Discussion

 Laterality of Different Action Units

Different actions showed different patterns of laterality and asymmetry. Analysis of variance on the continuous asymmetry scores showed a significant effect of AU for the bilateral asymmetry and better unilateral scores. Inspection of Tables 2 and 4 showed that AUs differed in the degree of asymmetry and the tendency to be lateralized. Some actions showed no tendency towards lateralization across subjects for any of the four measures. Laterality was evident (p<.05) for at least two of the individual actions in each of three of the asymmetry measures, excepting the ratings. For two of the measures, laterality of different AUs was opposite.

The prediction that laterality of deliberate actions, other than those related to blinking and winking, would be left was confirmed for some actions, disconfirmed for others. Bilateral asymmetry and better unilateral control scores showed that some actions were lateralized right while others manifested left laterality. The other two measures, preferred side and rated easier, showed a somewhat different pattern of laterality in that no individual Action Unit showed left laterality. These measures showed a weak tendency for right laterality, but they manifested less laterality than the bilateral asymmetry measure. Subjects significantly preferred moving the right side for two of the unilateral actions requested. Although nominally more subjects rated the right side easier to move for nine of ten actions (except smiling, which was equal), no tendency for an individual action was significant.

The prediction that actions of winking and blinking would be lateralized right was partially confirmed in that the subjects preferred winking with their right eye, but the other measures showed little asymmetry for these actions. Similarly, in blinking bilaterally, subjects kept their right eye closed longer. The finding that the right eye is preferred for winking replicates the findings of two other studies (Kohara, Note 2; Alford & Alford, 1982), but the other three measures of winking and blinking in Table 5 showed little asymmetry.

The unexpected evidence for some right laterality in facial function apparently contradicts reports about only left laterality, such as that the left side showed more EMG activity (e.g., Schwartz et al., 1979), was better coordinated (Chaurasia & Goswami, 1975), or was more expressive (Sacheim et al., 1978; Borod & Caron, 1980). In most cases, these contradictions could be due to different measurements of asymmetry. The modest correlations between measures used in this study, which on their face seemed so similar, suggest that the disparate measures used in previous studies may be even less related. Also, most measures employed previously have had serious limitations. As discussed in the introduction, ratings of which unilateral movement is more coordinated have appeared vulnerable to bias, especially with no explicit scoring rules. Similarly, judges' ratings of the intensity of expressions may be spuriously influenced by cues besides muscle actions, and are unlikely to reflect accurately the intensity of actions.

As an example of how results might differ because of differences in measures, consider the findings of this study about the asymmetry of deliberate smiles versus those of a study that employed EMG measurements. AU 12 (smiling action of zygomatic major) was lateralized stronger on the left in this study, but Sirota and Schwartz (1982) reported that EMG activity from the "zygomatic placement" of electrodes was lateralized greater on the right. EMG cannot precisely identify what muscle has acted, and they admitted that other muscles in the same area might have contributed to the activity recorded from this placement. Much of the muscular activity in this area of the face either was not lateralized or tended to be stronger on the right as measured in this study . Thus, this apparent discrepancy between studies might be explained by the lack of precision about the muscles measured by EMG. Another difference is the practice of averaging EMG activity over many seconds while asymmetry was measured in this study at a particular moment (apex of action).

Although different measures might account for some contradictions between studies, Ekman et al. (1981) used the same bilateral asymmetry measure as in this study and reported left laterality for deliberate actions. Their subjects, however, were much younger and included males rather than only females. Alford (in press) reported that males had more facility with the left side of their face than females. Also, the actions measured in the Ekman et al. study were only a subset of those measured here and the sample of all but one action was too small to permit analysis of each individually. Ekman et al. had a large sample of zygomatic major actions (AU 12) which they reported as lateralized stronger on the left side. This finding was replicated strongly here.

This study partially supported claims that the right side of the face was preferred in unilateral actions. The finding that subjects preferred moving the right side for two unilateral actions corroborated Alford's report (in press) that subjects generally found moving the right side of their face "more natural." On the other hand, the lack of laterality in ratings and most of the preferred side scores indicates that this effect is weak or largely absent.

Relations among Measures of Asymmetry

Conceptually, scores indicating the side of the face with a more intense bilateral action, the side that performs a unilateral action better, the side that subjects prefer for making unilateral actions, and the side rated easier to move unilaterally each took a somewhat different view of asymmetry in facial function. Intercorrelation of these four measures of deliberate facial actions showed that the different conceptual aspects of asymmetry were partially related. All the significant correlations among these asymmetry measures were positive, indicating that the separate measurements reflected some common variance rather than isolated phenomena. On the other hand, only half the intercorrelations were significant, and the magnitudes of significant correlations were modest. Artifactual problems, such as weaknesses in the measurements and low score variance, could have limited their magnitude, but these results were indications that the four measures might tap different factors affecting asymmetry.

