Cross-Cultural Emotion Recognition and In-Group Advantage in Vocal Expression: A Meta-Analysis

Abstract

Most research on cross-cultural emotion recognition has focused on facial expressions. To integrate the body of evidence on vocal expression, we present a meta-analysis of 37 cross-cultural studies of emotion recognition from speech prosody and nonlinguistic vocalizations, including expressers from 26 cultural groups and perceivers from 44 different cultures. Results showed that a wide variety of positive and negative emotions could be recognized with above-chance accuracy in cross-cultural conditions. However, there was also evidence for in-group advantage with higher accuracy in within- versus cross-cultural conditions. The distance between expresser and perceiver culture, measured via Hofstede’s cultural dimensions, was negatively correlated with recognition accuracy and positively correlated with in-group advantage. Results are discussed in relation to the dialect theory of emotion.

Keywords

culture dialect theory emotion in-group advantage meta-analysis speech universality voice

Introduction

Evidence for cross-cultural differences and universals in emotion has largely depended on the study of facial expressions (e.g., Elfenbein & Ambady, 2002; Scherer, Clark-Polner, & Mortillaro, 2011). Photographs created in the United States were shown to observers from countries as diverse as Argentina, Brazil, Chile, Japan, and Papua New Guinea, who were asked to make multiple-choice judgments of which emotional states were intended among the six “basic” categories of anger, disgust, fear, happiness, sadness, and surprise (Ekman, 1972; Izard, 1971). Observers generally achieved greater accuracy than would be expected due to chance guessing alone—that is, 16.7%—and so began an account of emotion as “universal.” Since that time, more nuanced accounts have advanced the question whether the nature of this evidence is sufficient to claim universality (e.g., Russell, 1994). Given how central this body of work has been to the continuing debate about universals and cultural differences, the current article asks to what extent such evidence exists for emotion expressed through the voice.

The voice provides a rich medium for nonverbal communication of emotion—whenever we vocalize, we do not only convey the meaning contained in the words we use, but also convey emotional information through speech prosody and nonlinguistic vocalizations. Recent studies have demonstrated how a wide variety of both positive and negative emotions can be recognized from vocal expressions (e.g., Cowen, Elfenbein, Laukka, & Keltner, 2019). There is also a sizable literature on cross-cultural vocal emotion recognition (for reviews, see Laukka et al., 2016; Scherer et al., 2011). In this article, we present a quantitative review of this work, in which findings from this literature are aggregated into a meta-analysis that assesses the extent to which emotions expressed through the voice can be recognized across cultures.

The present investigation is grounded in dialect theory (Elfenbein, 2013). Research in the field of linguistics proposes that spoken language continually evolves, which leads to different dialects across groups of people who become separated by cultural or geographical boundaries (O’Grady, Archibald, Aronoff, & Rees-Miller, 2001). Using a linguistic metaphor, dialect theory argues that there are subtle differences across cultures in the style of emotional expression—much like the dialects of a spoken language—which become greater the more culturally distant the groups are from each other. As in linguistic dialects, expressive dialects are largely recognizable across groups and yet some of the meaning can get lost along the way. This leads to an empirical observation of in-group advantage, namely that emotion recognition is more accurate when judging expressions from one’s own cultural in-group compared to cultural out-groups. The current meta-analysis was designed to test three corresponding hypotheses from dialect theory:

Hypothesis 1: There is accuracy at levels above chance in the cross-cultural recognition of emotion in the voice.

Hypothesis 2: There is in-group advantage across cultures in the recognition of emotion in the voice.

Hypothesis 3: In-group advantage in vocal emotion recognition is greater the more cultural distance there is between the expresser and perceiver cultural groups.

Two previous meta-analyses have considered cross-cultural recognition of emotion in the voice (Elfenbein & Ambady, 2002; Juslin & Laukka, 2003). Both reported that vocal expressions were recognized with better than chance accuracy across cultures, and also found evidence for in-group advantage. However, previous research syntheses were limited by the small number of studies available at the time, and thus were able to only report results for a limited number of individual emotions. For example, Juslin and Laukka (2003) only reported cross-cultural recognition rates for five emotions, and Elfenbein and Ambady (2002) for four emotions specifically for vocal expression. In our update, we benefit from all the cross-cultural studies published in the past 2 decades and consider a wide range of both positive and negative emotions.

