Facial expression of pain: "Just So Stories," spandrels, and patient blaming

Facial responses to pain might be the result of evolution but \Villiams' interesting "Just So" story provides no convincing evidence for her hypothesis. Contrary to her hope, casting facial action in an evolutionary perspective will probably not reduce the common practice of health care professionals blaming patients for their problems: instead, it may discourage appropriate treatment. Williams has done a thorough and creative job of assembling and integrating the extant data on facial expression and has shown that these data can be interpreted in evolutionary terms. She ill'gues that such ,m approach could reduce the common and pernicious practice of blaming patients who have chronic pain. However, Williams presents sc,mt evidence that facial expression is the result of evolution and no evidence that patients would be better treated if facial responses were shown to be evolutionmy. Williams alludes to the argument of Gould and Lewontin (1979) who have objected to "Just So Stories" in evolutionary psychology. 'just So Stories" a~ Kipling (1912) originally told them, are stories about characteristics of various animals and how they developed. For example, "How the Whale got his Throat" tells the story of how a shipwrecked sailor when swallowed by a whale created baleen by cutting up his shipwreck and making a grate in the whales' throats so that whales could not eat people. These fanciful stories are wonderfully entertaining, post hoc explanations that have as their justification the joy, not the scientific enlightenment, they bring to the reader. Williams' evolutionary account of facial response to pain is a great story, but I could not fmd in it the evidence that facial responses to pain are of evolutionary Oligin. Williams sets up operant conditioning as the only logical alternative illld shows that operant conditioning cannot explain the emergence of facial expressions to pain. Her dismantling of operant conditioning is well done, but unnecessary. Most people who have seen infants produce facial expressions of pain at one hour after birth, need no convincing that reinforcement is not the major element in the development of facial responses to pain. What is needed to make Williams' ari-,rument convincing is evidence that facial expressions have changed in response to evolutionary pressure. The argument that communication of pain ha~ current survival value is well made by Williams, but is not sufficient to show this is how it developed. Darwin (1873) discussed "con-elations of growth," and, even more eloquently, Gould and Lewontin (1979) pointed out that behaviour might develop as a side effect, ratller th,m as a direct effect, of evolution. Gould and Lewontin (1979) bon-owed tlle architectural tenn "spandrel," which refers to the triangular space created when an arch is placed in a rectangular wall. Spandrels are often the site of elaborate decoration. They point out that spandrels are the by-product of tlle development of arches. That is, tlley were not developed for tllemselves, but came along with the development of arches. Similarly many features that are attributed to evolution are simply the by-product of otller evolutionary changes and not the product of evolution itself. Gould (1997) suggests tllat there are two main ways of determining if a feature is a primary result of evolution or a secondary by-product. The first way is by evidence of an actual histOlical order of events. So, in our case, we would want to know when htcial action fi)r pain arose in the evolution of humans. This is unlikely to be available, a~ tllere is not likely a fossil record ~f pain faces. The second metllOd, which is more likely to be useful in the case of pain facial expression, but is still inferential, is to use current examples of anatomy illld behaviour across species that are more or less close to humans in evolutionary terms. This method could yield an inferred historical record. For this we need extensive cross-species studies that, as Williams notes, are not available at tlus time. My own, personal, anecdotal, observation suggests that many mammals, at least the domestic and lab animals, and many nOllmammals do not have facial expressions for pain. The most interesting data would, however, be from tlle lower and lugher nonhuman primates. So the question remains: Are facial actions evolved, or just an accidental spandrel left over from chance factors? Williams suggests a possible candidate, namely the close proximity of the motor and pain areas of tlle human brain. At tlus time, it is unclear why humans use facial expression ratller tlUII1relying on the more common vocal and other behaviours that other animals use to signal pain. The answer will likely never be answered deHnitively, but additional data might make tlle story more convincing. If facial pain expression were evolved, would it make a difference to the milllagement of pain? Williams seeks an evolutionary explanation for facial pain expression because she believes tllat an evolutionary approach will overcome tlle dominant openmt approach. She