یک حالت صورت از PAX: بررسی "بازشناسی" احساسات کودکان از صورتها

Abstract In a classic study, children were shown an array of facial expressions and asked to choose the person who expressed a specific emotion. Children were later asked to name the emotion in the face with any label they wanted. Subsequent research often relied on the same two tasks—choice from array and free labeling—to support the conclusion that children recognize basic emotions from facial expressions. Here five studies (N = 120, 2- to 10-year-olds) showed that these two tasks produce illusory recognition; a novel nonsense facial expression was included in the array. Children “recognized” a nonsense emotion (pax or tolen) and two familiar emotions (fear and jealousy) from the same nonsense face. Children likely used a process of elimination; they paired the unknown facial expression with a label given in the choice-from-array task and, after just two trials, freely labeled the new facial expression with the new label. These data indicate that past studies using this method may have overestimated children’s expression knowledge.

Introduction When presented with an array of facial expressions (such as those claimed to signal happiness, sadness, and fear) and asked to find the person who is afraid, children as young as 2 years select the predicted expression more often than would be expected by chance. The assumption has been that the task reveals children’s prior knowledge of what emotion is signaled by, in this case, the fear expression. Here we present evidence that at least some of this knowledge can instead be created by the experiment. Evidence from the choice-from-array task has been claimed to show that young children recognize happiness, sadness, anger, fear, surprise, disgust, and even pride (Denham et al., 1990, Gagnon et al., 2010, Gagnon et al., 2014, Harrigan, 1984, Izard, 1971, Kayyal and Russell, 2013, McClure, 2000, Nelson and Russell, in press, Székely et al., 2011 and Tracy et al., 2005). Izard (1971) showed that, following the choice-from-array task, children also freely label the facial expression with the predicted emotion label. Indeed, children as young as 2 years produce the expected label for at least some of the same expressions (Gates, 1923, Izard, 1971, Markham and Adams, 1992, Nelson and Russell, 2011b, Widen, 2013, Widen and Russell, 2003 and Widen and Russell, 2008), again complementing the conclusion that young children recognize emotions from facial expressions. Many subsequent studies of children used the same tasks: choice from array and free labeling. The idea of recognition of certain facial expressions during early childhood is embedded within the prevailing conceptual framework in the science of emotion, namely, basic emotion theory. Evidence with young children complements evidence from remote cultures (Ekman, 1980) and from people born with congenital blindness (Tracy & Matsumoto, 2008). Although basic emotion theory does not predict the precise age at which children can recognize basic emotions from their facial expressions, some researchers have offered evidence that such recognition occurs during the first year of life (Hoehl and Striano, 2010, Izard, 2011, Izard et al., 2010 and Martens et al., 2013). Such evidence resonates with basic emotion theory because early recognition fits with its evolutionary account in which caregivers can signal to a preverbal child that, for example, a certain animal is dangerous or a certain food is disgusting; the child who then avoids the current harm is more likely to survive (Izard, 1994). In this article, we argue that the choice-from-array task can yield illusory recognition; children can “recognize” various, even nonexistent, emotions from the same made-up nonsense expression—an expression that no one has proposed as a signal of any emotion. When children know some face–label pairings, they can then use a process of elimination to match the unknown face to a label not yet paired with a face. A fast-mapping process may then allow children to learn, with just one or two exposures, the new label associated with the new expression. When the choice-from-array task is followed by a free labeling task (e.g., as in Izard, 1971), children then use their newfound knowledge of the face and label pairing to label the expression with the expected label. If children used this strategy in emotion recognition tasks in prior research, they would have appeared to recognize the correct emotion from an expression—in both the choice-from-array and subsequent free labeling tasks—even if they were unfamiliar with either the emotion label or expression presented. Preschoolers and toddlers have been shown to use the process of elimination as a word learning strategy; children match unfamiliar labels to objects or properties that they do not yet have labels for, including color (Carey & Bartlett, 1978), objects (Au & Markman, 1987), animals (Au & Glusman, 1990), verbally provided facts (Waxman & Booth, 2000), adjectives (Diesendruck, Hall, & Graham, 2006), and actions (Waxman, Lidz, Braun, & Lavin, 2009). It is not known, however, whether the process of elimination and fast mapping occur with facial expressions. Of course, if children already know which basic emotion goes with a facial expression, an elimination strategy would not be necessary. Whether children use a process of elimination when confronted with an array of facial expressions has not, to our knowledge, been explored.