Materials and Methods

Study Details and Criteria for Selection and Inclusion

A total of 37 cross-cultural vocal expression studies were included in the meta-analysis (see online supplemental Table S1 for details). These studies provide data from more than 4,210 perceivers’ judgments of vocal expressions from more than 272 expressers (some studies do not report the numbers of perceivers and/or expressers). The median number of perceivers and expressers for each pairwise cultural comparison was 30 and 3, respectively. Expressers came from 26 different cultural groups or nations and perceivers came from 44 different cultures. Culture was usually defined by the original authors as either country of origin or as the native language of expressers and perceivers.

Studies for inclusion were identified through previously published reviews on the topic (Elfenbein & Ambady, 2002; Juslin & Laukka, 2003; Laukka et al., 2016; Scherer et al., 2011). In addition, computer-based searches of databases (PubMed, Web of Knowledge, PsycArticles, IEEE Explore) were performed; search string = (speech OR voice OR vocal OR prosody) AND (emotion* OR affect*) AND cultur*.

Only studies that provided estimates of the accuracy of emotion recognition were included. This entails that studies had to include an objective criterion for accuracy. Typically, this criterion was whether perceivers recognized the emotion that the expresser was trying to convey. All included studies used expressions portrayed by actors, which means that the intended expression was known to the experimenters. One possible exception was a study by Chung (2000) that used material selected from television interviews. The majority of studies also used a forced-choice method wherein participants were asked to choose their response from a list of predefined response alternatives. However, a couple of studies that used rating scales or free responses were included because they also provided data on the proportion of accurate responses (Abelin & Allwood, 2000; Gendron, Roberson, van der Vyver, & Barrett, 2014; Huang, Erickson, & Akagi, 2008). The accuracy criterion excluded studies that investigated emotion recognition using dimensional ratings (e.g., activation, valence, emotion intensity, appraisal dimensions)—which do not provide measures of accuracy (e.g., Koeda et al., 2013; Nordström, Laukka, Thingujam, Schubert, & Elfenbein, 2017; Pfitzinger, Amir, Mixdorff, & Bösel, 2011).

We included any emotion that individual researchers considered worthy of inclusion in their experiments. However, we excluded judgments of personality traits, attitudes, and linguistic functions of prosody. In total, the included studies provided accuracy estimates for 25 different emotions: achievement, amusement, anger, awe, calmness, contempt, contentment, desire (for food), disgust, doubt, embarrassment, fear, happiness, interest, irony/sarcasm, lust, pain, pride, relief, sadness, shame, surprise, sympathy, tenderness, and neutral. The median number of emotions in the included studies was 5. For more details about how different labels were classified, see online supplemental Table S2.

Further, only studies that provided data from at least one within-cultural condition and at least one cross-cultural condition, were included. This could be achieved in two different ways. The same set of vocal stimuli could be judged both by perceivers from the same cultural group as the stimuli and by perceivers from another cultural group. We call this type of study the “many-on-one” design in Table S1. Alternatively, the same group of perceivers could judge stimuli both from expressers from their own culture and from a different cultural group. This type of design is labelled “one-on-many” in Table S1. Balanced designs (where vocal stimuli from at least two different cultures are judged by perceivers from each culture) provide data for both types of cultural comparisons. Studies that investigated emotion perception in cross-cultural conditions only, were not included because they do not provide data on in-group advantage (e.g., Laukka et al., 2013).

Only studies that investigated healthy adults (approximately 18–65 years) were included. In cases where studies provided data from several participant samples (e.g., from different age groups), we only made use of the data based on adults (e.g., Chronaki, Wigelsworth, Pell, & Kotz, 2018; McCluskey & Albas, 1981). Studies on clinical populations were not included.

Finally, we only included studies which utilized nondegraded vocal stimuli, either in the form of short sentences or words (speech prosody) or in the form of nonlinguistic vocalizations (e.g., laughter, crying, shrieks, grunts, sighs). Studies that used stimuli that were manipulated (e.g., content-filtered) or based on speech synthesis were not included (e.g., Yanushevskaya, Gobl, & Ní Chasaide, 2018), because stimulus manipulations may obscure cultural variation in expressive style. In cases where several types of stimuli were used in the same study, we only used data for nonmanipulated speech (e.g., Kramer, 1964).