Ratings of the side easier to move and scores indicating the side subjects preferred for unilateral actions were similar measures. Averages of ratings and preferred side scores loaded highly on a factor with the average better unilateral score (Table 8). Factor analysis of scores for individual actions within each asymmetry measure showed that factors for ratings and the preferred side were similar to each other, but different from the other measures (Table 10). These results and the significant correlations between these two measures (Table 6) suggest that they tap the same set of variables.

Scores for the side with the better unilateral action were similar to rated easier and preferred side scores. The average better unilateral score loaded on the same factor as averages of ratings and preferred side scores (Table 8) and usually correlated significantly with these measures (Table 7). Better unilateral scores also measured some variance in common with the bilateral asymmetry scores as shown by the patterns of correlations (Table 6). Factors of individual scores within the better unilateral and bilateral asymmetry measures were different from each other and from the other two measures (Table 10). Disparity between measures was greatest for bilateral asymmetry scores compared with rated easier and preferred side scores because these measures had the fewest number of significant correlations and showed the least similarity in factor analysis. Like measures of facial asymmetry, measures of handedness also show varying degrees of intercorrelation (Johnstone et al., 1979).

Although preferred side and rated easier scores were very similar to each other, they have the most dubious value for assessing asymmetry in facial function. As discussed below, these two measures showed the least tendency towards laterality, and rated easier scores consistently showed the least asymmetry, as though subjects were not sensitive to differences in difficulty. On the other hand, the bilateral asymmetry score showed the most lateralized individual actions, and it is applicable to any facial behavior that can be adequately recorded, rather than only deliberate, requested actions. The better unilateral score appeared to tap variance that overlapped most with all three of the other measures, but it showed only modest laterality.

Asymmetry of Facial Actions and Hemispheric Specialization

The asymmetry of actions measured in this study highlights the inadequacies of several models of hemispheric specialization for explaining asymmetry and laterality of facial actions. One shortcoming is relying solely on the specialization of one process in a single hemisphere to explain asymmetry in facial actions. Another problem is attributing asymmetry to emotional processes. For both the bilateral asymmetry and better unilateral measures, asymmetries of some actions were lateralized left while others were lateralized right. This finding compels the conclusion that one kind of specialization in a single hemisphere cannot be the only factor that produces asymmetries in facial actions, given the assumption that such specialization would affect all actions the same way. Other evidence corroborates this conclusion. Specialization for one kind of process in a single hemisphere implies that a given individual subject would show the same asymmetry for all actions, but most subjects showed a mixture of left and right asymmetry that changed with the AU, and few subjects showed laterality except in self-report ratings. This pattern was evident for each of the four asymmetry measures.

Specialization for Emotion

Researchers who have found left lateralization of facial activity have speculated that specialization of the right hemisphere produced it. Some have hypothesized that the right hemisphere is specialized for emotion and that this asymmetry affects the symmetry of facial actions (Schwartz et al., 1979; Borod & Caron, 1980). Spontaneous emotional and reflex movements studied here were usually not lateralized, replicating the results of Ekman et al. (1981). Only action of orbicularis oculi in the startle showed lateralization, but the direction was opposite to that predicted by right hemispheric specialization for emotion. It is possible that this study did not measure enough spontaneous actions to detect significant tendencies for lateralization, but it did replicate the findings of Ekman et al. (1981) that there was generally less asymmetry in spontaneous than in deliberate actions. These findings suggest that at least some of the factors that produce asymmetry and laterality in deliberate movements are not related directly to emotional processes, at least for positive emotions involving smiling, nor to processes giving rise to negative, reflex-like startle reactions. Further research is needed to determine whether this pattern holds for negative emotions.

