facial efference and the experience of emotion

7/30/2019 Facial Efference and the Experience of Emotion

1/32

Ann, Rev. Psychol. 1989. 40.249-80Copyright1989by Annual eviewsnc. All rights reserved

FACIAL EFFERENCE AND THEEXPERIENCE OF EMOTIONPamela K. Adelmann and R. B. ZajoncDepartmentf Psychology nd he Research enter or Group ynamics, heUniver-sity of Michigan, Ann Arbor, Michigan 48109

CONTENTSINTRODUCTION.................................................................................... 249

Pre-DarwinianensoryTheoriesof Emotion............................................... 250Evolutionaryheory f Emotion.............................................................. 251Developmentf the FacialFeedbackHypothesis .......................................... 251INVESTIGATINGACIALEEDBACK...................................................... 257CorrelationalStudies of Facial Efferenceand EmotionalExperience................. 258ExperimentalStudies of Facial Efference and Emotional Experience:the Modulatingunction.......................................................... 264Experimental Studies of Facial Efference and Emotional Experience:the InitiatingFunction............................................................. 269SUMMARYF THE LITERATURE N FACIALFEEDBACK......................... 272CONTEMPORARYHEORETICALSSUES: WHATS FED BACK? ................ 274CONCLUSIONS...................................................................................... 276

INTRODUCTIONIn this review, we consider the nature and role of facial expression inemotionalprocesses. We xaminehe recent theoretical and empirical itera-ture for its bearingon the questionsof the proximal nddistal correlates offacial emotional ctions, particularlyon the questionof their modulatingndinitiating functions in the experience of emotion. We mphasizehe role ofemotional acial action in the subjective experienceof emotion.We void n this paper he convention f referring to emotional acial actionas "expression" ince that term imposes an a priori theory, implying that

2490066-4308/89/0201-0249502.00

www.annualreviews.org/aronlineAnnual Reviews
http://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronline


2/32

250 ADELMANN& ZAJONCcmotional acial action (facial efference) has as its major role the manifesta-tion of internal states. This as the primary role of facial action has not beenestablished as an empirical fact, and thus it is best for now o employ termthat does not prejudge the outcomeof empirical and theoretical analysis.While the term "facial efference" is less agile and less common,t has theadvantageof being neutral with respect to the kind of principles that wouldbecalled upon o explain it.Before examining he empirical evidence on facial efference and its corre-lates, we briefly review he major heoretical perspectives on facial efference:sensory, evolutionary, and facial feedback.Pre-Darwinian Sensory Theories of EmotionTwophysiologists working at the second half of the 19th century, TheodorPiderit (1858, 1888) and Pierre Gratiolet (1865), based the explanationfacial emotional action on the sensory system. According o both theories,facial movementsre generalizations of peripheral muscular actions elicitedin the course of the sensory and perceptual process. Thus, the facial actionaccompanyinghe emotion of disgust is similar to that occurring when heindividual reacts to an unsavory aste. Both theories point out that peripheralmovements an also be elicited by imagination; for example, ocular accom-modation and convergenceshowdifferent patterns whenone thinks of thread-ing a needle than whenone imagines a ship on the horizon. Gratiolet (1865)noted the close affinity betweensensation and sentiment, where the formerdealt with stimuli that camefrom the exterior whereas the latter with thosethat came rom the interior. Sentiment, or sens intime is an experience thatoriginates within the organism, and as such it constitutes the basic element ofemotion. He also distinguished symbolic and metaphoric movements,wherethe former is illustrated by a bowlers movementollowing his bowling balland the latter by the gesture of contempt hat is a metaphor or a reaction to anunpleasant odor.Gratiolet held that "no sensation, image, or thought.., can occur withoutevokinga correlated sentiment which ranslates itself directly . . . into allspheres of external organs... " (1865, p. 65). More nterestingly, however,he insisted that the converse is equally true. "The movements nd bodilyattitudes," he wrote (p. 66), "even if they arise from fortuitous causes, evokecorrelated sentiments, which in turn influence imagination, feeling, andthought."Gratiolet did not refer to Piderit, whose heory was strikingly similar,althoughPiderit (1888) claims to have presented his ideas in Pads in 1859atmeeting of the Biological Society, of which Gratiolet was a member ndwhichpublishes the Gazette M~dicale,which printed Piderits presentation inissue #46.



3/32

FACIAL EFFERENCE 251Evolutionary Theory of EmotionDarwinsviews on the communicative nd adaptive function of facial effer-ence, being well known, need not be described here. It is worth noting,however, hat he rejected both Piderits and Gratiolets writings. He says ofthe latter, "AlthoughGratiolet emphatically denies that any muscle has beendeveloped olely for the sake of expression, he seemsnever to have reflectedon the principle of evolution" (1896, p. 11). And o Piderit, who ent his bookto Darwin, the latter wrote:

I have copy ndknowf yourwork nMimic,tc. which have ound eryusefulandoftenquote.But am poorGermancholar ndyour tyle ... I find verydifficult tounderstand. ccordinglyemployedmano translate or me everal ages.These havegiven n myntroduction,n order o state, as far as possible ya few entences, ourviews. fear hatI mayotdoyou ull justice,butassuredlytriedmy est o do t (Piderit1888,pp.7-8).

In reference to this letter, Piderit says that even if Darwinhad knownGermanbetter he would not have paid more attention to the sensory theory ofemotional expression because, as in other work, Darwinwas only interestedin discovering newevidence for his theory of evolution ["Darwinne chercheici, comme ans tous ses autres travaux, quh d6couvrir de nouveauxdocu-ments en faveur de sa th6orie de l6volution" (1888, pp. 7-8).] Pideritsperception of Darwins motives seems accurate, as Darwin concludes hisbookby asserting that emotionalexpression is indeed a further demonstrationof how "man is derived from some lower animal form" (1896, p. 365).Despite Darwins preeminent nterest in promoting evolutionary theory inhis work on emotion, his admissionof a possible causal role of efference inthe emotional experience foreshadowed he developmentof the facial feed-back hypothesis. "The free expression by outward signs of an emotionintensifies it," he wrote. "On he other hand, the repression, as far as this ispossible, of all outward signs softens our emotions" (1896, p. 365).Development of the Facial Feedback HypothesisJAMESIANHEORYhe facial feedback theory of emotional efference de-rives in part from William James, who in proposing his famous theory ofemotion n 1884 ntroduced the possibility of a causal role of the face in theexperience of emotion. His often quoted statement that "the bodily changesfollow directly the perception of the exciting fact, and that our feeling of thesame changes as they occur is the emotion" (p. 13, 1922) was frequentlyconstrued later by critics to include only, or primarily, visceral changesasfeedback. Yet in his original formulation of the feedback heory he named otonly visceral but respiratory, cutaneous, and circulatory alterations. In addi-tion, he wrote, "what is really equally prominent, but less likely to be



4/32

252 ADELMANN ZAJONCadmitteduntil special attention is drawn o the fact, is the continuous oopera-tion of the voluntary muscles n our emotional states. Even whenno change ofoutwardattitude is produced, their inward tension alters to suit each varyingmood,and is felt as a difference of tone or of strain" (p. 15). Even n hissomewhatevised statement in The Principles of Psychology 1890), the roleof muscles in the experience of emotion was not discounted as thoroughly ashis critics contended(e.g. Cannon 927, 1931). He continued to refer to the"indefinitely numerous" nd "the immense umber"of bodily reverberationscorresponding to each emotion, including among hem visce(al, muscular,and cutaneous effects.As for the facial musculature,althoughhe did not specifically distinguish itfrom the skeletal musculature as a source of feedback in emotional experi-ence, virtually every example James employed o illustrate his hypothesisincluded somereference to facial efference. "Smooth he brow, brighten theeye, contract the dorsal rather than the ventral aspect of the frame, and speakin a major key, pass the genial compliment, and your heart must be frigidindeed if it does not gradually thaw!" he wrote (p. 1078). And, he asked,"Canone fancy the state of rage and picture no ebullition in the chest, noflushing of the face, no dilatation of the nostrils, no clenching of the teeth . ." (p. 1067-68)?Even o, in his most specific reference to facial efference in The Principles,James admitted that its influence in the generation of feeling was over-shadowedby the corresponding visceral and organic componentsof emotion.Anticipating the criticism that among ctors and others with extensive practicein posing facial emotional actions, voluntary efference does not alwaysproduce subjective feeling, he argued that these highly trained individualsmayhave learned to suppress the "natural association" betweenefference andthe visceral and organic components f emotion, on the latter of which"it isprobable that the chief part of the felt emotiondepends" (p. 1080). For lesspracticed individuals, he seemed o believe that voluntarily effecting the"so-called manifestations" of an emotion ought to give rise to that emotion.He also admitted, "Wemaycatch the trick with the voluntary muscles, butfail with the skin, glands, heart, and other viscera. Just as an artificiallyinitiated sneeze lack~ ~omething f the reality, so the attempt to imitate anemotion in the absence of its normal instigating cause is apt to be ratherhollow" (p. 1066)~Although amesascribed to the skeletal musculature lesser role in initiat-ing emotion han other organs, he nevertheless clearly included this source offeedback as an integral component f his theory. In 1890 as in his earlierformulation, he proposed our steps in the generation of subjective experienceof emotion: a sensory stimulus (of either external or internal origin)transmitted to the cortex and perceived; reflex impulses ravel to muscle, skin,