Recorded Variables and Data Collection

The most common way of reporting data in emotion recognition studies is to present the proportion of correct responses (percentage accuracy). Therefore, percentage accuracy provides a natural and inclusive effect size index that is easy to interpret and can be used in the analysis of both cross-cultural emotion recognition and in-group advantage (Elfenbein & Ambady, 2002). The included 37 studies provided data on 191 pairwise comparisons between expresser and perceiver culture. We extracted percentage accuracy for each reported emotion and Expresser × Perceiver combination from each included study. One problem with comparing percentage accuracy across studies is that different studies use different numbers of response alternatives in the judgment task. Following Juslin and Laukka (2003), percentage accuracy values were therefore transformed into Rosenthal and Rubin’s (1989) proportion index (PI). PI values were also used to calculate a measure of in-group advantage (defined as the difference between within-cultural and cross-cultural accuracy) for each pairwise cultural comparison.

PI presents an effect size index for one-sample, multiple-choice data, and enables the transformation of accuracy scores involving any number of response alternatives into a standard scale of dichotomous choice, where .50 corresponds to chance-level responding and 1.00 corresponds to perfect recognition. In other words, in cases where there are more than two response alternatives, the raw proportion of correct responses is converted to the proportion of correct responses made if there had been only two response alternatives. PI is calculated using the formula PI = P(k − 1) / (1 + P(k − 2)), where P is the proportion of correct responses and k is the number of response alternatives (Rosenthal & Rubin, 1989). When calculating PI values for studies that used free responses to collect perceiver judgments, and thus did not have a fixed number of response alternatives, we used the number of expressed emotions as a proxy for the number of response alternatives.

The following variables were extracted from each study using data available in tables and figures: (a) characteristics of the sample (expresser culture, perceiver culture, number of expressers, number of perceivers); (b) study design (study design: balanced or not balanced; vocal stimulus type: speech prosody or nonlinguistic vocalizations; perception test type: forced-choice or rating scale/free responses; number of emotions; number of response alternatives); (c) descriptive statistics of participants’ performance for each combination of emotion and expresser and perceiver culture (percentage accuracy). For two of the included studies (Jiang, Paulmann, Robin, & Pell, 2015; Sauter, 2013), authors kindly provided percentage accuracy rates not included in the original publications.

In addition, we used Hofstede’s (2001) values for the dimensions power distance, individualism, masculinity, and uncertainty avoidance to calculate an index of cultural distance for each combination of expresser and perceiver culture. This was done by calculating the Euclidean distance between culture pairs in a four-dimensional space using cluster analysis. A handful of studies included participants from isolated cultural contexts to investigate emotion recognition across cultures with very little common exposure, such as mass media (Bryant & Barrett, 2008; Cordaro, Keltner, Tshering, Wangchuk, & Flynn, 2016; Gendron et al., 2014; Sauter, 2013; Sauter, Eisner, Ekman, & Scott, 2010). For these cultures, there were no available data regarding Hofstede’s dimensions. In order not to lose these valuable studies from the analysis, we conservatively gave comparisons including remote cultures the same index value as the largest difference that was observed among the included globalized cultures (which was the difference between Japan and The Netherlands).

Results

Table 1 displays cross-cultural recognition accuracy in terms of Rosenthal and Rubin’s (1989) proportion index (PI) for the most frequently investigated emotions, as well as overall performance averaged across positive and negative emotions. Results for the full set of 25 emotions are shown in online supplemental Table S3. All individual emotions were recognized with cross-cultural accuracy above chance, as indicated by 95% confidence intervals (CIs) not including .50 (which is the level of chance responding for PI), although it should be noted that for a few emotions (calmness, love/tenderness), the number of data points was too small for calculating CIs. Negative emotions were overall recognized with higher cross-cultural accuracy (M = 0.849, SD = 0.095) than positive emotions (M = 0.797, SD = 0.098), t(107) = 4.74, p < .001, d = 0.561.

Table 1.

Cross-cultural recognition accuracy and in-group advantage for selected emotions.