It might be possible to explain the opposite laterality of different actions with some type of dual specialization where some actions are affected by the activity of one hemisphere while other actions are affected by the other hemisphere. Some researchers who have proposed hemispheric specialization for emotion have argued that the right hemisphere is specialized only for negative emotions, but the left is specialized for positive emotions (Schwartz et al., 1979; Reuter-Lorenz & Davidson, 1981; Sacheim & Gur, 1978). This theory predicts that actions related to positive emotion would be lateralized stronger on the right while actions related to negative emotion would be lateralized left. The finding that deliberate actions of AU 12 (the smile involved in positive emotion) are significantly lateralized stronger on the left is opposite to this theory's prediction for positive expressions, and the finding that spontaneous actions of AU 12 show only the same non-significant tendency lends no support to this position. The asymmetry of some deliberate actions (AUs 9, 15, and 20) which are often found in negative emotion expressions were significantly stronger on the right rather than on the left as their theory predicts for negative facial expressions. The right stronger laterality of orbicularis oculis (AUs 6 and 7) in the startle also does not support this model. While there is some dispute about whether startle is an emotion, almost all subjects said it was an unpleasant experience whether or not they knew when the noise would occur. However, subjects might have reported their negative expressions based on subsequent reactions to the startle reflex rather than the startle itself so whether the startle expression is negative is not certain.

A variant of this dual specialization for emotion theory is that the right hemisphere is specialized for avoidance emotions while the left hemisphere is specialized for approach emotions (Davidson & Fox, 1982). In this study, deliberate actions often involved in approach emotions (e.g., AU 4 in anger; AU 12 in happiness) were lateralized stronger on the left, and those often involved in avoidance emotions (e.g., AU 9 in disgust; AU 20 in fear) were lateralized right stronger. These relations are the opposite to those predicted for facial expressions by this theory. Again, these results indicate that asymmetry in deliberate actions or spontaneous smiling is not produced directly by the processes described by these theories of emotion and facial expression.

Specialization for Control of Actions

Some researchers have argued that rather than reflecting emotion, left lateralization of facial activity indicates that the right hemisphere is specialized to direct facial actions (Chaurasia & Goswami, 1975; Heller & Levy, 1981). They assumed without evidence that greater intensity on one side of a facial action indicates greater neural motor control over that side and, therefore, contralateral hemispheric specialization for controlling actions. The study presented here established relationships between the control of actions and asymmetry in their intensity. First, asymmetries in intensities of bilateral actions were generally related to measures of control over actions, i.e., which side had a better unilateral action, was preferred in making a unilateral action, and was rated easier to move. Second, in making unilateral actions, subjects were required to control intensity to augment movement on the onside (side where movement was requested) and minimize movement on the offside (side where movement was not to appear). Onside intensity was greater when the side that performed the better unilateral action was requested than when the poorer side was requested. This finding supports the notion that better control of an action is associated with greater intensity, at least on the side where movement is requested in unilateral actions. Conversely, offside intensities were greater when the poorer side was requested. This finding indicates that poorer control of an action is associated with greater intensity where the action is not supposed to appear. The finding that fewer unrequested actions occur with better than with poorer unilateral actions corroborates the assertion that this measure reflects subjects' ability to control facial actions.

Although this study supported a relation between intensity and control of an action, it did not support the hypothesis that the right hemisphere is specialized to direct facial actions. Looking across all the measurements in Tables 1, 2, and 4, more lateralization of deliberate actions was right than left, suggesting that if any hemisphere specializes in controlling facial actions, it might well be the left.

Although specialization for controlling facial actions might be one factor influencing asymmetry, one model of hemispheric control is ruled out by the relation between the offside intensity and the better unilateral action. If one hemisphere simply is better at directing the action, then we would expect the offside intensity to be greater when the better side does a unilateral action because the nonspecialized hemisphere is less able to keep the offside still. The opposite relation was found, which suggests that the offside intensities are determined by which hemisphere can inhibit the other's activity better.

Geschwind (1965, 1975) proposed a model of hemispheric specialization for control of facial actions that comes closest to fitting these findings. By studying facial apraxias, Geschwind inferred that a neural center in the left temporal lobe typically sends messages about making movements to the left precentral motor cortex, which controls facial movements on both sides. Geschwind argued that bilateral facial movements are typically integrated by the left hemisphere, particularly in response to verbal requests for movements, but that the right hemisphere can control movements in certain conditions, such as when the request is non-verbal. This model is compatible with the findings that laterality in this study was mostly right, but changed to the left depending upon the AU and the measurement. Geschwind did not indicate that hemispheric specialization could be observed in the normal actions of non-patients, nor did he indicate whether the left hemisphere typically inhibits any control the right hemisphere might attempt to exert.

Left hemispheric specialization for controlling deliberate facial actions can account for right facial laterality, but cannot obviously explain why actions of AUs 4 and 12 were lateralized left stronger. There are several possible explanations for the difference in laterality among actions. First, the right hemisphere could be specialized to control the actions that were lateralized left stronger. It is difficult to explain with this hypothesis why the rated easier and preferred side scores did not show the same differences among actions as the other two scores, unless the measures were insensitive. Second, the assumption that the specialization of one hemisphere affects the symmetry of all actions equally might be incorrect, perhaps because some actions are subject to different kinds of control. Ekman and Friesen (1975) have pointed out that the control of facial actions has several components. It is possible that control is different depending upon the action. For example, some actions might be inhibited more often and others might be put-on or intensified more often. Even if control were lateralized in one hemisphere, different actions might show different laterality depending upon how they were typically controlled.