5/32

FACIAL EFFERENCE 253and viscera; the resulting alterations in these targets are transmitted viaafferent pathwaysback to the brain; these return impulses are then corticallyperceived, and whencombinedwith the original stimulus perception, producethe "object-emotionally-felt."Sherrington (1900) and Cannon(1915, 1927, 1931), however, seizedJamess admission that felt emotion mayrely more heavily on visceral andorganic than other components s the basis of their attacks on his theory.Drawingon animal research, they built a convincing argument that visceralfeedback was an inadequate determinant of emotion. Sherrington (1900) citedhis research on animals in which the autonomic nervous system afferentpathways rom the shoulders downwere destroyed, thus eliminating feedbackfrom the viscera. When resented with an emotion stimulus, the dogs reactedwith all appearance of a normal "emotional psychosis," including facialefference, head and foreleg movements, nd vocalizations. His recommenda-tion was therefore to "accept visceral and organic sensations and the memo-des and associations of them as contributory to primitive emotion, but wemust egard themas reenforcing rather than initiating the psychosis"(p. 258).Cannon 1927) developed a five-point attack on the Jamesian theory. LikeSherrington, he named he animal research showing an unimpaired emotionalreaction in the absence of visceral feedback (Cannon1915). He also arguedthat visceral changes--heart ate acceleration, inhibition of digestive activity,sweating, and others---occur uniformly across a variety of emotional statesand are too diffuse to discriminate among hem. The viscera are relativelyinsensitive structures, he continued, and changes in them are muchslowerthan the average atent period of an affective reaction. Finally, he argued thatartificial induction by adrenalin of emotion-like visceral changes does notproduce the subjective experience of an emotion unless a mood s alreadypresent.The attacks by Sherrington and Cannonbrought a stream of response frompeople such as Angell (1916), Floyd Allport (1924), and Perry (1926),rising to Jamess defense offered theoretical interpretations placing moreemphasis on the role of the skeletal musculature in emotion. Angell, forexample, declared that James "nowhere et himself the task of attempting todifferentiate emotions on exactly the basis suggested by Dr. Cannons tate-ment" (1916, p. 259). James would have agreed, he contended, that in someemotions identical patterns of visceral excitement might occur, but mighthave then argued that "their distinction from one another in such cases maybefound n extra-visceral conditions, and particularly in the tonus of the skeletalmuscles" (p. 260). He further argued, of Sherringtons head-and-shoulderdogs, that "no evidence which left facial and cranial muscles unimpairedwould ever have seemed to him very convincing as ground for conclusionsunfavorable to his theory" (p. 261). Perry (1926), in a similar vein, responded



6/32

254 ADELMANN ZAJONCto Cannon hat distinctions among motions "may ie in the proprioceptiverather than in their interoceptivepatterns; that is to say, in the motor et ratherthan in the visceral reverberation" (pp. 300-1).

Allport (1924) argued that although the autonomicnervous system maynotdiscriminate among iscrete emotions, it does differentiate the class of posi-tive emotions from the negative. He assigned to the cranio-sacral divisionresponsibility for the "conscious quality of pleasantness," and to the sym-pathetic division the "visceral responses whichare represented in conscious-ness as unpleasant" (p. 90). Further, Allport proposed that within a singleaffective class "the differentiating factor arises from the stimulation of theproprioceptors in the muscles, tendons, and joints of the somatic part of theorganism; and that afferent impulses from these somatic patterns of responseadd to the autonomiccore of affectivity the characteristic sensory complexesby which one emotion is distinguished from another" (pp. 91-92). To thisadded that "the facial expressions as well as bodily movements re stronglydifferential" (p. 92).Indeed, Cannonhad a more difficult time in dismissing the muscularcomponent of Jamess theory. He claimed simply that "sensations whichunderlie the appreciation of posture are entirely lacking feeling-tone" (1927,p. 119). He acknowledged hat motor attitude seems to influence subjectiveexperience, but argued that instead of providing sensory feedback, certainpostures remove the usual motor cortex inhibition of the thalamus, thestructure at the center of his theory of emotion, thus facilitating subjectiveexperience.NeitherJamesscritics nor his defendersdefined a specific role of the facialmuscles n emotion, a definition that was not to appear for several decades.But in the interim, Nina Bull proposed her Attitude Theory of Emotion Bull]951; Pasquarelli & Bull 1951), which revived and extended the muscularaspect of Jamesian theory.TIlE ATTITUDEHEORYn the Attitude Theory, Bull argued that confusionabout whetherbodily changeor subjective experience comes irst was due to afailure to separate emotional efference into its component arts. James wasmistaken, she explained, only in that he focused on the action componentrather than on the preparatory motor attitude: "We eel angry as a result ofreadiness to strike, and feel afraid as a result of readiness tO run away,and notbecause of actually hitting out or running, as James explained the sequence"(1951, p. 6).

Bull postulated that the involuntary postural attitudes preparatory to actionare accompanied y appropriate organic changes, and that "feelings of theseorganic changescombinewith the feelings of the orienting posture itself--andwith someawarenessof the original exciting stimulus--to produce the famil-



7/32

FACIAL EFFERENCE 255iar experience knownas an emotion" (p. 5). She further suggested thatfeeling "may ollow and accompany motor attitude, but does not necessarilydo so; and cannot possibly precede it---cannot in fact appear at all without anantecedent motor attitude to fire the afferent pathways rom the muscles andviscera to the brain" (p. 19).To test the Attitude Theory, Pasquarelli & Bull (1951) first inducedsubjects an emotion and its corresponding motor attitude through hypnosis,using such directions as these (for anger): "Your hands are getting tense andyour arms are getting tense. Youcan feel your jaw tightening." They then"locked" this attitude, and suggested the feeling, but not the attitude, of acontrasting emotion. In some rials an unpleasant emotion (disgust, anger,fear, or depression) was locked and a pleasant one (joy, triumph) suggestedand on the others the reverse sequence was followed.Withoutexception, subjects reported they could not successfully "feel" thesuggested emotion while locked in the contrasting attitude. Those w~hochanged heir feeling to the newemotion could do so only by disobeying thesuggestion prohibiting changes n efference or organic sensation.Like those whoprecededher, Bull did not assign a special role to the face,although the hypnoticsuggestions to subjects included instructions for facialas well as bodily motorattitude. But within the next decade, emotion heoristsbegan o postulate a specific and central role of the facial musculature n theexperience of emotion. These postulates, embeddedwithin comprehensivetheories of emotion, came o be known s the facial feedback hypotheses.FACIAL EEDBACKYPOTHESESirst among these theorists was SilvanTomkins, who in a wide-ranging two-volume set of books (1962, 1963)introduced the notion of facial feedback. "Most contemporary nvestigatorshave pursued the inner bodily responses, after the James-Lange heory fo-cused attention on their significance," he explained. "Important as theseundoubtedly are, we regard them as of secondary importance to the expres-sion of emotion hrough he face .... the face expresses affect, both to others,and to the self, via feedback, which s more rapid and morecomplex han anystimulation of which the slower movingvisceral organs are capable" (1962,pp. 205-6). As providing feedback from the face he listed the tongue andfacial muscles, the sound of ones ownvoice in the ears, and changes inblood-flowand temperature of the face. He futher regarded the facial musclesas more nformative of affective state than those of the trunk and extremities,thereby making clear statement about the face apart from the other skeletalmuscles.Tomkins uilt his case for the primacyof the face uponseveral arguments.First, he argued the face is the most sensitive and dominantpart of the body,with a high density of "neural representation and firing" (1962, p. 208).



