|Asymmetries in Facial Actions|
Researchers in various disciplines that study the face have found asymmetry in the action of facial muscles an important issue. Behavioral scientists have discussed whether asymmetry is one of the cues that send messages in social interaction (Darwin, 1872) and whether it is related to personality or constitution (Lynn & Lynn, 1938, 1943). In biology, researchers have considered whether asymmetry is related to genetic factors (Papadatos, Alexiou, Nicolopoulos, Mikropoulos, & Hadzigeorgiou, 1974) and the growth and development of the face (Thompson, 1943). Medical investigators have tried to use asymmetry in facial actions to locate the focus of brain lesions (Tschiassny, 1953; Remillard, Andermann, Rhi-sausi, & Robbins, 1977), and oral surgeons and dentists have included asymmetry as an objective of corrective treatments (Janzen, 1977).
Recently, this topic attracted considerable attention because it may be related to asymmetries in cerebral hemispheric function. Several studies have reported lateralized asymmetries in facial actions that were claimed to support this relationship. Some studies reported that asymmetries in facial action are lateralized so that the left side has greater electromyographic activity (Schwartz, Ahern, & Brown, 1979), has more intense expression (Sacheim, Gur, & Saucy, 1978), has stronger muscular contractions (Ekman, Hager, & Friesen, 1981), or has more frequent unilateral actions during conversation (Moscovitch & Olds, 1981). One explanation for these findings is that the nervous system functions asymmetrically to produce observable asymmetries in muscle action. Since neural connections between the cortex and the face are mostly crossed, the right hemisphere has been thought to play a special role involving the production of facial actions. Some researchers (Chaurasia & Goswami, 1975; Heller & Levy, 1981) theorized that the right hemisphere has this role because of its purportedly greater involvement in cognitive, non-verbal processes such as the recognition of faces or facial expressions (Benton, 1980). Others (Schwartz, et al., 1979; Sacheim et al., 1978) have pointed to the evidence that the right hemisphere has an important function in emotional processes (Ley & Bryden, in press). They speculated that since facial expressions are an integral part of emotion, it is reasonable to expect the right hemisphere to have a special role in the production of facial expressions. These researchers, however, did not clarify whether this role is limited to emotion expressions or applies to all facial actions, even though different types of movement have different neural pathways. This ambiguity was exacerbated by inadequately specifying what type of movements were studied, e.g., whether emotional or not (see below).
In apparent contradiction to left laterality in facial actions, other studies reported a greater facility in performing deliberate actions on the right side of the face (e.g., Alford & Alford, 1982; Kohara, Note 2) or that moving the right side of the face is subjectively more "natural" (Alford, in press). These findings might indicate that the left hemisphere is specialized for producing facial actions.
These contradictory results may be due to serious problems in these studies (see Hager, 1982, for a review). First, most reports failed to distinguish adequately which of the many facial characteristics that manifest asymmetry, such as facial structure, permanent facial features, scars, blemishes, or muscular actions, were measured. This deficiency makes it difficult to know whether the results obtained can reasonably be attributed to nervous system activity, which has direct effects on the action of muscles, or to some other factor unrelated to muscle or neural activity. Second, the type of facial movement studied, such as deliberate imitative actions versus spontaneous emotional expressions, was often not adequately considered or specified. This issue is important because different types of facial movements have different neural substrates (e.g., Tschiassny, 1953). For example, deliberate facial actions are mediated by the classical pyramidal system or corticobulbar pathways. Spontaneous emotional movements, however, are mediated by other, non-pyramidal pathways. When discussing the implications of asymmetry for neural organization, as in hemispheric specialization, the types of movement and their associated neural pathways should be distinguished. Knowing which types of movement manifest asymmetry might also indicate whether a theory relating asymmetry to emotional processes or to processes which involve deliberative control of actions is more tenable.
To compound these problems, only one measure, unique to the researcher's program, was typically used to assess asymmetry. Thus, there is no evidence about the relation of different measures of asymmetry to each other. The discrepancy between reports in the literature of right versus left laterality could arise because each different measure tapped unrelated aspects of facial asymmetry. For example, the studies which showed right sided facial laterality examined unilateral muscular actions, but only one of the studies showing left sided laterality studied unilateral actions. Multiple measures of asymmetry are useful to weigh the validity of individual measures and to assess the degree to which asymmetry in facial function is a general characteristic or is a set of discrete, unrelated phenomena.