Other explanations of the differences in laterality between actions were found to be inadequate. The area of the face was not a factor since both left and right laterality was found in the upper and the lower face. The possibility that some kind of emotional process entered into the process of deliberately making actions of smiling and brow lowering was rejected. First, this explanation implies that spontaneous actions of smiling would be lateralized, but they were not. Second, many actions that were lateralized right are involved in emotions that are probably as common as those in which smiling and brow lowering are found. How common the action is for other functions besides emotion also seems inadequate. Although there are no norms available for the frequency of occurance of actions, brow raising, which was lateralized right, is probably as common as brow lowering and smiling, which were lateralized left. Galin has suggested (personal communication) that the right hemisphere is specialized for inhibiting or modulating emotion expression, rather than for emotion itself. Smiling (AU 12) and frowning (AU 4) are actions that appear in emotion expressions that are often controlled (e.g., feigning happiness or masking with a smile and anger, respectively). Whether these expressions and actions are more commonly controlled than others is an issue for further research.

The point of this speculation is that more sophisticated ideas about the causes of facial asymmetry are required to explain the difference in laterality between actions. The results do not support theories about right hemispheric specialization for emotion or controlling facial actions, nor more complex theories about one hemisphere specialized for positive or approach emotions and the other for negative or avoidance emotions. Geschwind's theory about typical left hemispheric control of facial movements is the model of hemispheric specialization most compatible with the results, but what conditions make control shift to the right hemisphere and how this switch is related to the actions of AUs 4 and 12 need to be clarified.

Other Possible Causes of Asymmetry in Facial Actions

Verbal vs. Visual Processes

There was no evidence in this study that asymmetries of actions were mediated differently by verbal versus visual processes. Evidence from brain-damaged patients suggests that neural organization for processing verbal requests is different from processing nonverbal information (Geschwind, 1965). No difference in the laterality of actions that were verbally versus visually requested was detected, and these scores were significantly correlated. These results suggest that it does not make any difference in the asymmetry measured here whether actions are mediated by verbal or nonverbal requests.

Structural Asymmetry

Some evidence indicated that one of the factors affecting asymmetry might have been physical, structural characteristics. First, asymmetry scores for different actions tended to be unrelated except for certain clusters of actions which could be seen in correlation matrices and factor analyses. Many of these relationships reflected the anatomy of the muscles and nerves underlying different actions. For example, scores for AU 1, 2 and 1+2 tended to be related (Table 9 and 11) and to load on the same factor (Table 10). AUs 1 and 2 are the inner and outer strands of the same muscle (frontalis) and AU 1+2 is the brow raise action of both these strands. Likewise, the correlation of AUs 9 and 10 might be attributed to their being different heads of the same muscle (levator labii superioris).

Another finding suggesting a physical, structural basis for asymmetry was the analysis of the action's intensity on each side of the face during unilateral actions (Table 12). The onside intensity was greater when the side with the better unilateral was requested, and the offside intensity was greater for the request for the poorer side. In other words, when comparing either onside or offside intensities, intensity was less on the poorer side of the face. The consistent tendency for one side to have weaker contractions might be attributed to a structural factor.

Unilateral movements also might have reflected known anatomical differences between the innervation of the brow versus the lower two thirds of the face (Thompson, 1982). As discussed in Appendix Y, most subjects could produce asymmetrical actions of lower face muscles with greater intensity on either side of the face (Table Y1). Few subjects could do the same for brow actions. This pattern might be related to the largely crossed corticobulbar pathways to the lower face, versus the greater bilateral innervation in the upper third of the face.

The possibility that peripheral factors influence asymmetry raises interesting questions about the antecedents of the asymmetry of structural components and why they are lateralized. Little is known about the causes of asymmetry in structural tissues, but the action of muscles is an important factor influencing their own size and strength as well as the growth and shape of bone. Asymmetries in the action of muscles might produce asymmetries in structural tissues. This relation could explain why there is more laterality in measures of normal, bilateral muscle action than in preferred side scores and ratings. The effects of weak laterality in the use of muscles could have accumulated over the years to yield a more marked laterality in structural tissue. Bilateral asymmetry scores would be influenced more by these factors than ratings of difficulty if the structural asymmetries do not affect a subject's evaluation. In other words, an observer who compares intensity on each side might see more asymmetry than the subject feels.