8/32

256 ADELMANN& ZAJONCnoted that in contrast to other responses, involuntary facial responses arehighly resistant to habituation. In addition, the facial muscles ack a fascialcover that binds muscleselsewhere n the body ogether into groups, so that inthe face, "smaller muscle portions or even single musclebundles maycontractindependently of the rest of the muscle" (p. 225) in a variety of complexpatterns.In the Tomkinseedback cycle, a stimulus activates an innate, subcortical"affect program," which emits messages through the motor and circulatorypathways o the entire body. The responses of the affected motor and glandu-lar targets--the face primarily, other sites secondarily--supply sensory feed-back to the brain, which, if it reaches consciousness, is subjectively ex-perienced as emotion. Tomkins rgued that this feedback may be acted uponwhether or not it reaches awareness, that voluntary facial efference maynotaccurately duplicate the innate pattern, and that the sequencemaybe initiatedby retrieved conscious affect or central imagery as well as by an externalemotion stimulus.He later modified his theory somewhat o downplay the facial muscles["Muscles appear to be specialized for action and not for affect" (1980, p.149)] and focused instead on the facial skin as playing the greatest role inproducing eelings of affect. Hespecifically argued hat in an expressive face,receptors normallyhidden n the skin changeposition in response o the facialmuscle patterns; the feedback is therefore from these cutaneous receptorsrather than from the muscles of the face.Both muscles and skin played important roles in Gellhorns (1964) viewfacial feedback, and both bodily and facial muscles were assigned importantroles. Gellhorn argued that body posture influences affective arousal throughthe proprioceptive discharges feeding back to alter hypothalamicbalance. Butbecause such diverse states as happiness and tenseness are both associatedwith increased proprioceptive postural discharges, while low muscle toneaccompaniessadness as well as "postprandial happiness," he considered theproprioceptive feedback provided by bodily posture to be insufficient todistinguish among iscrete emotions. For this differentiating information heturned to the face."The great density of the cutaneous receptors in the face and the consider-able variety of the patterns of contraction of the facial muscles uggest hat theresulting patterns of neocortical excitation and hypothalamic-cortical dis-charges will match in diversity that of the emotional expression," he wrote(1964, p. 465). He argued that the proprioceptive facial-muscle dischargesarouse the hypothalamic-cortical system, while tactile impulses are conveyedto sensorimotor cortex. In combination, these two sources of feedback "playan important role in the development f the emotionsand a subsidiary role intheir reinforcement after they have been established," he claimed (p. 468).



9/32

FACIALEFFERENCE 257Izard voiced a similar view in his Differential EmotionsTheory (1971).contended hat the primary components f emotion are neural activity, striatemuscleor facial-postural activity, and subjective experience, augmented y

the brain stem reticular system and the glandular-visceral system. In hissequenceof events, a stimulus perception activates central neural activity (inan unknownorder to the brain stem, hypothalamus, and limbic cortex),producing a global pleasant or unpleasant feeling. The hypothalamus ignalsthe smoothand striate muscles(perhaps in a discrete emotion-specificpatternto the face). The specificity of the facial musclefeedback o the brain stem,hypothalamus, limbic system, thalamus, and possibly cortex determines thespecificity of the felt emotion. Feedback rom the auxiliary systems (includ-ing visceral, glandular, cardiovascular, and respiratory) helps sustain andamplify he subjective experience. He added hat for a specific facial patternto match subjective experience, it must correspond to the original efferentpattern, the neural messagemust travel innate pathways or the emotion, andthe feedback must be reasonably complete (e.g. slight or micromomentaryefference mayproduce only fleeting awareness).In sum, although their ideas varied somewhatn the particulars, these threetheorists in quick successionproposed specific and central role of the face inthe experience of emotion. Their thoughts inspired a wealth of empiricalresearch on the facial feedback hypothesis.INVESTIGATING FACIAL FEEDBACKIn the course of empirical investigation of facial feedback,several versions ofthe hypothesis have evolved. The most basic, drawndirectly from the theoriesof Tomkins,Gellhorn, and Izard, is considered first. It proposes simply thatin the process of a naturally occurring emotionalexperience, there will be acorrespondencebetweenfa~:ial efference and subjective experience. Implicitin this version are the notions that strength of efference and intensity ofsubjective experience covary, and that specific efference patterns correspondwith specific subjective states.Although n the original facial feedback hypotheses autonomicarousal wasassigned only an auxiliary role, at most amplifyingor sustaining an emotionalexperience, others have pursued the Jamesian notion that facial efferencepatterns correspond with patterns of autonomicarousal as well. Thus, re-search has also addressed whether facial efference and arousal covary inintensity and whether rousal patterns are differential, at the least, for positiveversus negative emotional efference patterns, and, at most, for a variety ofdiscrete facial efference patterns. Thecorrelational literature on facial effer-ence, physiological arousal, and subjective experience addresses these ques-tions arising from the original facial feedback hypotheses.



10/32

258 ADELMANN ZAJONCCorrelational Studies of Facial Efference and EmotionalExperienceIn the correlational literature on facial feedback, external stimuli such asfilms, electric shock, and slides are most ypically used as elicitors of facialemotional efference, but some researchers have had success using othertechniques such as imagery and reinforcement. In general, only physiologicalor subjective correlates of facial efference have been measured, but in a fewstudies both types of information have been collected from the samesubjects.PHYSIOLOGICALXPERIENCEhe literature on emotional efference andphysiological arousal preceding the facial feedbackhypotheses suggested thatany correlation between acial expressivity and arousal wouldbe negative. Inmost of this research (Prideaux 1920; Landis 1932; Jones 1948; Block 1957;Learmonth et al 1959), subjects who were most expressive showed littleautonomic arousal, and those least expressive showed he most autonomicactivity.

As a logical extension of this early research, initial studies on facialefference and physiological arousal adopted the same between-subjects meth-odology. Lanzetta & Kleck (1970), for example, were among he firstexamine he association between acial efference and arousal in the context ofa study on encodingand decoding ability. In their study, subjects undergoingshock trials were unknowingly ideotaped. Their degree of facial expressive-ness was determinedby the ability of a set of judges to accurately discriminateshock from nonshock rials from the subjects faces. Subjects with the highestgalvanic skin responses (GSR), indicating sympathetic autonomic nervoussystem activity, were the least facially expressive, paralleling the inverserelationship betweenemotional expressivity and GSReported by the earlierresearchers.Notarius & Levenson (1979), on the other hand, found no significantassociation between GSR esponses and expressiveness of response to thethreat of shock. But consistent with the general pattern, less facially ex-pressive subjects showed reater heart rate and respiration rate responses tothe threat than did more expressive subjects.

Bucket al (1972), in the context of research on nonverbal communicationaccuracy, looked at GSR nd heart rate correlates of facial efference elicitedby sexual, scenic, maternal, disgusting, unusual, and ambiguous slides.Expressivity of subjects was measured by howwell observers could guess,from subjects facial efference, the category and pleasantness of slides beingviewed. Expressive subjects showed lower GSR han less expressive ones,but heart rates did not differ. Buckand colleagues (Bucket al 1974) replicatedthis study using male-female pairs of senders and observers as well assame-sex pairs. They again found lower GSRamongmore expressive sub-



11/32

FACIAL EFFERENCE 259jects (this time for menonly) and no differences in heart rate in between-subjects comparisons. More ecently, Buck(1977) found a negative associa-tion between ndicators of expressivity and skin conductance n both male andfemale preschoolers in response to slides.Whyhe degree of an individuals emotional expressivity, facial or other-wise, should inversely vary with autonomicarousal, has been explained in anumberof ways. The basic discharge or cathartic-hydraulic view is thatemotion must find an exit, and if it cannot be vented outwardly throughefference, it must be routed inward. Jones (1948) labeled individuals whooutwardlydisplay emotion but show ittle arousal "externalizers;" those whobehaviorally manifest little emotionbut show ubstantial autonomicactivityhe called "internalizers." Lanzetta & Kleck (1970) proposed hat individualswhoare socialized to inhibit their outward emotional displays evidenceincreased arousal from the combination of redirected emotion and conflictexperienced over competing tendencies to express and to inhibit emotion.Bucket al (1974) suggested that during the emotional socialization of suchindividuals, the stress of parental rebukes becomes ssociated with emotion-eliciting situations, and that the arousal they showmayarise from thisassociation rather than from evoked emotion. Buck also later suggested(1977) that innate determinants mayplay a role in whetheran individual isinternalizer or extemalizer.What s important for the facial feedback hypothesis, however, s that theinverse relationship foundbetween ubjects does not rule out the possibility ofa positive correlation betweenexpressivity and arousal within subjects, aswouldbe predicted by the hypothesis. In fact, in nearly all within-subjectsanalyses, that is the association that has been found.Vaughan& Lanzetta (1980) recorded both autonomic arousal and facialexpressivity as judgedby activity in musclesaround the eyes and jaw (indicat-ing pain) of subjects watchinga modelsreactions to shocks. Onshock rials,subjects showedboth greater facial activity (resembling pain) and increasedGSR,compared to nonshock trials.Contradictory results have been reported in two studies with infants. Brocket al (1986) found that 3-month-old nfants smiles in reaction to a femaleexperimenter corresponded with an increase in heart rate. But Cohenet al(1986) found in a study of 4-month-oldnfants facial efference in responsetheir mothers facial poses that infants heart rate increased during angerefference but decreased when their faces showed interest and joy. Skintemperature was highest during angry faces and lower during joy.