suggests tlus approach is too frequently applied and overemphasizes the benefits or gains that patients have from pain, and disregards the costs tllat patients in pain must endure. She hopes to use an evolutionmy approach to reduce tlle preoccupation that m,my clinicians have with malingering, and to combat the ignoring of patient complaints and the under-treatment of pain. I share Williams' concern about doctors blalning patients for their condition, and about tlle preoccupation with unjustified and untllerapeutic search for malingering. An additional patient-blaming tactic is to asclibe psycholOgical causes a~ tlle source of pain. Williams does not e:>,:plainhow an evolutionary model will reduce patient blaming. Health professionals have, from ancient times, blamed patients for not getting better when treatments failed. Prior to the poorly applied, operant model that is currently used to blame patients, we had a poorly applied psychodynamic model that also blamed patients and empha~ized secondary gain from pain. Of course our Freudian colleagues did not originate patient blmning. Prior to Freud, patients were blamed on grounds of moral weakness illld before that, because of supposed possession by evil spiIits. I believe tl1at if facial reaction to pain were shown to be evolved, patients would continue to be blamed. Moreover, because of tlle general but erroneous perception tllat anytlung genetic cannot be modified by lemning, it is likely that pain patients who did not get better would be seen to deserve blame because of tlleir evolutionary inferiority; it is also likely that they would be denied tlle psychosoeial treatments that have been shown to decrease disability a1ld improve quality of life (Williams et al. 1966). A behavior-analytic developmental model is better Gary Novaka and Martha Pelaezb aDepartment of Psychology, California State University, Stanislaus, Turlock, CA 95382; bDepartment of Educational and Psychological Studies, University Park Campus, College of Education, Miami, FL 33199. [email protected] [email protected] Abstract: Behaviorists accept, but go beyond, Williams' notion that there is an evolutionary origin to some unlearned pain behaviors. Abehavior-analytic developmental model is a better fit for explaining the totality of pain behaviors. This model focuses on respondent-operant interactions and viewsmuch pain behavior as "mands" (Le" demands). Behaviorally based explanations from the crying and social referencing literature support this model. Behaviorists accept, but go beyond, Williams' notion that there is an evolutionary origin to some unlearned pain behaviors. Abehavior-analytic developmental model is a better fit for explaining the totality of pain behaviors. This model focuses on respondent-operant interactions and viewsmuch pain behavior as "mands" (Le" demands). Behaviorally based explanations from the crying and social referencing literature support this model. ViItually every behavior analyst would agree that facial expressions of pain probably have an evolutionary Oligi.n. Skinner (1969; 1984) wrote extensively on the phylogenie basis of many hUmail behaviors and supported the notion of "unlearned" or "innate" behaviors in addition to learned ones. The main weakness of Williams' article is her misunderstanding of "behaviOlism," as demonstrated by her followingcontradictory statements: "operant mechanisms act on unconditioned pain behaviors. But where the implicit assumption is of a tabula rasa at biJ.th, openmt mechanisms become the exclusive explanation for pain, invoked with minimal or no evidence" (target article, sect. 2). Unconditioned behaviors (responses) are, bydefmition, unlearned or innate. This contradicts Williams' second statement that behaviorists make the implicit assumption of a tabula rasa. Skinner long ago addressed this misperception of behavioral theory in responding to a similar misunderstanding made by Breland and Breland (1961) when he said: "And it is a myth. No reputable student of animal behavior has ever taken the position 'that the animal comes to the laboratory as a virtual tabula rasa'." (Skinner 1969). While Williams asselts that she only takes exception with the way behavioral theory has been misapplied to the clinical treatment of pain, and accepts other aspects of behavioral theory in general, her call for an evolutionary prewired account of the facial expression of pain makes it clear that she sees little utility in behavioral accounts and treatments of pain behaviors. However, the issues Vvilliamsraises are better addressed by contemporcuy behavior cu1alyticapproaches (Novak 1996; 1998; Gewirtz & Pelaez 20(0). Behavior analysis provides more consistent, parsimonious, cu1dempirical explanations of tlle development of behavior signaling perinand other emotions tllat emerge as a result of ontogenic emdphylogenic histories. The available evidence on some specifIc pain behaviors, including facial vocal e}"1)ressions, is that tlley have a phylogenie source. Asnoted byWilliams, infant facial expressions serve many functions that logically would have adaptive advantages. Primary among these functions