Peripheral factors such as structural asymmetries can not explain the entire pattern of results in this study. First, such factors alone do not explain why some actions are lateralized right, some left, although different structural factors might affect each separate action differently. Second, peripheral factors are likely to affect all types of movement equally, but asymmetry of spontaneous actions was different than deliberate actions. Exactly what structural properties are contributing to laterality and how well they can account for the laterality in preferred side and rated easier scores also need clarification.

Opposite laterality for different AUs might be explained by structural factors opposing the effect of specialization of one hemisphere and shifting laterality of some deliberate actions. This explanation, however, is inconsistent with the possibility that hemispheric specialization causes structural asymmetry. Also, structural asymmetries such as sizes of muscles, should be reflected in the symmetry of spontaneous actions, but they were not as spontaneous actions generally showed little asymmetry and no laterality. It is possible that spontaneous actions might, contrary to expectation, not reflect asymmetries in structural factors that affect deliberate actions. This could happen, for example, if there were differences between the neural impulses involved in spontaneous versus deliberate actions that interact with structural characteristics to reduce the asymmetry of spontaneous compared to deliberate actions. Although such interactions are hypothetical, spontaneous emotional actions have different neural substrates than deliberate actions. These differences might not be limited to neural pathways, but could include differences in the ratio of crossed versus uncrossed connections, rates or sequences of motor unit firing, etc. If so, asymmetries in the intensity of actions caused by structural asymmetries might be reduced in spontaneous actions by the effects of different neural activity.

Emotional vs. Cognitive Activity

The finding that spontaneous actions were generally more symmetrical than deliberate actions corroborated the results of Ekman et al. (1981) who found this relation for the smiling action (AU 12). They speculated that the greater involvement of cortical processes in deliberate actions as opposed to emotional actions could have induced a greater degree of asymmetry. This hypothesis was tested by seeing whether the average degree of asymmetry during simulations of each emotion was related to how much emotion the subject felt during the simulation. The hypothesis was confirmed for the happy simulation, but was the opposite for the fear simulation.

This same idea suggested the prediction that actions in the unanticipated startle might be more symmetrical than actions in the other startle conditions because anticipating the noise or trying to inhibit it might involve more cognitive processes. The finding that there were no differences among conditions indicates that whatever subjects thought or did during these startles did not affect the symmetry of their actions and suggests that the symmetry of the startle pattern is not very vulnerable to modification.

The disconfirmation of this prediction is not strong evidence against the hypothesis that cognitive processes account for the difference in asymmetry between emotional and deliberate actions for two reasons. First, Ekman et al. (in press) have presented evidence that the startle reaction is unlike emotional expression. Thus, startles may not have the same neural substrates nor reflect the same asymmetry patterns as emotion expression. Second, deliberately producing actions may involve different neural processes from reacting naturally to actions elicited by a reflex or trying to inhibit these actions, as subjects did in the anticipate and inhibit startle conditions. Further research is needed to clarify what factors underlie the difference in asymmetry between spontaneous and deliberate actions.

Need for Better Concepts

In summary, this study has shown that the asymmetry of certain deliberate actions is consistent across individuals in this sample of women subjects. This consistency implies that the subjects have in common some functional asymmetry related to differential use of the hemispheres or some structural, anatomical asymmetry or both kinds of asymmetry. Evidence from this study indicated that although the specialization of a single hemisphere for one kind of process might explain some asymmetry and laterality of facial actions, it cannot be the only factor. Laterality might differ among actions because both hemispheres are specialized, but too little evidence is available that can clarify how this dual specialization might work. There was no evidence that asymmetry is a mere by-product of whether the actions are mediated by verbal or visual requests.

Physical, structural asymmetries in nerves, muscles, or other tissues can account for some patterns of asymmetry and differences in laterality among actions, but the cause of such structural lateralization calls out for explanation. Differential use of the hemispheres in motor actions is one explanation. Structural asymmetries cannot be the whole story either because all types of movement should reflect this asymmetry, but spontaneous actions did not show the same pattern of asymmetry as deliberate actions. Why the pattern of asymmetry differs between spontaneous and deliberate actions is not yet clear.

With so many factors that could affect observed asymmetry in the face, it seems reasonable that the results are more complex than originally hypothesized. The asymmetry and laterality observed in this study might be a product of of different factors or a fluctuating combination of them. These factors could interact in complex ways with the action studied, the type of movement observed, and the specific measurement that is made. Clearly, more research is needed to clarify precisely how these factors influence facial asymmetry.