Dimberg 1982) used photographs of emotional faces as stimuli and mea-sured both heart rate and skin conductance responses. Happy face photo-graphs produced increased zygomatic activity and angry photographs in-creased corrugator activity, indicating that subjects adopted happyand angry



12/32

260 ADELMANN& ZAJONCfacial patterns, respectively. Heart rate and skin conductancedropped n bothpatterns of emotional efference (and did not vary by emotion).In the studies by Buck nd colleaguesnoted earlier, greater facial activity inresponse to the stimulus slides corresponded with higher GSR eactivitywithin subjects (Bucket al 1972, 1974). In addition, GSRncreased with theunpleasantnessof an individuals facial reaction to slides (Buck1977; Buck tal 1972, 1974), as did heart rate (Buck et al 1972, 1974).The within-subjects research on physiological correlates of spontaneousfacial efference is mixed. Indicators such as GSR nd heart rate do not seemto consistently discriminate positive from negative facial affect, but they doappear to generally increase with intensity of facial efference.SUBJECTIVEXPERIENCEore directly pertinent to facial feedback hy-potheses is whethersubjective experience corresponds with facial efference.In contrast to the research on physiological correlates, greater facial ex-pressivity is uniformly associated with greater subjective experience bothbetween and within subjects, and the reported emotion tends to match theparticular facial efference pattern in both valence and category.In two between-subjects studies, Cupchik & Leventhal reported associa-tions between facial expressive behavior and evaluations of cartoons inresearch on audience and sex effects. In one (1974), the intensity of spontane-ous smiling and laughing at cartoons (manipulated by playing canned laugh-ter) was correlated with higher funniness ratings by women ut only forpoor-quality cartoons for men. These results were replicated in a second study(Leventhal & Cupchik 1975) in which type of canned laughter was varied.Kleinke & Walton 1982) manipulated subjects spontaneous facial effer-ence of happiness using reinforcement techniques. Subjects were not awarethat this was influencing the amount hey smiled, nor could they accuratelyguess howoften they smiled during the study. Those whowere reinforced tosmile frequently reported more positive feelings and rated the interview andinterviewer higher than did those not reinforced.Ekman t al (1980) examinedboth positive and negative affect in a studyusing films as stimuli. Facial efference patterns were coded using their FacialAction Coding System (FACS). Subjects who showed a happiness facialaction while watching positive films reported themselves as happier thanthose whodid not show t, and the frequency, duration, and intensity of thisaction were positively correlated with self-reported happiness. On he otherhand, subjects whoshowed acial actions of anger, fear, disgust, sadness, orcontempt o a negative film reported more negative affect than those whodidnot, and frequency and duration of expressed negative emotion were corre-lated with greater self-reported negative affect. In particular, expressed andfelt disgust were significantly related.



13/32

FACIAL EFFERENCE 261In several within-subjects analyses, facial efference patterns as assessed byelectromyography (EMG) orresponded with self-reported mood. TeasdaleBancroft (1977) compared corrugator EMG nd depressed mood n a smallsample of depressed subjects during happy and unhappy thoughts. The un-happy imagery increased both depressed mood nd corrugator activity (nega-tive facial efference) compared o happy thoughts, and depressed moodandcorrugator activity were highly correlated. McHugo1983) found that activityof the zygomatic muscles (happy facial efference) while watching positivefilms was correlated with positive self-reports, while corrugator activity

during a negative film was linked to higher self-reported anger. In a thirdstudy (Cacioppo et al 1986), although observers could not discriminate thevalence and intensity of subjects facial displays to slides, EMGctivityreliably did so; corrugator activity (anger) was higher for negative thanpositive self-reported affect, and the more he corrugator activity, the morenegative the affect. Similarly, zygomatic esponses correspondedwith greaterpleasant than unpleasant affect. In two experiments, Dimberg 1987b) alsofound that subjects corrugator and zygomatic ctivity in reaction to picturedstimuli corresponded o negative and positive self-reports, respectively.Schwartz and colleagues have published a number of studies on facialefference generated by imagery-induced affect and assessed with EMG.n a1976 study, depressed and normal subjects imagining happy, sad, and angrysituations generally showed he correspondingfacial efference patterns andsubjective reports for each emotion, although the magnitudeof each responsediffered for depressed and normal subjects. In a normal student sample,Brown& Schwartz (1980) found increased self-reported happiness and happyfacial patterns (zygomatic)following standardized happy magery, and great-er reported sadness and sad facial patterns (corrugator) following sadnessimagery. Anger and corrugator activity, and fear and zygomatic activityfollowed the corresponding magery or anger and fear. Differing intensitiesof imagery evoked parallel intensity of both facial activity and subjectiveexperience. In another sample, self-reported happiness and zygomatic ctivityboth increased with happy imagery, and self-reported sadness and corrugatoractivity resulted from sadness imagery (Schwartzet al 1980). Similar resultswere reported using self-referent statements to generate affect (Sirota et al1987).An magery paradigmwas also used by Sutherland et al (1984) in a sampleof black women. magining fearful and racially derogatory scenes increasedcorrugator activity and reduced pleasantness ratings compared o imagining aneutral scene.Only one report departed from this consistent positive association betweenfacial efference and subjective experience. Kleinke & Walton 1982) founda within-subjects analysis of their data (for both subjects whowere einforced



14/32

262 ADELMANN& ZAJONCfor smiling and nonreinforced subjects) that amount of smiling was notsignificantly correlated with self-reported positive feelings.PHYSIOLOGICAL AND SUBJECTIVE EXPERIENCE Finally, in several studiesboth physiological and self-report correlates of facial efference were assessedin the samesubjects. Three of these involved between-subjects ests. In thefirst, researchers divided their sample into groups varying in rated facialexpressivity (Notarius et al 1982). In reaction to an angry scolding byexperimenter, minimally expressive subjects showed significant heart rateincrease compared o both nonexpressive and highly expressive subjects, andreported significantly higher guilt feelings (but did not differ on nine otheraffect scales). In the second, Winton et al (1984) found no significantcorrelations of facial expressivity with heart rate or skin conductance nbetween-subjects analysis. Ridgeway& Waters (1987) found that childrenasked to think about exciting experiences showedmore facial pleasure andheart rate variability than children in a calm-imagery group, and a sad-imagerygroup showed ess facial pleasure and slightly less heart rate changethan in the calm group.Whenhe relationship betweenfacial efference and its correlates is ex-amined within subjects, the picture again changes. Kleck and colleagues(1976) measured ubjective and autonomic esponses in a study on the impactof an observer on expressiveness. The presence of another reduced theexpressive response to shock, and both subjective and autonomic responsesdecreased as well.Subjects in a study by Dimberg 1987a) showed reater corrugator activityand skin conductance responses and more rated unpleasantness to a high-intensity tone than to one of low intensity. Heart rate decelerated to thelow-but not the high-intensity tone.As part of a larger study, McHugot al (1985) showed ubjects taped silenttelevision segments of Ronald Reaganexpressing happiness, fear, or anger.Subjects imitated the facial patterns, with elevated zygomatic ctivity duringhappy segments, elevated corrugator and reduced zygomatic activity duringanger segments, and both moderately raised corrugator and moderately lo-wered zygomatic activity during the fear segments. Corresponding to theirhappy acial imitations, subjects self-reported joy and warmthwere high andskin conductance was low. For anger, reported negative affect and GSRresponses were high. When roducing fearful faces, both positive and nega-tive self-reported affect were moderatelyhigh, and the samewas true of GSR.Heart rate dropped or all three facial patterns, but more o for the negative

ones.Winton t al (1984), using slides as stimuli, found hat heart rate increased



15/32

FACIAL EFFERENCE 263linearly with pleasantnessof subjective report and facial efference, while skinconductance ncreased with the intensity of facial efference and self-report,regardless of valence.Finally, in a 1981study, Schwartz and colleagues used the imagery tech-nique to examine he blood pressure and heart rate correlates of happiness,sadness, anger, and fear in comparison o relaxed and control trials. Fear andhappiness both were linked with increased heart rate and systolic bloodpressure; but in fear, heart rate increased more. Sadness and happinessproducedsimilar heart rate and systolic blood pressure levels, but sadnessproduced lower diastolic blood pressure. Anger, like fear, was associatedwith high heart rate and systolic blood pressure, but it produced higherdiastolic blood pressure. Self-reported affect corresponded ppropriately witheach pattern of facial efference.IMPLICATIONS FOR THE FEEDBACK HYPOTHESIS From the correlationalliterature it is clear that in between-subjectsests, facial expressiveness snegatively correlated with autonomicarousal. Morecrucial for the feedbackhypothesis, however, are within-subjects comparisons. These show con-vincingly that increased facial expressiveness of emotion is correlated withincreased physiological arousal, as the hypothesis predicts. Generally, itappears that facial efference of either positive or negative affect is associatedwith autonomicarousal, although there is someevidence that increased GSRis linked specifically to negative facial affect (Buck1977; Bucket al 1972,1974; Dimberg 1987a; Kleck et al 1976; McHugo t al 1985; Vaughan& Lanzetta 1980) and that increased heart rate is specific to pleasantnessof the facial efference pattern (Brock et al 1986; McHugot al 1985; Ridge-way & Waters 1987; Wintonet al 1984), although the reverse has also beennoted for heart rate (Buck et al 1972, 1974; Cohenet al 1986; Schwartz etal 1981).