Cross-cultural accuracy (proportion index)							In-group advantage
Emotion	Mean	SD	−95% CI	+95% CI	Median	Number of cross-cultural comparisons	Emotion	Type of contrast	Mean	SD	−95% CI	+95% CI	Median	Number of cross-cultural comparisons
Overall^a	0.843	0.075	0.830	0.856	0.860	132	Overall	Many-on-one	0.055	0.067	0.043	0.067	0.039	119
								One-on-many	0.051	0.065	0.035	0.067	0.038	69
Positive emotions^b	0.797	0.098	0.778	0.815	0.809	110	Positive emotions	Many-on-one	0.071	0.091	0.052	0.089	0.046	97
								One-on-many	0.072	0.091	0.047	0.096	0.059	55
Negative emotions^c	0.849	0.095	0.831	0.867	0.878	109	Negative emotions	Many-on-one	0.052	0.084	0.035	0.069	0.035	98
								One-on-many	0.046	0.070	0.027	0.065	0.043	53
Anger	0.854	0.152	0.825	0.883	0.905	107	Anger	Many-on-one	0.058	0.119	0.034	0.082	0.029	96
								One-on-many	0.048	0.115	0.016	0.080	0.036	53
Contempt	0.720	0.128	0.678	0.762	0.711	38	Contempt	Many-on-one	0.053	0.121	0.012	0.093	0.065	37
								One-on-many	0.078	0.112	0.026	0.129	0.047	21
Disgust	0.726	0.327	0.599	0.853	0.852	28	Disgust	Many-on-one	0.086	0.195	0.005	0.166	0.031	25
								One-on-many	0.049	0.108	−0.024	0.121	0.032	11
Fear	0.837	0.114	0.813	0.861	0.871	90	Fear	Many-on-one	0.065	0.092	0.044	0.086	0.043	80
								One-on-many	0.066	0.092	0.041	0.092	0.042	52
Happiness	0.799	0.102	0.777	0.820	0.808	87	Happiness	Many-on-one	0.076	0.087	0.057	0.096	0.045	77
								One-on-many	0.074	0.096	0.046	0.101	0.055	50
Interest	0.853	0.066	0.830	0.877	0.862	32	Interest	Many-on-one	0.052	0.058	0.031	0.073	0.037	32
								One-on-many	0.037	0.068	0.005	0.069	0.036	20
Irony	0.665	0.158	0.578	0.753	0.702	15	Irony	Many-on-one	0.119	0.117	0.054	0.184	0.103	15
								One-on-many	0.103	0.184			0.103	2
Lust	0.824	0.138	0.780	0.869	0.853	40	Lust	Many-on-one	0.076	0.124	0.035	0.116	0.037	38
								One-on-many	0.057	0.093	0.018	0.096	0.036	24
Relief	0.813	0.164	0.757	0.870	0.854	35	Relief	Many-on-one	0.041	0.179	−0.023	0.104	0.006	33
								One-on-many	0.012	0.131	−0.043	0.067	−0.004	24
Sadness	0.890	0.103	0.870	0.910	0.909	101	Sadness	Many-on-one	0.029	0.102	0.007	0.050	0.009	90
								One-on-many	0.019	0.083	−0.004	0.042	0.012	53
Shame	0.700	0.098	0.657	0.743	0.739	22	Shame	Many-on-one	0.082	0.077	0.048	0.116	0.092	22
								One-on-many	0.066	0.135	0.003	0.129	0.101	20
Surprise	0.810	0.172	0.754	0.866	0.851	39	Surprise	Many-on-one	0.083	0.173	0.027	0.139	0.041	39
								One-on-many	0.056	0.086	−0.017	0.128	0.041	8
Neutral	0.885	0.083	0.864	0.906	0.907	62	Neutral	Many-on-one	0.053	0.056	0.038	0.069	0.044	55
								One-on-many	0.020	0.049	0.004	0.036	0.016	39

Note. ^aOverall recognition accuracy is based on all included emotion categories in each respective study. ^bPositive emotions included: achievement, amusement, awe, calm, contentment, desire (for food), happiness, interest, love, lust, pride, relief, and sympathy. ^cNegative emotions included: anger, contempt, doubt, disgust, embarrassment, fear, pain, sadness, and shame.