is to signal to caretakers the existence of aversive stimuli. Animals (including humans) are born witll many reflexes tllat have an apparent functional advcultage for the species. These behavior-environment relationships are seen, in behavioral theory, to be the result of environmental contingencies tl1athave worked, distally, in tl1ehistOly of the species. In behavioral terms, we have an unconditioned response (e.g., griJ.11ace) to emunconclitioned stimulus (e.g., pin prick, stomachache). Altl10ugh, as Williams points out, there is no consensus as to what constitutes a facial expression of pain, several of its phenotypic traits appecu'to be universal. In the history of the species, the function of pain facial expressions as of otl1erpain behaviors would evolve if tl1eyprovided defensive functions, such as signaling the presence of hann-producing stimuli (includiJ.1ginternal stimuli like pain). Tllis is particularly functional for species with young, including humans witl1 infants, who have limited repertoires of self~defensivebehaviors, and survival reflexes that disappear quickly. Fixed responses to fixedstimuli, however, have linlited utility over the course of a lifetime, particularly among species having long lifespans and inhabiting varied species-typical environments. These conditions give selection advantages to individuals in species who could also change responses to stimuli, or learn. While pain reflexes to unconditioned pain stimuli remain potent across the humcu1lifespan, both respondent and operant learning are relevant to tl1eshapiJ.1g of pain behaviors. In respondent conditioning, a neutral stimulus (e.g., lifting tl1e cl1ild'sleg) occurring in close temporal proximity to the unconditioned perinstimulus (e.g., heel-stick) acquires a new function of bemg a conditioned stimulus that elicits pain behaviors (Goubet et al. 2(01). Thus, eventually, just seeing a needle can make an infimt cry. Opercu1t lecuning is the focus of most behavior-analytic approaches, as it is iJ.1Fordyce's operant model of pain behavior, which is at the center of Williams' criticisms. From a behavioraldevelopmental perspective, operant-respondent mteractions cu'e involved (B~io.u& Baer 1961). Not only are pain stimuli unconditioned stimuli f<lrreflexive pain behaviors, they are also primary universal negative reinforcers. That is, because of phylogenic contingencies, all members of the species are more likely to engage m behaviors that terminate, remove, eliminate, or avoid these aversivepainful stimuli. Some of these negatively reinfclrcingconsequences may be automatic, aswhen moving an injured limb relieves the pain stimulus. Other consequences may consist of socially remforcing events, as when crying causes a parent to move a cl1ild'sinjured limb to a more comfortable position. In addition to the negative remforcement resultmg from ternlinating the perin stimulus, positive reinforcement for pain reporting behaviors may occur, such a~when the mother kisses the scraped knee of tl1CClYing child. Behavior analysts (Hayes & Hayes 1992; Michael 1984; Novak 1996) apply Skinner's analysis of verbal behavior to these situations where the behavior is maintained by the mediation of a speaker who understculds the function of the speaker's behavior. Therefore, the opercu1tfunction of much pain behavior is that of a "mculd,"or dememdfor removing the pain stimulus. AsWilliams suggests, both speaker and listener must be involved, but from a behavior-analytic stand, tl1ese behaviors are learned (Gewiltz & PeIaez-Nogueras 1992a). Note that behaviors might have both respondent and openmt components contemporaneously. The development of Clymgbehavior, which ha~ the advantage of bringing assistance from a greater distemce them facial expression, follows this pattern of transition from respondent to operant crymg. ACIYof pain is among tl1euniversal distinctive cries of human neonates. Although pain crying is distinct, these differences are small (Fuller 1991), and parents (especiallymothers) quickly]e,ml to identify and respond to their own infants' pain cries (Wiesenfeld et al. 1981), The origillal respondent crying may be shaped into operant behavior reinforced by either positive or negative reinforcement. Operant crying may grow to be so problematic tl1at it becomes tl1e focus of clmical intervention (Etzel & Ge\virtz 1967; Gewirtz & Boyd 1977; Hart et al. 1964). The effects of intennittent positive, social reinforcement by parents with regard to crying have been demonstrated in laboratory studies tllat explain the formation of infant attachment patterns to tl1eirmothers, emd the conditioniJ.1gof separation protests (GewiJ.tz & PelaezNoguenlS 1991). LikeWillicuns,Campos (1983)has postulated that the responses and perceptions comprismg social referencing are "prewired" (i.e., unconditioned). As an alternative to tl1at nativistic theory, C,ewirtz and Pelaez-Nogueras (1992b; Pelaez-Nogueras 1992) demonstrated that iJ.1fantsocial referencing results from the infant's contingency-based learnmg. That is, m contexts of uncertainty,maternal expressive facial cues of joy and fear come reliably to predict