Asfor subjective experience, specific affective self-reports increase almostwithout exception with facial efference patterns of the corresponding emo-tion. This is not surprising given that in most research paradigms theaccompanying motional stimulus could easily inform subjects of the appro-priate feeling even if facial efference did not. But in studies wherespontaneous acial efference was intensified without subject awareness(i.e.by using reinforcement, canned laughter, or the presen~ce of an observer),self-reported affect increased correspondinglyabove he level reported to theemotionalstimulus alone. These results suggest the possibility not only of acorrespondence, but of a modulating influence of facial efference in theexperience of emotion, a hypothesis explored more thoroughly in researchusing experimentally manipulated facial efference.



16/32

264 ADELMANN& ZAJONCExperimental Studies of Facial Efference and EmotionalExperience: the Modulating FunctionThe possibility that facial efference plays a causal role in emotion has longbeen a subject of theoretical speculation (Gratiolet 1865; Darwin1896; James1890, 1922). Although he original facial feedback hypotheses restrict con-sideration of this role to the context of ongoing, spontaneous emotion,research in which facial emotional action is experimentally manipulated hasmushroomedn the past decade. Part of the reason is undoubtedly he inabilityof research using spontaneousefference to separate correlation from causal-ity. Interest in the possibility of modulating, nd, perhaps, initiating effects offacial efference is such that the term "facial feedback hypothesis" is nowcommonly efined in these terms (Buck 1980; Laird 1984; Matsumoto 987).

Mostwork examining he role of voluntary facial efference focuses on itsmodulating function in emotional experience. Typically, an emotional stimu-lus (in the form of an external stimulus, such as a film, or an internal one,such as imagery) is introduced, and the effects on emotional experience ofexaggerating or inhibiting a congruent facial efference pattern or simulatingan incongruent pattern are examined.PHYSIOLOGICALXPERIENCEn the literature relating facial efference toother components f emotional experience, correlational studies have focusedlargely on the physiological correlates while experimentalstudies have tendedto focus on subjective experience. True to this split, only a few experimentalstudies on facial efference examineonly physiological correlates.In the first, Colbyet al (1977) required subjects to facially pose high,moderate, and no pain on different shock trials. They ound that as intensityof posed pain to a shock increased, so did subjects GSResponses, but in theabsence of shock, posed pain did not affect skin conductance. In the secondstudy, a variant of their 1980 correlational study using spontaneous ex-pressions, Vaughan & Lanzetta (1981) measured autonomic responsessubjects watchinga modelsdisplay of pain in response to electric shock. Thesubjects were instructed either to inhibit their own acial patterns, to displaypain when he modeldid so, or were given no instructions. The display groupshowed ncreased skin conductance and heart rate compared o the other twogroups, which did not differ from each other.Two ecent studies used imagery to generate emotions. Ianni et al (1986)comparedvoluntary facial anger, disgust, neutral, and control conditionsduring anger imagery. Facial anger augmentedcardiovascular activity com-pared to the other facial actions, particularly reducing finger blood flow; butskin conductance and heart rate were not significantly affected. Lanzetta,Kleck, and colleagues (Kappas et al 1987) reported that adding matching



17/32

FACIALEFFERENCE 265overt facial action to self-generated emotions (happiness, sadness, anger,peacefulness) increased heart rate but did not change skin conductance.SUBJECTIVEXPERIENCEn contrast to the physiological data, a numberofstudies have focused on the effects of suppressing, exaggerating, or dis-simulating facial efference on subjective experience during an emotionalstimulus. One of the earliest was the Leventhal & Macebetween-subjectsexperiment (1970) in which the evaluations of a comedy film by schoolchildren asked to smile and laugh as muchas they could were comparedwiththose of children asked not to laugh. Amongirls, positive ratings of the film.were higher in the exaggeration than the inhibition group, but the oppositewas true for boys, although they laughed more.

Rhodewalt & Comer (1979) manipulated smiling, frowning, or neutralpatterns using Lairds (1974) technique while subjects wrote a counterattitu-dinal essay. Subjects moodwas most positive in the smile and least positivein the frown conditions, with neutral in between. Moodwas most negative inthe frowning group, least negative in the smiling group, and in betweenforthe neutral group. Attitude changewasalso greatest in the smiling and least inthe frowning groups.Laird and colleagues have madeseveral within-subjects investigations offacial efference and mood. Laird (1974) first induced subjects to displaysmiles and angry frowns by manipulating muscle contraction patterns, andthen exposed them to pictures of children and of the Ku KluxKlan. Aggres-sion scores were higher on the frowning than smiling trials, and scores onelation and surgency scales were higher on smiling than frowning trials.(Anxiety, remorse, and social affection were not differentially affected byefference.) The difference between self-reported moodon smiling versusfrowning trials was significantly larger for manipulated han for nonmanipu-lated observer subjects, suggesting an effect of efference above he effects ofthe slides. In a second experiment, subjects rated cartoons as funnier when nthe smiling than in the frowningcondition; elation was also higher (but notsignificantly) and aggression lower while smiling than when rowning. Laird& Crosby (1974) reported one successful and one unsuccessful replicationthis work using humorous artoons as stimuli.Again using a technique similar to Lairdss, McArthur t al (1980) inducedsmiles, sad frowns, and neutral patterns while subjects viewed positive,negative, or neutral slides. Self-reported moodof normal-weight but notoverweight subjects was less happy when frowning but not significantlyhappier when smiling compared o neutral across stimuli. In a replicationreported in the same article the same pattern was found.Rutledge & Hupka 1985) also used Lairds technique to pose joy and anger



18/32

266 ADELMANN& ZAJONCin subjects viewingneutral slides and high and low anger and joy slides. In awithin-subjects comparison, subjects reported feeling more joyous and lessangry when posing joy, and more angry and less joyous when in an angerpose. The difference held across all stimulus intensities. The esearchers alsocomparedhe self-reported affect of posed subjects with that of observers whohad undergone the same experimental session but were instructed to keepfacial musclesrelaxed. The subjects whenposing joy reported greater feelingsof joy than their paired counterparts across stimuli, and whenposing angerreported greater anger.Kraut (1982) varied the experimental paradigm by using pleasant andunpleasant odors as affective stimuli in a within-subjects experiment. Posingapleasant reaction increased the evaluated pleasantness of the odors whileposing disgust decreased it compared o the ratings madewhile spontaneouslyreacting to the odors.Three experimentshave tested for modulating nfluences of facial efferenceduring self-generated emotion. McCanne& Anderson (1987) asked subjectsfirst to imagine a positive or negative situation, then to imagine the scenewhile enhancingeither zygomaticor corrugator tension. In a third trial, theywere instructed to suppress tension in that muscleduring imagery. The musclechanges were verified by EMG ut were otherwise not readily detectable.Suppressed zygomatic activity during positive affective scenes decreasedself-reported enjoymentand increased distress; no other effects of muscleactivity were significant. Klions and Dale (Riccelli et al 1984; Antila et al1988) used self-referent statements to induce mood nd posed facial actionsmonitoredby EMGn two studies. In the first, positive and negative emotion-al facial actions (zygomaticand corrugator) increased self-reported elationand depression, respectively, compared o the reports of nonmanipulatedsubjects reading the statements. In the second, facial actions incongruentwiththe statements reduced the contrasting self-reported mood omparedo reportsof unmanipulated ubjects, but the differences were not significant.Rutledge and colleagues (1987) reported effects on felt emotion fromisolated facial musclecontractions while viewingslides used in their earlierstudy. Across all slides, contraction of forehead muscles (occipitofrontalis)was associated with greater reported surprise; contraction of the corrugator(brow)was linked with feelings of anger. Triangularis contraction (pullingmouth down) and zygomaticus contraction (creating a smile) were linkedrespectively with sadness and joy. This study was unusual in movingbeyonda positive/negative (dimensional) distinction by linking specific formsfacial emotional efference with reported changes in discrete (categorical)emotions.Strack and Martin (Strack et al 1988; L. L. Martin, T. F. Harlow, F.Strack, in preparation) have also tested dimensional versus categorical ef-



19/32

FACIAL EFFERENCE 267fects of facial efference using a technique requiring subjects to hold a pen intheir mouths n different ways.In two initial experiments Strack et al 1988),subjects held the pen either in their teeth (simulatinga smile) or in their lips(simulating a frown). Their ratings of cartoon funniness were higher duringthe "smile" and lower during ther "frown" patterns compared o ratings in acontrol condition. Using the pen technique to simulate smiling, angry, andfrowning efference (L. L. Martin, T. F. Harlow, F. Strack, in preparation),they first found evidence only for a dimensional perspective: "smiling"subjects reacted morepositively to stories than did "frowning" or "angry"-faced subjects; but whenstories were changedso that only angry (not sad)evaluations were appropriate, angry poses produced more angry evaluationsbut sad facial efference did not. The researchers also found that addingphysiological arousal that could not readily be attributed to exercise increasednegative and reducedpositive ratings of stories for anger-posingsubjects andproduced he reverse effects for smile-posing subjects.The data in this section are largely consistent in showing hat voluntaryfacial efference, whetherproduced y direct request to express or inhibit or bymore subtle muscle manipulation, changes subjective experience. Exaggera-tion of facial efference congruentwith an emotional timulus increases corre-sponding subjective experience of an emotion while inhibition reduces it.Simulating efference incongruent to a stimulus also reduces stimulus-consistent subjective experience. In the only study looking at more han twoemotions, the results suggest that manipulationof emotion-specific efferenceincreases the corresponding emotion-specific subjective experience.PHYSIOLOGICAL AND SUBJECTIVE EXPERIENCE TWOearly studies dealtwith the modulating effects of facial pose on subjective and physiologicalreactions to electric shock. In a between-subjects experiment by Kopel &Arkowitz 1974) subjects role-played either a calmor upset reaction or did notrole-play during shock. Subjects posing an upset reaction reported feelingmore pain and had a lower pain threshold than the neutral controls, whilethose posing calm reported less pain and a higher threshold. The calm facialpattern looked no different to observers than did the control, although theupset pattern did. Pulse rate wasnot affected by the role-playing. Lanzetta etal (1976) asked subjects to conceal or clearly reveal their reactions to shocksof varying intensit3~. Independent udges confirmed hat facial displays forhiding and exaggerating trials differed in the appropriate directions frombaseline patterns. Comparedo the baseline (previously collected spontaneousreactions), hiding a facial response to shock reduced both GSRand thesubjective report of shock painfulness; exaggerating the facial reaction in-creased both GSR nd self-reported pain.Subjects also suppressed or exaggerated the congruent patterns for