Conclusion During the past 40 years, research on children’s understanding of facial expressions has found that they reliably select, for example, the fear face when asked to find someone who is scared. Researchers then concluded that children “recognized” the fear in the fear face. Five studies here showed that this same method can create illusory recognition. Across the studies here, the puffed-checks novel expression shown in Fig. 1 was “recognized” as an expression of pax, tolen, fear, and jealousy. As shown in Table 1 and Table 3, after just a few trials, 74% of the 120 children tested selected that novel expression when presented with a label that did not apply to the other expressions in the array regardless of whether that label was novel (pax or tolen) or familiar (scared or jealous). By the logic of the standard method, we would conclude that children “recognized” the pax in the pax face. The explanation we offer for this illusory recognition is that the choice-from-array task allows children to use a process of elimination to guess the expected answer, particularly if children are already familiar with the other expressions in the array. Our data show that even with just two trials, children’s matching of the expected label and the novel expression was robust and was not contingent on whether the label presented was an emotion word or not, whether children were familiar with the concepts conveyed by the target label, or whether the label was negative or positive (Nelson & Russell, 2015). Although we included only one novel expression in these studies, we speculate that process of elimination might produce above-chance “recognition” for a greater number of expressions that are unknown to children. Prior research has used many choice-from-array trials such that children see the same expressions repeatedly; for example, Izard (1971) presented 39 triads composed of nine expressions, and children saw each expression 13 times (either as target or as distractor). If children used a process of elimination to learn one new label early in a trial, they could then use that information in later trials to eliminate distractors and perhaps learn additional labels. Thus, children may learn new expressions across trials, divining the expected answer for several unfamiliar expressions by the end of the task, by simply using a process of elimination. Indeed, when a study includes many trials, a greater number of participants appear to recognize each expression (Nelson & Russell, 2013). However, even with just one exposure, children could still produce illusory recognition; if children match a label randomly between two unknown expressions, this would produce 50% recognition, which would exceed the 33% typically expected due to chance. Past research has hinted at the possibility that the choice-from-array task might exaggerate the amount of recognition. Specifically, the percentage of children scored as recognizing a given emotion varies with the number of expressions in the array. When preschoolers were presented with arrays of two expressions, their performance ranged from 72% to 85% (Gagnon et al., 2010 and Russell and Bullock, 1985). When preschoolers were presented with arrays of three or four expressions, their performance ranged from 63% to 71% (Denham and Couchoud, 1990, Harrigan, 1984, Izard, 1971, Tracy et al., 2005 and Walden and Field, 1982; but see Székely et al., 2011). When preschoolers were presented with arrays of 10 expressions, their performance averaged 44% (Bullock & Russell, 1984). That performance decreases as the number of faces in the array increases suggests that performance in choice-from-array tasks is inflated by guessing, an issue that is receiving renewed attention in the field of emotion (Cassels and Birch, 2014 and Gendron et al., 2015). We do not claim that children are as certain about the association between the novel expression and the label pax or scaredas they are about the association between familiar expressions such as smiles and labels such as happy. This is shown in children’s greater likelihood of labeling familiar expressions as expected than labeling the novel expression as pax in many of the studies we present. However, our data call into question prior research using choice-from-array tasks; children may have appeared to recognize expressions by relying on a process of elimination. In past research, the choice-from-array task is sometimes followed by a free labeling task (Harrigan, 1984, Izard, 1971, Markham and Adams, 1992 and Walden and Field, 1982). However, as shown in Table 1 and Table 3, after just two choice-from-array trials, 61% of the 120 children in these studies went on to spontaneously label the novel expression with the expected label, behavior that increased with age. These data suggest that the choice-from-array task can teach children to associate specific expressions and labels even after only a few trials and call into question the veracity of free labeling data collected after children completed a choice-from-array task. Although the results here were robust across various changes in method, an uncountable number of such method changes can be explored. It might be argued that the methods used in the studies presented here differ in some ways from the methods used in prior research using choice-from-array tasks. For example, in some prior research, children were asked to select only one emotion from an array of three expressions (e.g., Harrigan, 1984, Izard, 1971 and Tracy et al., 2005), were presented with expressions posed by several posers (e.g., Bullock and Russell, 1984, Gagnon et al., 2010, Izard, 1971 and Russell and Bullock, 1985), or were presented with the labels in a random order (Denham and Couchoud, 1990, Harrigan, 1984 and Izard, 1971). Each of these methodological differences may influence the likelihood that children will use a process of elimination strategy, a question that can be answered with future research. However, Izard (1971) found that 62% of preschoolers matched the expected expression with the expected label, whereas 74% of preschoolers in the studies presented here matched the novel expression with the expected