Regarding in-group advantage, recognition accuracy was overall higher in within-cultural conditions (M = 0.897, SD = 0.061) compared to cross-cultural conditions (M = 0.843, SD = 0.075; more details about within-cultural recognition rates in the current sample are available in Table S3). In-group advantage (defined as the difference in recognition accuracy between in-group and out-group comparisons) was overall significant for both “many-on-one” contrasts (M = 0.055, SD = 0.067, d = 0.821) and “one-on-many” contrasts (M = 0.051, SD = 0.065, d = 0.785), as indicated by 95% CIs not including zero (see Table 1). In-group advantage was significantly larger for positive emotions (M = 0.072, SD = 0.091) than for negative emotions (M = 0.046, SD = 0.070) for “one-on-many” contrasts, t(52) = 2.04, p = .046, d = 0.325. However, the difference between positive (M = 0.071, SD = 0.091) and negative (M = 0.052, SD = 0.084) in-group advantage was only marginally significant for “many-on-one” contrasts, t(96) = 1.96, p = .053, d = 0.219. Across all emotions, cross-cultural recognition accuracy was negatively correlated with in-group advantage for both “many-on-one” contrasts (r = −.74, p < .001, N = 119) and “one-on-many” contrasts (r = −.61, p < .001, N = 69), suggesting that in-group advantage was larger for expressions that were relatively harder to recognize.

We further explored differences in accuracy beyond the dichotomous status of in-group versus out-group, and examined to what extent cultural distance led to relatively lower accuracy. Overall recognition accuracy (including both within- and cross-cultural data points) was negatively correlated with cultural distance (r = −.37, p < .001, N = 186). A negative correlation was also observed between cross-cultural recognition accuracy (including only cross-cultural data points) and cultural distance (r = −.22, p = .014, N = 129). We also directly calculated the correlations between cultural distance and in-group advantage, and observed positive correlations for both “many-on-one” contrasts (r = .21, p = .026, N = 116) and “one-on-many” contrasts (r = .22, p = .072, N = 69), although the latter association did not reach conventional criteria for statistical significance. As a whole, these correlations suggest that emotion recognition accuracy decreases, and in-group advantage increases, as cultural distance between expresser and perceiver increases.

Finally, we investigated the effect of moderator variables (see Table 2). Large studies (as indicated by larger numbers of expressers, perceivers, and emotions) were associated with higher cross-cultural recognition rates and smaller in-group advantage. Balanced studies were further associated with smaller in-group advantage compared to unbalanced studies, and forced-choice studies were associated with smaller in-group advantage compared to studies using free responses and rating scales. No significant associations were found for stimulus type (speech prosody or nonlinguistic vocalization).

Table 2.

Correlations (Pearson r) between cross-cultural recognition accuracy and in-group advantage and moderator variables.

Moderator variable	Cross-cultural recognition accuracy	In-group advantage (many-on-one)	In-group advantage (one-on-many)
Stimulus type(0 = prosody, 1 = vocalizations)	.06N = 132p = .464	.04N = 119p = .689	.06N = 69p = .603
Study design type(0 = balanced, 1 = not balanced)	−.01N = 132p = .943	.17N = 119p = .058	.26 N = 69p = .028
Perception test type(0 = forced choice, 1 = rating scale/free responses)	−.13N =132p = .151	.23 N = 119p = .011	n/a
N expressers	.13N = 117p = .179	−.27 N = 108p = .005	−.21N = 63p = .091
N perceivers	.30 N = 130p < .001	−.27 N = 117p = .003	−.30 N = 67p = .014
N emotions	.21 N = 132p = .017	−.18N = 119p = .052	−.32 N = 69p = .007

Note. Boldfaced results indicate p < .05.

Discussion

In order to supplement a large volume of research on facial expression, the goal of this study was to document the state of the evidence for cross-cultural similarities and differences in emotion expressed via the voice. Drawing from dialect theory, we hypothesized and found that emotion expressed through the voice was recognized across cultures at levels substantially above what would be expected due to chance guessing alone. However, we also found evidence for cultural differences in the form of in-group advantage, in that vocal emotion was recognized more accurately when expressers and perceivers were from the same cultural group. Further, this in-group advantage was larger the greater cultural distance the expresser and perceiver groups had from each other. Taken together, the results of this study support the dialect theory of emotion (Elfenbein, 2013; Elfenbein & Ambady, 2002).