The study reported here attempted to clarify what factors are associated with asymmetry of facial actions by correcting the problems of previous studies. The major problems are: 1) lack of multiple measures of asymmetry, 2) failure to use well-defined, sensitive measures which can discriminate specific facial actions, 3) inadequate specification and isolation of the type of movement studied, such as spontaneous versus deliberate or emotional verses non-emotional.
To remedy the lack of multiple measures of asymmetry, this study employed four measures of asymmetry in deliberate facial actions: which side of bilateral muscular actions had stronger contractions, which side was able to show a more purely unilateral action of the requested, specific muscles, which side subjects chose to use for unilateral actions when the side to move was not specified, and the side on which subjects rated unilateral actions easier to do. When discussing these measures, it is useful to distinguish between asymmetry and laterality. A particular measurement indicates whether an action is symmetrical or not. Laterality is indicated by a consistent, significant pattern of asymmetry across many measurements. Of course, there may be much asymmetry, but no consistent tendency for one side or the other, i.e., no laterality.
1. Bilateral Asymmetry
The first measure reflected asymmetry in the strength of contraction of bilateral actions. This measure was the basic measure of asymmetry used in this study. All the different types of movement examined in this study were scored with this measure, so comparisons of different types of actions were based on this measure. The other measures described below necessitated deliberate, requested unilateral facial actions and, thus, could not be applied to spontaneous actions or other expressions in the whole face. A problem with this measure is that the underlying processes it reflects are unclear. For example, previous researchers have assumed that a stronger contraction on one side of the face indicates that this side is better controlled, but there is little reason to assume this relation.
2. Better Unilateral Control
The second measure assessed how much better one side could perform a unilateral action than the other. The better unilateral side was defined by whether the request for a right or a left unilateral movement showed greater asymmetry, i.e., was more nearly unilateral. Some researchers (Chaurasia and Goswami, 1975; Alford, in press) asked subjects to perform unilateral actions on each side of the face and rated how coordinated and controlled each movement was. These researchers attempted a more direct assessment of the control exerted over each side of the face rather than making inferences based on asymmetry in the intensity of actions, but their ratings were subjective, without explicit rules, and vulnerable to bias. Measurement of the better unilateral side was more objective in the study reported here. The better unilateral action was predicted to reflect better control over that side of the face. If so, the better unilateral action should be accompanied by fewer actions of unrequested muscles than a unilateral action on the poorer side. In addition, if the strength of contraction of actions is related to factors influencing their control, the better unilateral action should have a stronger contraction than the poorer unilateral.
3. Preferred Side
The third measure of asymmetry in facial function was identifying whether there was a side of the face that subjects preferred when making unilateral actions. This measure is analogous to measures of hand, ear, eye, foot, or paw preference.
4. Side Rated Easier
The fourth measure was how much more difficult subjects rated making a unilateral action on one side than on the other. This measure can be interpreted as reflecting subjective estimation of control over unilateral actions, although subjects could have based their ratings on other criteria such as a preference or habit to use one side.
These four measures helped to show whether different aspects of asymmetry were part of a general pattern of asymmetry in facial function or whether there are different, unrelated aspects of asymmetry in facial action. The measures involving unilateral actions are likely to tap variance related to control of the action, so a strong relation of these measures to the measure of asymmetry in the strength of bilateral actions would indicate that this latter measure also reflects these control factors.
One of the most important features of this study was measuring the symmetry of each muscular action separately. This approach mitigated the problem, crippling in other techniques, of defining what features in the face are measured for asymmetry. Some studies of asymmetry in facial action (e.g., Sacheim et al., 1978; Campbell, 1978) had observers make judgments about the intensity of expressions on each side of the face, but whether observers were judging the intensity of muscular actions or some other asymmetrical characteristics was not clear, because physiognomic features, scars, hair style, lighting, blemishes, and many other characteristics could have made one side of the face different from the other and might have misled judges about the intensity of muscular action (see Hager, 1982, for a discussion). Even objective physical measurements of facial asymmetry are susceptible to this problem. EMG measurements, for example, are influenced by the amount and kind of tissue between the electrode and the muscle. The asymmetry scoring procedures used in this study minimized the influence of such extraneous factors on the measurement of actions and their intensity. A skilled facial scorer, trained to consider the effects of various characteristics on facial actions, can filter out these factors by comparing the appearances produced by the action to the inactive, baseline face.