positive or aversive consequences for tl1emfant's operant (reachmg) responses. Using a conditioning-reversal (ABAB) design with eighteen 4to .5-month-old infants who showed no social referencing, Pelaez-Nogueras (1992) demonstrated that maternal emotional facial expressions can become conditioned cues for infant referencing. Initially, during pretreatment/ba~eline a~sessment, no difference existed in tl1eiJ.1cidenceof mfants reaching for ambiguous objects followmgeither maternal joyful or fearfUl facial expressions. However, in the next phase, the infants learned to reach for ambiguous objects when reaclling was cued by a joyful maternal facial expression and followed by extlinsic positive reinforciJ.lgconsequences, and to avoid those ambiguous objects when reaching was cued by a fearful maternal expression and followed by extIinsic aversive consequence contingent on their reachiJ.lg.In tlle tl1ird phase of tlle experiment tIllSdifferential reaclling pattern in tl1epresence of tlw two facial-emotional expressions were eAtinguished. Finally, in tl1elast phase, the cues recovered tl1eirpredictive power when contingent reiJ.lforcement was remtroduced. These results supported tl1e hypothesis that maternal facial emotional expressions selve ilSconditioned cues for mfant social referencing and their reaching or avoiding responses in ambiguous contexts. In the same manner, infants learn to use their mother's facial expressions a~Signals.One can easily observe tI1esocial referencing of toddlers who fall, look to their mother's face for a cue as to whether to get up and go on, or cry. In turn, the toddlers' facial expressions of pain are likely to serve as signals for the parents. These behaviors are subject to reinforcement by parents who typically relieve pain, and whose behavior is concurrently shaped by the child's termination of the grimaces and crying. Therefore, some long-term pain behaviors may be the result of direct pain stimuli, but they may also be the result of an operant history of reinforcing pain behaviors, as initially fonnulated by Fordyce. In sum, the goal of behavior analysis is to identifY the functional relationships between pain behaviors and their maintaining contingencies, both distal phylogenic and proximal ontogenic contingencies. We believe that the existing, well-established principles of behavior analysis are sufficient to account for the wide range of pain behavior phenomena. Compared with the evolutiomllY account proposed by Williams in the target article, we believe that behavior analysis provides both a more parsimonious account of the source of pain behaviors, and a more efficacious program for their clinical treatment. Jaak Pankseppa and Marcia Smith Pasqualinib aOepartment of Psychology, Bowling Green State University, Bowling Green, OH 43403; bDepartment of Psychology, Centre for the StUdy of Emotions, University of Portsmouth, Portsmouth POt 20Y, United Kingdom. [email protected] [email protected] Abstract: No adequate evidence exists f(x the evolution of facial pain expression and detection mechanisms, as opposed to social-learning processes. Although brain affective/emotional processes, and resulting whole body action patterns, have surely evolved, we should also aspire to monitor human sufIering by direct neural measures rather than by more indirect indiees. No adequate evidence exists f(x the evolution of facial pain expression and detection mechanisms, as opposed to social-learning processes. Although brain affective/emotional processes, and resulting whole body action patterns, have surely evolved, we should also aspire to monitor human sufIering by direct neural measures rather than by more indirect indiees. There is probably no bigger issue for human and animal welfare than understanding and controlling pain and sufferillg. Amanda Williams has shared a wide-ranging evolutionary perspective on the sources of facial expressions that commonly accompany pain, and the associated psychological-communicative processes that, she suggests, may have evolutionary underpinning. It is an admirable call fi)r more and better research, rather than a defining summary in a field where substantial evidence remains meager. The practical issue is to develop external measures ofinternally experienced suffering, independent of pervasive social desirability ,md second,uy gain issues that may affect outward emotional expressions. Williams £:wors the view that facial expressions function in a social-communicative role in addition to being an unconditioned consequence of the experience of suffering. Her approach leads to one clear prediction: Facial expressions of pain will be more evident in supportive social circumstances than in nonsupportive ones. Indeed, since Kleck et al. (1976) have found fewer facial expressions of pain in tlle company of others, \Villiams suspects that those eflects are due to the presence of strangers, who might be prone to take advantage of the situation, rather than the presence of friends, who would be more likely to help. Williams suggests tllere are evolved communicative tendencies that regulate such eJl.1Jressions;hence, the issue of detecting lying ,md deceit becomes essential fiJr her analysis. We leave aside tlle conspicuous dilemma tlmt tlle modulation of facial expressions by social context, including possible deliberate manipulations, still presents many problems for using facial expressions as a "gold standard" for evaluating the intensity and duration of affective experience, at least in adults. Instead, we will focus on critical evolutionary questions tllat Williams' perspective brings to tlle fore: Is there evidence to choose between evolution and le~rning as explanations for the regulation and detection of pain eJl.