Although individual studies usually included a fairly limited number of emotions, the aggregated results suggest that a rich repertoire of emotions—not limited to the usual six or so basic emotions, and including several positive emotions beyond happiness—can be recognized cross-culturally with accuracy well above chance. Indeed, out of 25 included emotions, all were recognized with better than chance accuracy. (Note that for two emotions, there were not enough data to calculate interval estimates.) In terms of Rosenthal and Rubin’s (1989) PI, our estimate of overall cross-cultural accuracy (M = 0.843) was very similar to the one obtained in a previous meta-analysis (see Juslin & Laukka, 2003)—although the current meta-analysis included 5 times more studies and emotions. A PI value of 0.84 corresponds to a raw accuracy score of 0.57 in a forced-choice task with five response options (which was the median number among the included studies), which in turn is approximately 3 times higher than the 1 in 5 rate that would be expected by chance responding. Based on the current evidence, this seems to be a fairly robust estimate of average cross-cultural accuracy for vocally expressed emotions.

In terms of cultural differences, our estimate of overall in-group advantage (M ≈ 0.05) was also very similar to that reported by Juslin and Laukka (2003). Vocal expressions are considered to be at least partly based on appraisal and emotion-related physiological changes (Scherer, 1986), and we have previously argued that this may prevent drift of expressive styles past the point of mutual unintelligibility (Laukka et al., 2016). Accordingly, overall in-group advantage was not large in terms of proportion of correct responses, although it was systematic enough that it did represent a large effect size (d ≈ 0.80).

Results further underscore that accuracy and in-group advantage are not static phenomena, but vary across cultural conditions. Dialect theory proposes that differences in expressive style grow as groups become more culturally distant, resulting in less accurate cross-cultural communication. We accordingly observed that increased cultural distance was associated with decreased accuracy and increased in-group advantage. Although effect sizes were only small to medium, these findings demonstrate the benefits of meta-analysis, because it has been difficult for individual studies to obtain samples with enough cultural variability to investigate this issue (e.g., Laukka et al., 2016; Scherer, Banse, & Wallbott, 2001).

In-group advantage also seemed to vary across emotions. The number of observations was rather small for several emotions (see Table S3), which makes it difficult to draw conclusions about the significance of in-group advantage for individual emotions. However, we note that among the most frequently investigated ones (presented in Table 1), relief and sadness did not show consistent evidence for in-group advantage. Interestingly, sadness had small in-group advantage for vocal expression in both Elfenbein and Ambady (2002) and in Juslin and Laukka (2003). Results further suggest that in-group advantage may be larger for expressions with positive versus negative affective valence. This observation can be interpreted in terms of the functions of positive and negative emotions. Sauter et al. (2010) suggested that communication of positive emotions facilitates social cohesion with group members, but that such affiliative signals may not always be shared with nonmembers. Signals of negative emotions may instead be more closely related to biological reactions to various threats, and less influenced by cultural learning. On a more general level, in-group advantage was strongly negatively correlated with cross-cultural recognition accuracy. We speculate that knowledge of cultural differences in expressive style may become more important when judging expressions that are more ambiguous. Clearly, more research is needed to understand how and why in-group advantage varies across emotions.

Results are in line with dialect theory, but do not provide direct evidence for the mechanisms of the theory. However, we and others have recently tested propositions of dialect theory more directly. For example, Laukka, Neiberg, and Elfenbein (2014) used speech stimuli from Australia, India, Kenya, Singapore, and the United States, and trained acoustics-based classifier programs to recognize emotional expressions. For each model, training was based on stimuli from one cultural group, and recognition was based either on stimuli from the same group on which it was trained or on stimuli from a different group. Accuracy was at greater than chance levels regardless of whether classifiers were trained and tested on stimuli from the same versus different cultures. This provides evidence for basic universality in acoustic patterns. However, accuracy was higher when classifiers were trained and tested on stimuli from the same culture versus different cultures. This, in turn, provides direct evidence for systematic cultural differences in expressive style.

In another study, we examined patterns of fundamental acoustic properties in emotional expression within and across cultural groups (Laukka et al., 2016). It was found that acoustic cues were used in a relatively similar way across groups both to produce and judge emotions, but there were also subtle cultural differences. Findings thus suggested that expressers appear to have culturally nuanced schemas for enacting vocal expressions, and perceivers appear to have culturally nuanced schemas for judging expressions. Consistent with dialect theory, in-group judgments showed a greater match between these schemas used for emotional expression and perception.