20/32

268 ADELMANN& ZAJONCpleasant, unpleasant, and neutral film segmentsor reacted spontaneously inanother study (Zuckerman et al 1981). Finger blood volume, skin con-ductance, and heart rate were measured, and subjects made self-reports.These researchers reported greatest autonomicarousal across stimuli in theexaggeration group, followed by the spontaneous expressers, and the lowestarousal among he suppressors. The differences were greater in groups view-ing affect-laden films than in those viewingneutral films. Films were ated asmore pleasant for pleasant scenes and more unpleasant for unpleasant scenesby the exaggerators than by the suppressors or spontaneous eactors, althoughthe differences were not significant. Reported intensity of reactions washigher for the affective films when eactions were exaggerated. Higher levelsof facial expressivity were associated with greater autonomic arousal andsubjective experience within as well as betweensubjects.

In two studies, the effects of spontaneous eactions were comparedo onlyone other posing condition. In the first, Ochsmann& Henrich (1984) foundthat female subjects (but not males) rated accident pictures as more righten-ing and showed ncreased physiological arousal when hey exaggerated theirreactions. (In men, self-reported frightfulness did not differ, and physiolog-ical response was lower when efference was amplified.) In the second,self-reports and autonomic responses corresponding with spontaneouslyoccurring facial efference to pleasant, neutral, and unpleasant slides werecompared to responses of subjects asked to display an incongruent pose(Putnamet al 1982). Self-reported pleasantness corresponded o pleasantnessof facial efference in the unmanipulated ondition, but decreased in the posedcondition. Skin conductance was higher for affective than neutral slidesamong he spontaneous reactors, while heart rate accelerated more forpleasant than unpleasant slides. Posing incongruent efference increased skinconductanceresponses but not heart rate.Tourangeau & Ellsworth (1979) compared the effects of congruent andincongruent posed faces, spontaneous efference, and nonemotional poses onself-reports and physiological responses to fearful, sad, and neutral stimulusfilms. Whenwatching a neutral film, subjects posing fear reported highersubjective fear and subjects posing sadness reported greater sadness than didsubjects in other face conditions, but the differences were not significant. Forfearful and sad films, a congruent acial pose did not increase self-reports ofthe matching emotion above the levels reported by the unmanipulated andnonemotional-faced groups, and incongruent poses did not decrease self-reports of the filmed emotion. On the physiological measures, both fear-posing and sorrow-posingsubjects showed reater drops in heart rate than didthe nonemotional-faced group across all films, but less decrease than thespontaneous expressers. The spontaneous expressers showed he biggest GSRresponse, the fear-posing subjects the least.



21/32

FACIAL EFFERENCE 269IMPLICATIONSORTHE MODULATINGYPOTHESISlthough the originalfacial feedback hypothesis wouldnot have considered voluntarily modulatedfacial efference to be a theoretical equal to spontaneouslyproduced fference(Tomkins1962, 1981; Izard 1971, 1981), many esearchers have neverthelessrelied on manipulatedefference in an attempt to clarify whether acial effer-ence can play a modulating ole in ongoingemotional experience. In general,in the subjective experience of emotion, it may;with only a few nonsignifi-cant results (Laird & Crosby 1974; McCanne& Anderson 1987; Tourangeau& Ellsworth 1979) and one contradictory effect (Putnam et al 1982),literature suggests that intensification of a congruent acial pattern enhancessubjective experience, while an inhibited congruent pose or the pose of anincongruent emotion reduces subjective experience.Exaggeration of facial efference generally increases GSR nd inhibitionreduces it, regardless of valence of efference. Only four attempts haveaddressedheart rate effects, with three revealing nondifferentiating increasesfor any efference pattern (Tourangeau& Ellsworth 1979; Vaughan& Lanzet-ta 1981; Zuckerman et al 1981) and one showing no significant change(Putnam et al 1982).Experimental Studies of Facial Efference and EmotionalExperience: the Initiating FunctionA final and more difficult question to answer s whetherfacial efference caninitiate emotional experience in the absence of an emotional stimulus. Someevidence relevant to this issue can be gleaned from conditions included instudies mentionedabove, and in recent years a few attempts have been madeto focus specifically on this question. In these studies, subjects are sometimesasked simply to pose a particular emotional face, are trained to contractparticular muscles or combinationsof muscles, or perform tasks that involve.the facial muscles in ways approximating emotional postures.PHYSIOLOGICALND UBJECTIVEXPERIENCEvidence of the initiatingpotential of facial efference can be found in those studies mentionedabovewhereone of the conditions involved posed efference during a neutral stimu-lus. (For these studies, unless otherwise noted, only trends can be ascertainedbecause significance levels were not reported.)Colbyet al (1977) foundno significant effects of posing pain facial patternson GSR n the absence of actual shock. When resented with neutral pictures,subjects in McArthur t als study (1980) reported slightly higher happinesswhen smiling than with a neutral face, and lower happiness when rowning.In the Rutledge & Hupka study (1985), in a neutral-stimulus condition,subjects posing oy reported slightly higher joy (but also slightly higher anger)than subjects with a neutral pose; whenposing anger they reported more elt



22/32

270 ADELMANN& ZAJONCanger and less felt joy than neutral posers. Fear posers reported more ear, andsadness posers more sadness, during a neutral film in the Tourangeau &Ellsworth study (1979), but the differences were not statistically significant.And, in response to a neutral slide in the Rutledge et al (1987) research,subjects contracting forehead muscles (resembling surprise) reported signifi-cantly more surprise than unposedsubjects, and those posing joy (mouth up)and fear (mouth down) eported significantly more oy and fear, respectively,than unposed subjects.Other studies have focusedspecifically on the potential initiating functionof facial posture. McCault al (1982) asked subjects on different trials eitherto portray fear, to portray calm, or to show heir usual face. In a firstexperiment, subjects showedhigher pulse rates and skin conductance whenthey were portraying fear than for either the calm or normal expressions.Self-reported anxiety was not affected, a result that led the researchers toconclude that physiological changes were due to effort rather than change inemotion. To test this, in a second experiment they added trials in whichhappiness was portrayed. They also added noise as a situational manipulationof negative affect. Subjects rated the noise as less loud while portrayinghappiness (self-reported emotion was not directly assessed in this study).Pulse rate increased under both happiness and fear portrayals, but skinconductance did not change. The researchers believed the greater movementinvolved in these expressions as compared o calm or normal expressionsaccounted for pulse changes.In two studies, Duncan& Laird manipulated acial posture to differentiateself-produced cue users from situational cue users for other purposes. In thecontext of a first study on self-attribution and attitude change(1977), theyinduced muscle poses of smiles, frowns, and neutral faces. Manipulatedsmiles corresponded o higher and frowns to lower elation and surgency thanneutral (relaxed) efference in the presence of a neutral stimulus. A frowningface producedhigher aggression scores than a neutral one, but smiling did notlower aggression comparedo a neutral expression. In a later study on placeboeffects (1980), the researchers successfully replicated this effect using slidesof geometric forms as stimuli and using a group manipulation. Elation-surgencywas significantly higher on smiling than frowning rials, and aggres-sion was higher for frowning than for smiling trials; both smiles and frownsproducedmood cores significantly different from those of a neutral facialpattern.Four studies on biofeedback raining as asthma reatment also fit within theconsideration of facial efference and physiological consequences(GlausKotses 1983; Harver & Kotses 1984; Kotses & Glaus 1981; Miller & Kotses1987). This research involves relaxation or tensing of the frontalis muscle nthe forehead, with dependentvariables including respiratory expiration flow