label. If the use of methods more similar to those used in previous research decreases children’s likelihood of matching a novel expression and a target label, their performance may simply drop to be more similar to that of prior research. Unless performance were to decrease to levels lower than expected by chance guessing, these data would still be consistent with the idea that children in prior research could have used a process of elimination strategy to appear to “recognize” expressions. As with most research, these studies had limitations. We used static facial expressions presented in isolation and in an array—factors that are likely to differ from children’s daily experiences with expressions. Thus, the external validity of these studies is unknown. We also primed children with only the three labels used in each study and presented the expression labels in the same order for each study (i.e., for the first trial, pax was always presented last). It is possible that had we expanded the list of priming labels or varied the order of trials, children’s responses may have differed. We also had small sample sizes for each study (precluding the analysis of gender effects due to the small number of male and female participants per cell), which might call into question the validity of our findings. Our results were replicated across several studies, however, providing some evidence of robustness of the effect. Finally, it is likely that children’s response patterns would vary if two novel expressions were included in the array; future research should examine these variations. Finally, it might be argued that the results of these studies were due entirely to task demands placed on children or that children were insufficiently encouraged to respond in the negative to the experimenter’s questions and, thus, that these results do not truly reflect their reasoning about emotional expressions. Relatedly, it might be argued that the task presented here, in which an expression not intended to convey an emotion was presented as emotional, is a situation children are unlikely to encounter and that, in general, their emotion matching skills are well represented by the choice-from-array task. Although we agree that the choice-from-array task can place demands on participants (child or adult) and cause them to respond in ways they would not do spontaneously, we do not agree with the supposition that this concern applies only to “fake” expressions rather than “real” expressions. Children are unlikely to be equally familiar with all expressions presented in prior research (e.g., children may witness expressions of fear less often than they witness expressions of happiness); arrays that contain unfamiliar expressions would be similar to the arrays presented here. If children are subject to task demands when presented with an unfamiliar expression in a choice-from-array task, previous findings could be inflated or even primarily driven by the demands of the task and children’s use of the process of elimination, as was the case for children’s recognition of pax. To the extent that basic emotions theory assumes that even young children innately recognize particular novel facial movements as displaying a discrete emotion—for example, recognizing a fearful expression as displaying fear in their first encounter with the expression—the parallel assumption must also be made; children must then ignore many nonemotional facial movements during their interactions with others. Nonemotional expressions would be recognized as having no value as emotional signals. The data presented here, however, suggest that children do not view nonemotional expressions as something to be ignored but rather view them as something to be incorporated into their existing emotion knowledge. Although many variations of basic emotions theory have been suggested during the past 40 years, the core prototypical version of the theory remains the most influential (Russell & Fernández-Dols, 1997). Even when an author modernizes his or her theory or tempers it in the face of new evidence, these changes might not be noted by the field at large; for example, over the past 3 years, Izard’s (1994) original theory has been cited 207 times, whereas his more modern theory, outlined in Quinn et al. (2011), was cited just 15 times. The field continues to debate issues related to children’s recognition of emotion partly because, as was nicely articulated by Quinn and colleagues, although the evidence is weak for infants’ recognition of discrete expressions, the possibility still remains and researchers continue to search for evidence of this phenomenon. Thus, our results are an important addition to a discussion that continues to progress within the field; extra care must be taken in interpreting data that appear to show that young children recognize emotion from facial expressions but may be, instead, an artifact of method. Our studies suggest a direction for future research; the process of elimination might be one way in which children acquire emotion labels for facial expressions in nonlaboratory settings. Children encounter a constant stream of facial movements that may include emotional expressions but also talking, eating, nonemotional situation-specific behaviors (e.g., squinting in the sun), and any number of additional facial movements. One question is how children learn which particular movements constitute an emotional expression and are associated with a particular emotion. Perhaps early in their life span, all facial movements are novel for children; certainly. None has a known label. Children might rely on adults’ labeling of an expression to determine which movements they should attend to and remember as indicative of emotion. Once children are familiar with one expression and its label, they would use a process of elimination to match other expressions with other labels. This hypothesis is consistent with prior research documenting children’s use of an elimination strategy to learn categories in a variety of other situations. This process may also ensure that children’s expression knowledge is well-tailored to their culture, environment, and individual caregivers.