As a final example, Sauter (2013) showed that in-group advantage was observed when perceivers judged vocalizations from their own cultural group, even though they were unable to reliably infer the group membership of the expresser. Taken together, these studies speak against perceiver biases—such as increased motivation for judging in- versus out-group expressions (e.g., Thibault, Bourgeois, & Hess, 2006) or possible effects of decoding rules on emotion inferences (e.g., Matsumoto, 2002)—as the main origin of in-group advantage. Instead, they suggest strongly that the mechanism is subtle differences in expressive style.

Like any meta-analysis, we are limited by the availability of original empirical work. As such, this quantitative review also exposes several important gaps in the literature. Notably, nearly all data points included in the meta-analysis were based on recognition of expressions portrayed by actors. This means that studies using spontaneous expressions will be an important next step, especially because recent research has revealed small but systematic differences in the acoustic patterns between posed and spontaneous expressions (e.g., Juslin, Laukka, & Bänziger, 2018). The majority of included studies also used a forced-choice methodology with a relatively small number of response alternatives, which has been criticized to artificially inflate recognition rates and impose categorical interpretations on affective stimuli (e.g., Russell, 1994). Our analysis of moderator variables suggested that forced-choice studies were associated with smaller in-group advantage compared to studies using other response formats (see Table 2), but this finding is limited by the very small number of studies using other formats included in the current analysis. Measures that account for the possibility that systematic biases in participants’ use of forced-choice response alternatives could inflate apparent accuracy, have been proposed (see e.g., Wagner, 1993), and future cross-cultural work could report such sensitivity measures in addition to traditional accuracy rates. In a broader perspective, we believe that studies that combine forced-choice methods with open-ended responses and continuous scale ratings will be important for making progress in understanding cultural similarities and differences with regard to how perceivers infer meaning from vocal expressions (Cowen, Laukka, Elfenbein, Liu, & Keltner, 2019).

We did not find any effects of type of stimulus (speech prosody or nonlinguistic vocalizations) on cross-cultural recognition accuracy. This is in contrast to several within-cultural studies proposing higher recognition rates for vocalizations compared to speech prosody (e.g., Hawk, van Kleef, Fischer, & van der Schalk, 2009). However, the number of cross-cultural studies on vocalizations is currently rather small, which calls for more cross-cultural research that directly compares different types of vocal expressions. Such studies could preferably employ balanced designs, which are especially useful for evaluating in-group advantage. A balanced design can control for possible extraneous main effects across expresser and perceiver groups that are not related to culture, and test cultural differences in the form of an interaction effect. As more balanced cross-cultural studies become available, future meta-analyses could employ these interaction effects as a measure of effect size.

Finally, we used Hofstede (2001) as the source for calculating our index of cultural distance, but there may be other ways of measuring cultural distance that may capture aspects that are more relevant for nonverbal behavior, and we welcome more research that tries to link various measures of cultural differences with emotion recognition.

We conclude this review by noting that, in contrast with hundreds of studies on facial expressions, the collection of 37 studies of vocal emotion that we located is a far smaller number—but worthwhile in the corroborating evidence it provides. Taken together, the evidence reveals that perceivers can recognize emotional expressions via nonverbal behavior across cultural divides, and yet they do so better when staying within their own cultural boundaries.

Supplemental Material

LaukkaandElfenbein_meta-analysis_suppl_material_R2 – Supplemental material for Cross-Cultural Emotion Recognition and In-Group Advantage in Vocal Expression: A Meta-Analysis

Supplemental material, LaukkaandElfenbein_meta-analysis_suppl_material_R2 for Cross-Cultural Emotion Recognition and In-Group Advantage in Vocal Expression: A Meta-Analysis by Petri Laukka and Hillary Anger Elfenbein in Emotion Review

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Marianne and Marcus Wallenberg Foundation (MMW 2018.0059).

ORCID iD

Petri Laukka

Supplemental Material

Supplemental material for this article is available online.

References

*Abelin

Å.