-pressions?We think not. And even if certain such processes have been subject to some type of overall evolutiomuy pressures, we still have a long way to go to conclude that individual facial and perceptual components have been tlle objects of selection. Even as we respect Williams' struggles with such distinctions, we. believe that most findings in evolutionalY psychology, including those summarized by Williams, can at present be seen as largely reflecting leaming-based processes based perhaps on the reinforcing consequences of evolved affective processes, of which there are bound to be mally. Should we have any confidence in the assumption that "the pain face" evolved? Although Williams does not take a strong position on this, her extensive coverage of reliable pain action units could, without further clarification, be seen as an implicit acceptallCe that the expressions have undergone natural selection. Even though there may be substantial evolutionary underpinnings for many other facial expressions, such as smiling, which clearly facilitates certain social communications, one could easily suggest that the facial glimacing that accomp,mies acute pain may be part ,md parcel of a whole body defensive response (global flexor contraction) which lTIayspill over into the facial musculature. From this perspective, pain may have highly predictable facial action units, without necessitating the conclusion that such expressions evolved in some ancestral deep time where the survival adValltage of this or that face was penetrating into the genome. \Villiams herself asserts that evolutionary "selection operates at the level of function, not at the level of physical structures or behaviours that subserve the function" (target article, sect. 3). Perhaps for this reason, she focuses more of her effi)rts on the socio-cognitive rules that regulate facial displays, than on those that generate tlle displays. We think it is quite reasonable to postulate that the whole body affect-linked expressions of pain have in fact evolved. This seems evident from the flailing and wailing of all infant in acute pain. Such responses al'e seen even in inf,mt rats, whose facial expressions of pain would surely be lost on potential c,rretakers (albeit not their crying). Might not the facial accompaniments in creatures possessing the necessalY facial muscles be evolutionary byproducts of the global distress response, ones that can easily be molded by social learning, but by-products nonetheless? Therefore, wIllie we accept that tlle global affective-distress response is surely evolved and full of internally experienced affective "meaning" for many animals, we would hesitate to accept the evolution otspecific facial action units and related cognitive behaviors until simpler, more p,mimonious affect-based le,mling interpretations have been evaluated and excluded. In short, where evolution has generated global action patterns within the brain, the components of tlle patterns need not have been objects of selection with singular representations in the genome. Rather, they may be stable features of a general "instinctual" response tendency. vVe suspect there are many such broad-scale heuristic principles in mind/brain evolution, with the global responses generated by basic emotional systems comprising prime examples (Panksepp 1998). This is where modem evolutionary psychology may have already gone massively astray (P,mksepp & P,mksepp 2000). Although there are certainly various evolved emotional systems shared, in principle, by all mammals, the cognitive manifestations of tllOse systems in their pervasive interactions \Vitll neocortical tissues tlmt mediate general-purpose learning abilities may have few evolutionarily built-in strategies, except for conditional strategies to minimize distress and to maxim.ize pleasure. The increasingly popuhrr intellectual assumptions of mainstremn evolutionary psycholot,'Y are likely to remain in the realm of ideas alld modest statistical trends, rather than ofbiological subshUlce, for a long time to come. Most plausibility arguments in the area remain more heavily conceptual than empirical. To take one final pain example consider the phenomenon of limping. All vertebrates, when they have an injured leg, exllibit a limping gait that presumably has little, if any, intrinsic social communicative value. Either tlle limping reflects tlle effects of internally experienced pain OIl the central motor apparatus, or simply damaged peripheral structural supports. Very little about limping