. D.C. Turk & R. Melzack, Guilford Press.

Craig, K.D., Whitfield, M.F., Grunau, R.V.E., Linton, J. & Hadjistavropoulos, H.D. (1993) Pain in the preterm neonate: behavioural and physiological indices. Pain 52: 287-299.

Crawford, C.B. (1989) The theory of evolution: of what value to psychology? Journal of Comparative Psychology 103: 4-22.

Crombez, G., Eccleston, C., Baeyens, F. & Eelen, P. (1996) The disruptive nature of pain: an experimental investigation. Behaviour Research & Therapy 34: 911-918.

Crombez, G., Eccleston, C., Baeyens, F. & Eelen, P. (1998) When somatic information threatens, catastrophic thinking enhances attentional interference. Pain 75: 187-198.

Cronin, H. (1991) The ant and the peacock: altruism and sexual selection from Darwin to today, Cambridge University Press.

Daly, M. & Wilson, M. (1988) Homicide, Aldine de Gruyter.

Daly, M. & Wilson, M. (1991) Anti-science and the pre-Darwinian image of mankind. American Anthropologist 93: 162-165.

Damasio, A.R. (1994) Descartes’ Error: emotion, reason and the human brain. Macmillan Papermac.

Damasio, A. (1999) The feeling of what happens: body, emotion and the making of consciousness. Vintage.

Dar, R., Beach, C.M., Barden, P.L. & Cleeland, C.S. (1992) Cancer pain in the marital system: a study of patients and their spouses. Journal of Pain & Symptom Management 7: 87-93.

Darwin, C. (1872/1965) The expression of the emotions in man and animals. University of Chicago Press.

Dawkins, R. (1976) The selfish gene, Oxford University Press.

Demyttenaere, S., Finley, G.A., Johnston, C.C. & McGrath, P.J. (2001) Pain treatment thresholds in children after major surgery. Clinical Journal of Pain 17: 173-177.

Dimberg, U. & Ohman, A. (1996) Behold the wrath: psycho-physiological responses to facial stimuli. Motivation & Emotion 20: 149-182.

van Dongen, K.A.J., Abu-Saad, H.H. & Hamers, J.P.H. (1999) On the development of an observational scale to measure pain in non-verbal children with severe or profound cognitive impairment; collecting the indicators. Abstracts, 9th World Congress of Pain, IASP Press, p.87.

Eccleston, C. & Crombez, G. (1999) Pain demands attention: a cognitive-affective model of the interruptive function of pain. Psychological Bulletin 125: 356-366.

Eccleston, C., Crombez, G., Aldrich, S. & Stannard, C. (1997) Attention and somatic awareness in chronic pain. Pain 72: 209-215.

Ekman, P. (1989) The argument and evidence about universals in facial expressions of emotion. In: Handbook of Social Psychology, eds. H. Wagner & A. Manstead, John Wiley & Sons.

Ekman, P. (1992) An argument for basic emotions. Cognition & Emotion 6: 169-200.

Ekman, P. (1993) Facial expression and emotion. American Psychologist 48: 384-392.

Ekman, P. (1994) Strong evidence for universals in facial expressions: a reply to Russell's mistaken critique. Psychological Bulletin 113: 268-287.

Ekman, P. & Friesen, W. V. (1978) Facial Action Coding System, Consulting Psychologists Press.

Ekman, P., Friesen, W.V., O'Sullivan, M., Chan, A., Diacoyanni-Tarlatzis, I., Heider, K., Krause, R., LeCompte, W.A., Pitcairn, T., Ricci-Bitti, P.E., Schere, K.R., Tomita, M. & Tzavaras, A. (1987) Universals and cultural differences in the judgments of facial expressions of emotion. Journal of Personality & Social Psychology 53: 712-717.

Ekman, P. & O'Sullivan, M. (1991) Who can catch a liar? American Psychologist 46: 913-920.

Etcoff, NL. & Magee, J.J. (1992) Categorical perception of facial expression. Cognition 44: 227-240.

Fernandez-Dols, J.M. & Ruiz-Belda, M-A. (1997) Spontaneous facial behavior during intense emotional episodes: artistic truth and optical truth. In: The psychology of facial expression, eds. J.A. Russell & J.M. Fernandez-Dols. Cambridge University Press.

Ferrell, B.R., Rhiner, M., Cohen, M.Z. & Grant, M. (1991) Pain as a metaphor for illness, Part I: impact of cancer pain on family caregivers. Oncology Nursing Forum 18: 1303-1309.

Finkenauer, C. & Rimé, B. (1998) Socially shared emotional experiences vs. emotional experiences kept secret: Differential characteristics andn consequences. Journal of Social & Clinical Psychology 17: 295-318.

Finley, G.A., McGrath, P.J., Forward, S.P., McNeill, G. & Fitzgerald, P. (1996) Parents' management of children's pain following 'minor' surgery. Pain 64: 83-87.