The ability to distinguish individual muscular actions depended upon using Ekman and Friesen's Facial Action Coding System (FACS) (1976, 1978). FACS measures the visible action of facial muscles with "Action Units" (AUs) that indicate what muscles have contracted to produce the expression. Action units correspond to the anatomy of facial muscles, but rather than measuring every change in muscular action, they differentiate what skilled scorers can reliably discriminate when movements are inspected repeatedly in stopped and slowed motion. Using this technique, an expression can be decomposed into the elemental muscular actions that produced it. After identifying the particular AU to be scored with FACS, it is possible to determine whether or not there is asymmetry in the intensity of the action. Figure 1 describes the Action Units (AUs) that were measured in this study and indicates the conditions in which they were typically elicited.
By enabling accurate measurement of individual muscular actions, FACS helps to alleviate other measurement problems. Some researchers have attempted to measure the symmetry of whole expressions (e.g., Borod & Caron, 1980), but this measure could be inaccurate or inefficient. For example, actions comprising an expression might manifest different degrees of asymmetry, as suggested by Ekman et al. (1981), or even have the greater intensity on opposite sides. These differences among actions might be masked by a summary score. Also, asymmetry of only one of several actions in the expression might not be detected when looking at the expression as a whole. If the actions that show more asymmetry were known, it could make scoring more efficient by concentrating on the more asymmetrical actions. Finally, researchers comparing asymmetry of different expressions, such as positive versus negative (Borod, Caron, & Koff, 1981), cannot determine whether differences obtained are due to the hedonic quality of expressions or simply to different muscles involved in them. Using surface electromyography (EMG) to measure the activity of facial actions also cannot precisely identify what muscles acted (see Ekman, 1982, and Hager & Ekman, 1983, for a discussion).
Finally, this study solved the remaining major problem of previous studies by carefully distinguishing different types of facial movements and comparing their asymmetry. Ekman (1980) hypothesized and Ekman et al. (1981) showed that the pattern of asymmetry in strength of contraction differed between spontaneous emotional and deliberate requested actions: deliberate smiling actions were more asymmetrical than spontaneous emotional smiles and deliberate actions were lateralized while spontaneous were not. This finding is important because these two types of movement correspond to different neural processes and bear on the two theories of the relation between asymmetry in facial actions and hemispheric specialization for emotion verses for deliberative, cognitive processes. In the present study, relatively pure samples of spontaneous and deliberate actions were obtained by using eliciting conditions that maximized the opportunity to observe only one type. In addition, samples of actions were obtained that mixed the types of actions. Previous studies used similar eliciting conditions so the inclusion of these actions allowed a point of comparison. These conditions were also included to test some ideas about the causes of asymmetry described below.
A relatively pure sample of deliberate actions was obtained by requesting the subject to perform specific facial actions. Performing some of these actions might have suggested an emotion expression, but making individual actions as requested is a task that minimizes the likelihood of producing an emotional experience. Requests for deliberate actions were made in two ways, verbally and visually. For verbal requests, the experimenter described each action with phrases such a "pull your upper lip up." For the visual requests, the subject saw each action performed on television and tried to imitate it without any verbal description or coaching. These two conditions were intended to maximize the differences in the kind of information processing that occurred with the facial actions. Translating the verbal requests into motor actions could have involved relatively more left hemispheric processing than imitating the visual, nonverbal stimuli. Examining actions in these two conditions checked the possibility that asymmetry might be affected by different processes that mediate the motor actions.
A relatively pure sample of spontaneous actions was obtained by creating conditions that would minimize the likelihood of reflection or deliberate control. One type of spontaneous action, the startle expression, was elicited by startling subjects with a loud noise. Another spontaneous action, smiling related to emotion, was elicited by a mildly amusing comment by the experimenter.
Conditions that attempted to produce a mixture of spontaneous and deliberate actions were created by two general strategies. First, subjects were asked to simulate six emotion expressions and a startle expression. The type of movement obtained in this task was more ambiguous because subjects could use at least two different strategies for producing the actions, one involving emotion and one not (Ekman, Roper, & Hager, 1980). Several previous studies of asymmetry have used such requests to elicit facial actions (e.g., Borod & Caron, 1980; Schwartz et al., 1979). A second approach was to manipulate the natural response to the startle noise. Subjects were told when two noises would occur and were asked to suppress their reaction to one of them. It was thought that the process of anticipating or inhibiting a startle reaction might introduce processes that would affect the symmetry of startle actions.