23/32

FACIAL EFFERENCE 271rate, respiration rate, and heart rate. Relaxation of the frontalis muscleproducedan increase in peak expiration flow rate and tension decreased flowrate; no changes n respiration rate or heart rate have been found. In compari-son, forearm muscle tension or relaxation had no effect on any measure.Ekmant al (1983) used facial poses of six different emotions nd collectedfive physiological measures. Facial surprise, disgust, sadness, anger, fear,and happiness were manipulated using both directed facial action (a musclecontraction technique similar to those used by others) and relived emotion(resembling imagery techniques). In the posed efference condition, threesubgroups of emotions were discriminated based on physiological measures:poses of happiness, disgust, and surprise wereassociated with low heart rate;fear and sadness were linked with high heart rate and low skin temperature;and anger was accompaniedby high heart rate and high skin temperature.Forearm muscle tension did not vary. Of importance to the facial feedbackhypothesis is the finding that autonomicchanges were more clear-cut in thefacial action condition than in the relived emotion imagery) task.Smith et al (1986) also used both posed and imagery-induced efference(happy, sad, angry, and neutral) and measured skin conductance and tieartrate. As determined by EMG agnitudes, stronger efference patterns wereproducedby voluntarily posing, but the patterns were similar to the imagery-induced patterns. They split their sample into expressive and nonexpressiveposers, The nonexpressives produced similar levels of facial activity whenhappiness was posed and when it was induced by imagery, but less activitywhen posing the negative emotions of sadness and anger compared tospontaneousefference. Skin conductancedid not significantly differ by emo-tion when osed. Heart rate increased over the neutral condition for all threeemotions; there was no difference in heart rate increase betweenexpressivesand nonexpressives, even though the latter showed ess facial activity fornegative emotions than did expressives.

In two recent studies, subjects performed asks involving facial muscles nways that approximateemotionalfacial efference patterns, and thereby allowthe experimenter to examine the effects of facial efference alone, in-dependently of its emotional content. Strack et al (1988) using the pen-holding technique, noted evidenceof an effect of facial posture on self-reportseven in the absence of their cartoon stimuli. AndZajonc and colleagues (R.B. Zajonc, S. Murphy,M. Inglehart, submitted), in the context of research onthe vascular theory of emotional efference (Zajonc 1985), compared thesubjective experience of subjects pronouncing or listening to variousphonemes, omeof which involve the action of muscles that are dominant inemotional expressions. For example, the production of the phoneme resem-bles the smile. Photographerselicit smiles from their subjects by requiringthem o say "cheese". The German honemei, on the other hand, has just the



24/32

272 ADELMANN& ZAJONCopposite action. Repeatedpronunciation of e resulted in positive subjectivereports as measured y ratings of liking, pleasantness, and preferences for thesound, whereas /i was judged unpleasant and was disliked, not only byAmerican but by German ubjects as well.IMPLICATIONS FOR THE INITIATING FUNCTION The evidence in this sec-tion suggests that facial efference mayplay an emotion-specific nitiating roleas well as a modulating role in the subjective experience of emotion. Somesignificant results support this conclusion (Duncan & Laird 1977, 1980;Rutledge et al 1987; Strack et al 1988; R. B. Zajonc, S. Murphy, M.Inglehart, submitted), and trends apparent in the nonsignificant results are atleast consistent with the hypothesis (McArthuret al 1980; Rutledge & Hupka1985; Tourangeau & Ellsworth 1979; but not McCaulet al 1982).As in most of the research on facial efference and physiological arousal,little autonomicdifferentiation by the particular form of efference has beenapparent. Either no significant effects on arousal are producedby efferencealone or it increases uniformly across different emotions (Colby et al 1977;Glaus & Kotses 1983; Smith et al 1986). The exceptions are the Ekman t al(1983) study, and the Zajonc-Murphy-lnglehart esearch. In the first, enoughdifferent physiological indexes were used to detect three differentiating pat-terns: one for happiness, disgust, and surprise, characterized by low heartrate; one for fear and sadness, with high heart rate and low skin temperature;and one for anger, with high heart rate and high skin temperature. In thesecond, temperature of the forehead showed lear distinction betweenpositiveand negative affect, cooling being associated with pleasant states and warm-ing with unpleasant hedonic states. Thus, temperature--unlike GSR,heartrate, and similar autonomic indicators that do not discriminate hedonicpolarity---offers a new, important, physiological index of emotion.

SUMMARY OF THE LITERATURE ON FACIALFEEDBACKWehave divided our discussion of the literature of facial feedback into twoclasses: that which examines the correlates of facial efference under con-ditions that elicit it spontaneously, and that in which the modulating andinitiating functions of facial ef~ference can be examinedhrough experimentalmanipulation.

In the correlational literature, we drew, as Buck 1980) did in his review,distinction betweenwithin- and between-subjects ests of association betweenfacial efference and emotional experience. Weargued that the former aremore pertinent to the original facial feedback hypothesis as variously pro-



25/32

FACIAL EFFERENCE 273posed by Tomkins 1962), Gellhorn, (1964), and Izard (1971), and thatgenerally support the hypothesis. Not only does intensity of facial efferencecorrespondwith greater subjective emotional experience in general, but par-ticular facial efference patterns are positively correlated with subjective expe-rience of the same emotions. Physiological arousal does not appear to varydifferentially with the nature of the efference pattern but seems o increasewith any increase in emotional efference.More nterest has been displayed recently in the modulating nd initiatingpotentials of facial efference, explored through experimentalmanipulation ofthe face. The term "facial feedback hypothesis" has come o be defined bythese functions, although the original proponents of the hypothesis disavowthis usage (Tomkins1981; Izard 1981). Although the experimental evidenceis less unanimoushan the correlational data, it appears o us that the literaturesupports these versions of the hypothesis, perhaps more convincingly forsubjective experience than for physiological arousal.It should be noted that in going from spontaneous efference to posing anemotion, contracting muscles in an emotion-like face, or performing otherfacial motor tasks, two things are likely to decrease simultaneously: theinferences subjects can consciously make bout their feelings from the situa-tion and from what their faces are doing, and the closeness of the facialefference to a spontaneous display. The first is an advantage in that itminimizesself-perception interpretations (Laird 1974) and has the strongestcausal implications, while the second poses a clear disadvantage n attemptingto generalize to naturally occurring facial emotionalefference. Some videncesuggests that voluntarily posed efference patterns and spontaneous ones areinnervated through different pathways (Monrad-Krohn1924, 1939). How-ever, compared o spontaneousefference, posed faces tend to be quite similarboth in appearance (Borod et al 1986a, b) and EMGatterns for each emotion(Schwartz et al 1979) and thus, perhaps, in their feedback patterns to thebrain. Oneof the main differences betweenspontaneous and posed efferenceseems to be the greater asymmetry f the latter (Ekman t al 1981).To the extent that voluntarily produced efference can be assumed toadequately correspond to spontaneous emotional efference, the experimentalevidence suggests that facial efference mayplay an important causal role inthe subjective experienceof emotion. n this, other reviewersat least partiallyagree. Laird (1984) concluded that these experimental studies "have demon-strated effects of varying the magnitude of expressive behavior on bothself-reports of emotionalexperience and on various measuresof physiologicalarousal such as heart rate and skin conductance" (p. 910). Winton (1986)agreed that the literature supports the modemacial feedbackhypothesis, butcautioned hat it only supports a "dimensional" iew, in that almost all studiescompare nly one positive to one negative emotion. The exception at that time



26/32

274 ADELMANN& ZAJONCwas the Tourangeau & Ellsworth study (1979), which in comparing the twonegatively valenced emotionsof fear and sadness failed to support a "categor?ical" feedback hypothesis; their study was criticized, however, on boththeoretical and methodological grounds (Hager & Ekman 981; Izard 1981;Tomkins1981). The more recent studies by Rutledge et al (1988) and Stracket al (1988) do support a categorical version, but morestudies in this vein areclearly called for. Finally, Matsumoto1987) movedbeyond a conventionalreview by submitting the studies reviewed by Laird (1984) to meta-analysis,and concluded that "the meta-analytic procedures indicate that the effect offacial manipulation on self-reported emotional experience is of moderatevalue" (p. 772).CONTEMPORARY THEORETICAL ISSUES IN FACIALFEEDBACK: WHAT IS FED BACK?The empirical literature bearing on the facial feedback hypothesis stronglysuggests that facial emotionalefference is not only correlated with emotionalexperience but maymodulateand initiate it. The evidence on the physiolog-ical componentf that experience s weaker. f facial efference plays a causalrole in the subjective experience of emotion, as the empirical literaturesuggests it does, it is likely to do so directly rather than by first initiatingphysiological arousal that is then subjectively experienced as emotion. Per-haps the most interesting contemporary heoretical question in this domain,therefore, is how acial efference mayplay a causal role in the subjectiveexperience of emotion.In general, theory has lain dormant n this question since the original facialfeedback hypotheses, in which muscular proprioceptive patterns (Gellhorn1964; Izard 1971; Tornkins 1962) and cutaneous sensation (Gellhorn 1964;Tomkins 1980) were proposed as mechanisms. Ekman 1984) has suggestedthat motor cortex directing facial muscle activity simultaneously connectswith hypothalamic areas to stimulate ANS ctivity, but his modeldoes notdirectly address subjective experience. Laird (1974, 1984), drawingon self-perception as a possible mechanism, oes specifically focus on subjectiveexperience; but the most recent facial feedback studies using mechanicalmanipulations of the face nowmakehis position less tenable.Oneexception to the current dearth of theoretical progress on facial effer-ence in the experience of emotion is a recently reclaimed theory (Zajonc1985) that links emotional efference to vascular processes. The author of thevascular theory of emotional efference (VTEE), Israel Waynbaum1907),argued that facial movementsn general, and emotional efference in particu-lar, have regulatory and restorative functions for the vascular system of thehead. He noted the intimate relationship between facial and cerebral blood