Fishbain, D.A., Cutler, R., Rosomoff, H.L. & Rosomoff, R.S. (1998) Chronic pain and disability exaggeration/malingering research: a review. American Pain Society 17^th Annual Scientific Meeting Program Book, p. 144.

Fleckness, P.A. & Molony, V. (1997) Pain and injury. In: Animal welfare, eds. M.C. Appleby & B.O. Hughes, CAB International.

Flor, H., Breitenstein, C., Birbaumer, N. & Furst, M. (1995) A psychophysiological analysis of spouse solicitousness towards pain behaviors, spouse interaction, and pain perception. Behavior Therapy 26: 255-272.

Flor, H., Kerns, R.D. & Turk, D.C. (1987) The role of spouse reinforcement, perceived pain, and activity levels of chronic pain patients. Journal of Psychosomatic Research 31: 251-259.

Fordyce, W.E. (1976) Behavioral methods for chronic pain and illness, C.V. Mosby & Co.

Fordyce, W.E., Fowler, R.S., Lehmann, J.F. & DeLateur, B.J. (1968) Some implications of learning on problems of chronic pain. Journal of Chronic Diseases 21: 179-190.

Fordyce, W.E., McMahon, R., Rainwater, G., Jackins, S., Questad, K., Murphy, T., & De Lateur, B. (1981). Pain complaint-exercise performance relationship in chronic pain. Pain 10: 311-321.

Fordyce, W.E., Roberts, A.H. & Sternbach, R.A. (1985). The behavioural management of chronic pain: a response to critics. Pain 22: 113-125.

Forrest, M., Hermann, G. & Andersen, B. (1989) Assessment of pain: a comparison between patients and doctors. Acta Anaesthesiologica Scandinavica 33: 255-256.

Frank, R.H. (1988) Passions within reason: the strategic role of the emotions. W.W. Norton & Co.

Fridlund, A.J. (1994) Human facial expression: an evolutionary view, Academic Press.

Fridlund, A.J. (1997) The new ethology of human facial expression. In: The psychology of facial expression, eds. J.A. Russell & J.M. Fernandez-Dols, Cambridge University Press.

Fridlund, A.J. & Duchaine, B. (1996) ‘Facial expressions of emotion’ and the delusion of the hermetic self. In: The Emotions: Social, Cultural & Biological Dimensions, eds. R. Harre & W.G. Parrott, Sage.

Frijda, N.H. (1994) Emotions require cognitions, even if simple ones. In: The nature of emotion, eds. P. Ekman & R.J. Davidson, Oxford University Press.

Galin, K.E. & Thorn, B.E. (1993) Unmasking pain: detection of deception in facial expressions. Journal of Social & Clinical Psychology 12: 182-197.

Gilbert, P. (1992) Depression: the evolution of powerlessness, Lawrence Erlbaum Associates.

Goodenough, B., Addicoat, L., Champion, G.D., McInerney, M., Young, B., Juniper, K. & Ziegler, J.B. (1997) Pain in 4- to 6-year-old children receiving intramuscular immunization injections: a comparison of the Faces Pain Scale with other self-report and behavioral measures. Clinical Journal of Pain 13: 60-73.

Goodenough, B., Champion, G.D., Laubreaux, L., Tabah, L. & Kampel, L (1998) Needle pain severity in children: does the relationship between self-report and observed behaviour vary as a function of age? Australian Journal of Psychology 50: 1-9.

Gosselin, P., Kirouac, G. & Dore, F.Y. (1997) Components and recognition of facial expression in the communication of emotion by actors. In: What the face reveals, eds. P. Ekman & E. Rosenberg, Oxford University Press.

Gould, S.J. & Lewontin, R.C. (1979) The spandrels of San Marco and the Panglossian program: a critique of the adaptationist programme. Proceedings of the Royal Society of London 205: 281- 288.

Grossman, S.A., Sheidler, V.R., Swedeen, K., Mucenski, J. & Piantodosi, S. (1991) Correlation of patient and caregiver ratings of cancer pain. Journal of Pain & Symptom Management 6: 53-57.

Grunau, R.V.E. & Craig, K.D. (1987) Pain expression in neonates: facial action and cry. Pain 28: 395-410.

Hadjistavropoulos, H.D. & Craig K.D. (1994) Acute and chronic low back pain: cognitive, affective and behavioral dimensions. Journal of Consulting & Clinical Psychology 62: 341-349.

Hadjistavropoulos, H.D., Craig, K.D., Grunau, R.V.E. & Johnston, C.C. (1994) Judging pain in newborns: facial and cry determinats. Journal of Pediatric Psychology 19: 485-491.

Hadjistavropoulos, H.D., Craig, K.D., Hadjistavropoulos, T. & Poole, G.D. (1996) Subjective judgements of deception in pain expression: accuracy and errors. Pain 65: 251-258.

Hadjistavropoulos, H.D., Hadjistavropoulos, T. & Craig, K.D. (1995) Appearance-based information about coping