One hypothesis was that different muscular actions would show different patterns of lateralization. This prediction conflicts with models of the relation between hemispheric specialization and asymmetry of facial action that have implicated a single process and specialization of a single hemisphere as underlying asymmetry. Such models imply that all muscular actions have the same laterality. The methods researchers used to explore these ideas often reflected this assumption. For example, asymmetry of entire expressions was usually measured rather than that of individual muscular actions. Other models have hypothesized specialization of both hemispheres for different processes and predict that different actions might have varying laterality. This study attempted a partial test of these independent theories by measuring individual actions and comparing the laterality observed. There were two empirical reasons for predicting laterality that depends upon the action. First, Ekman et al. (1981) found evidence that some actions showed less tendency to show laterality than others, but their samples were to small to be conclusive. Second, each of the studies that examined winking showed greater skill and/or preference for winking the right eye, and this result agrees with informal observation. Studies of other facial actions have usually found lateralization favoring the left side of the face.
Thus, a specific prediction was that deliberate facial actions that showed laterality would be lateralized left across subjects, except for actions related to blinking and winking, which should show right laterality. The preponderance of previous evidence, including the Ekman et al. (1981) study that used procedures similar to the present study's, suggested the prediction of left laterality (see Hager, 1982).
The second hypothesis was that the four measures of asymmetry would reflect the same underlying processes. If so, these measures should have significant positive intercorrelations and the same pattern of laterality for each action. Specifically, for all actions except winking and blinking, subjects should show greater strength of contraction on the left side of bilateral deliberate actions, prefer using the left side to make unilateral movements, rate the left side easier to move unilaterally, and show better unilateral actions on the left side.
The third hypothesis was that asymmetry in the strength of contraction of muscles would be related to measures of control over the action. This relation should be shown, in part, by the correlations between the bilateral asymmetry measure and the other measures of asymmetry, which more obviously tapped control factors. In addition, this relationship was predicted to be apparent in a more detailed analysis of better unilateral control scores. The unilateral action on the better unilateral side should be stronger than the unilateral action on the other side. Secondly, unilateral actions on the poorer unilateral side, as opposed to the better side, should be accompanied by stronger contractions on the side that is not supposed to move. Finally, the action on the better unilateral side should be accompanied by fewer unrequested actions than the action on the poorer unilateral side.
The fourth hypothesis was that the pattern of asymmetry would differ between spontaneous and deliberate actions in two ways. First, spontaneous actions would show less asymmetry than deliberate actions. Specifically, spontaneous smiles in response to the experimenter's humorous comment would be less asymmetrical than deliberate smiles, and actions in response to a startle noise would be less asymmetrical than when these same actions were deliberately performed individually. Second, spontaneous actions, unlike deliberate actions, were not expected to show laterality.
Related to the hypothesized differences in asymmetry between spontaneous and deliberate actions are predictions about conditions where these two types of actions were mixed, i.e., simulations of startle or emotions, and startle reactions where the subjects expected the noise. Since spontaneous and deliberate actions were thought to be mixed in simulations of startle and emotions, these actions were predicted to show fewer asymmetries and less laterality than deliberate actions, but more than spontaneous actions. Specifically, smiling actions in simulations were expected to be more asymmetrical than spontaneous smiles of amusement, but less asymmetrical than requested actions of individual muscles. Likewise, actions in simulations of startle were expected to be more asymmetrical than actions in genuine startle reactions, but less asymmetrical than the requested actions of individual muscles. Startle actions in the two conditions where subjects knew when the noise would occur were predicted to be more asymmetrical than actions in the unanticipated startle. This prediction bears on Ekman et al.'s (1981) proposition that more cognitive control and the involvement of higher nervous centers, rather than emotion, underlies asymmetry of facial actions.
The final hypothesis concerned the two modes of requesting individual deliberate actions, verbal and visual. Since viewing a model and imitating the action might have involved more right hemispheric activity than following the verbal instructions of the experimenter, visually requested actions were predicted to show more left-sided asymmetry.
The hypotheses described above are outlined below:
FIGURE 1 (cont.)