27/32

FACIALEFFERENCE 275flow (CBF), and suggested that facial muscular movements ontribute to theregulation of CBFby pressing against facial veins and arteries and thusshunting blood to the brain when eededor diverting it awaywhen he brain isthreatened with excess. The face, according to Waynbaum,cts as a safetyvalve for the brain, where blood supply can vary only within very narrowlimits. He also suggested that these regulatory muscular actions of the facehave subjective consequences: Changes n CBF aused by facial motor move-ment are reflected as changes in feeling states. He did not disagree withDarwin(1896) that the function of emotional facial gestures is to com-municate the individuals internal states to others, but he held that thecommunicative function was secondary.Several of Waynbaumsssumptions are questionable and others are wrong(Burdett 1985; Fridlund & Gilbert 1985; Izard 1985; Zajonc 1986). Forinstance, arterial flow is unlikely to be much ffected by muscular action ofthe face. Furthermore, arterial blood flow is under the control of so manyother central factors that peripheral action could only have negligible directeffects. Facial muscles, however, can affect venous flow. Regardless of theparticular physiological processes that maybe involved, it is both plausibleand theoretically important that facial efferents mayhave direct regulatoryfunctions and subjective consequences. If true, VTEE rganizes diversefindings such as biofeedback, placebo effects, unconscious preferences andaversions, growth of preference with repeated exposure, empathy,etc (Zajonc1986). The particular neurophysiological and neurochemicalprocesses are yetto be specified by empirical investigations. Useful speculations about suchprocesses that would guide future research, however, can be made.A testable hypothesis that follows from VTEEs that facial efferents canproducechanges n brain blood temperature which, in turn, can facilitate andinhibit the release and synthesis of a variety of neurotransmitters. Thus, if acertain facial muscleaction changes he temperature in a brain region whereserotonin is released, for example, then the resulting serotonin regulationmight cause the individual to experience joyful or depressive affect. Not allneurochemicalshat have subjective effects are region specific. Peptides, forexample,are found in profusion throughout the entire brain, and a change intemperature might change he threshold of the enzymaticactions that releasethem. R. B. Zajonc, Sheila Murphy, and Marita Inglehart (submitted)observed systematic correlations betweenchanges in temperature and hedonictone as simultaneous reactions to uttering various phonemes.They explainedthe results by assuming hat the productionof various phonemesmay acilitateor impair the air cooling of the venousblood that enters the cavernoussinus.The atter is a venous tructure that cools arterial blood as it enters the brain.In an experiment which subjects thought involved the psychophysics ofolfaction, cool (19 C) and warm 32 C) air was introduced into subjects



28/32

276 ADELMANN& ZAJONCnostrils and ratings of the odors were collected. On some rials no odor waspresent, yet subjects rated cool air as decidedly pleasant and warm ir asdecidedly unpleasant. To be sure, the conjecture that brain temperaturechanges can influence the release and synthesis of neurohormones andneuroenzymes ssociated with subjective emotional states still needs empiri-cal documentation.CONCLUSIONS

1. There are no sufficient grounds hus far to reject any theory of the roleof facial efference in the experience of emotion. The neglected early theoriesbased on the sensory process have a great deal to offer given that emotionalexperience dependsheavily on sensory input and its derivatives; the neuroana-tomical connections and processes indicate a powerful role of the sensoryprocess in emotion (LeDoux 987). Hence, the work of Piderit and Gratioletdeserves greater attention in contemporary esearch. In addition, there is noconflict among he sensory theories of emotional efference, Darwins evolu-tionary perspective on efference, and the facial feedbackhypotheses. Clearly,nothing prevents a facial emotional action from depending on a peripheralprocess that is allied to sensation, having adaptive communicative alue, andalso arousing in the actor the subjective experience of emotion.2. The correlational evidence reviewed here clearly indicates a positiveassociation between facial efference and emotional experience within sub-jects, particularly for the subjective component f emotion, in support of thefacial feedbackhypothesis. Intensity of facial efference of a specific emotioncorresponds with increasing subjective experience of the same emotion.3. Although he experimental evidence on facial feedback is less conclu-sive than the correlational literature, it tends to support he notion that facialefference plays not only a modulating unction but an initiating function in theexperience of emotion, particularly for subjective experience. Some nitialevidence suggests that facial efference maycausally differentiate not onlypositive from negative subjective experience, but mayproduce emotion-specific effects. More esearch comparing acial efference patterns for two ormore emotions of the same valence is needed.4. The facial feedback hypothesis does not explain why some facialactions "feel" good and others "feel" unpleasant. There is somepromise inthis respect from the vascular theory of emotional efference, whichattributeschanges n subjective hedonic states to changes n neurochemistry f the braincaused by changes in temperature reaching the hypothalamus via thecavernoussinus. If facial action can influence the thermoregulatoryaction ofthe cavernous sinus it might thereby influence the release and synthesis ofsomepeptides and neurotransmitters that are highly temperature dependent



29/32

FACIAL EFFERENCE 277and have been found to produce hedonic changes. The most significantfinding here is that for the first time a physiological indicator, foreheadtemperature, has been found to discriminate reliably between positive andnegative affect.5. The mounting evidence that facial efference under some conditionsmay modulate or even initiate subjective emotional experience suggests that atheoretical position that clings to the term "expression" misrepresents thecomplex and varied roles of the face in the experience of emotion.Literature CitedAllport, F. H. 1924. Social Psychology. Chi-cago: HoughtonMifflinAngell, J. R. 1916. A reconsideration ofJames theory of emotion in the light ofrecent criticisms. Psychol. Rev. 23:251-61Antila, C. E., Dale, J. A., Klions, H. L.1988. Depressive and elative mood nduc-tion as a function of antagonistic facial ex-pressions. Presented at Ann. Meet. EastemPsychol. Assoc., Buffalo, NYBlock, J. 1957. A study of affective respon-siveness in a lie detection situation. J.Abnorm. Soc. Psychol. 55:11-15Borod, J. C., Koff, E., Buck, R. 1986a. Theneuropsychologyof facial expression: data

from normal and brain-damaged adults. InNonverbal Communication n the ClinicalContext, ed. R. Blanck, R. Buck, N.Rosenthal, pp. 196-222. University Park,PA: Penn. State Univ. PressBorod, J. C., Koff, E., Lorch, M. P., Nicho-las, M. 1986b. The expression and percep-tion of facial emotion in brain-damagedpatients. Neuropsychologia 24:169-80Brock, S. E., Rothbart, M. K., Derryberry,D. 1986. Heart-rate deceleration and smil-ing in 3-month-old infants. Infant Behav.Dev. 9:403-14Brown, S., Schwartz, G. E. 1980. Rela-tionships between facial electromyographyand subjective experience during affectiveimagery. Biol. Psychol. 11:49-62Buck, R. 1977. Nonverbal communication ofaffect in preschool children: relationshipswith personality and skin conductance. J.Pers. Soc. Psychol. 35:225-36Buck, R. 1980. Nonverbal behavior and thetheory of emotion: the facial feedback hy-pothesis, d. Pers. Soc. Psychol. 38:811-24Buck, R., Miller, R. E., Caul, W. F. 1974.Sex, personality, and physiological varia-bles in the communicationf affect via fac-ial expression. J. Pers. Soc. Psychol.30:587-96Buck, R., Savin, V., Miller; R., Caul, W.1972. Communicationf affect through fac-

ial expressions in humans. J. Pers. Soc.Psychol. 23:362-71Bull, N. 1951. The attitude theory of emotion.J. Nerv. Mental Dis. Monogr. 81. NewYorkBurdett, A. 1985. Emotionand facial expres-sion. Science 230:608Cacioppo, J. T., Petty, R. E., Losch, M. E.,Kim, H. S. 1986. Electromyographicactiv-ity over facial muscle egions can differenti-ate the valence and intensity of affectivereactions. J. Pers. Soc. Psychol. 50:260--68Cannon, W. B. 1915. Bodily Changes n Pain,Hunger, Fear, and Rage. NewYork: Ap-

pletonCannon, W. B. 1927. The James-Langetheory of emotions: a critical examinationand an alternative theory. Am. J. Psychol.39:106-12Cannon, W. B. 1931. Again the James-Langeand the thalamic theories of emotion. Psy-chol. Rev. 38:281-95Cohen, B., Izard, C. E., Simons, R. F

facial efference and the experience of emotion

Documents