Allwood

. (2000). Cross linguistic interpretation of emotional prosody. In Cowie

Douglas-Cowie

Schröder

(Eds.), Proceedings of the ISCA Workshop on Speech and Emotion (pp. 110–113). Belfast, Northern Ireland: International Speech Communication Association.

*Albas

D. C.

McCluskey

K. W.

Albas

C. A.

(1976). Perception of the emotional content of speech: A comparison of two Canadian groups. Journal of Cross-Cultural Psychology, 7, 481–490. http://doi.org/10.1177/002202217674009

*Altrov

Pajupuu

(2015). The influence of language and culture on the understanding of vocal emotions. Journal of Estonian and Finno-Ugric Linguistics, 6(3), 11–48. http://doi.org/10.12697/jeful.2015.6.3.01

*Bailey

Nowicki

Cole

S. P.

(1998). The ability to decode nonverbal information in African American, African and Afro-Caribbean, and European American adults. Journal of Black Psychology, 24, 418–431. http://doi.org/10.1177/00957984980244002

*Braun

Heilmann

C. M.

(2012). SynchronEmotion. Bern, Switzerland: Peter Lang.

*Bryant

G. A.

Barrett

H. C.

(2008). Vocal emotion recognition across disparate cultures. Journal of Cognition and Culture, 8, 135–148. http://doi.org/10.1163/156770908X289242

*Chronaki

Wigelsworth

Pell

M. D.

Kotz

S. A.

(2018). The development of cross-cultural recognition of vocal emotion during childhood and adolescence. Scientific Reports, 8. http://doi.org/10.1038/s41598-018-26889-1

*Chung

S.-J.

(2000). L’expression et la perception de l’émotion extraite de la parole spontanée: Evidences du coréen et de l’anglais [Expression and perception of emotion extracted from spontaneous speech in Korean and English] (Unpublished doctoral dissertation). Université Paris III – Sorbonne Nouvelle, Paris, France.

*Cordaro

D. T.

Keltner

Tshering

Wangchuk

Flynn

L. M.

(2016). The voice conveys emotion in ten globalized cultures and one remote village in Bhutan. Emotion, 16, 117–128. http://doi.org/10.1037/emo0000100

10.

Cowen

A. S.

Elfenbein

H. A.

Laukka

Keltner

(2019). Mapping 24 emotions conveyed by brief human vocalization. American Psychologist, 74, 698–712. http://doi.org/10.1037/amp0000399

11.

Cowen

A. S.

Laukka

Elfenbein

H. A.

Liu

Keltner

(2019). The primacy of categories in the recognition of 12 emotions in speech prosody across two cultures. Nature Human Behaviour, 3, 369–382. http://doi.org/10.1038/s41562-019-0533-6

12.

*Davitz

J. R.

(1964). Minor studies and some hypotheses. In Davitz

J. R.

(Ed.), The communication of emotional meaning (pp. 143–156). New York, NY: McGraw-Hill.

13.

Ekman

(1972). Universals and cultural differences in facial expressions of emotion. In Cole

(Ed.), Nebraska Symposium on Motivation, 1971 (Vol. 19, pp. 207–282). Lincoln: University of Nebraska Press.

14.

Elfenbein

H. A.

(2013). Nonverbal dialects and accents in facial expressions of emotion. Emotion Review, 5, 90–96. http://doi.org/10.1177/1754073912451332

15.

Elfenbein

H. A.

Ambady

(2002). On the universality and cultural specificity of emotion recognition: A meta-analysis. Psychological Bulletin, 128, 203–235. http://doi.org/10.1037/0033-2909.128.2.203

16.

*Frye

C. I.

Creel

S. C.

(2013). Emotional speech processing and language knowledge. In Knauff

Pauen

Sebanz

Wachsmuth

(Eds.), Proceedings of the 35th Annual Meeting of the Cognitive Science Society (pp. 2333–2337). Austin, TX: Cognitive Science Society.

17.

*Gendron

Roberson

van der Vyver

J. M.

Barrett

L. F.

(2014). Cultural relativity in perceiving emotion from vocalizations. Psychological Science, 25, 911–920. http://doi.org/10.1177/0956797613517239

18.

*Graham

C. R.

Hamblin

Feldstein

(2001). Recognition of emotion in English voices by speakers of Japanese, Spanish, and English. International Review of Applied Linguistics in Language Teaching, 39, 19–37. http://doi.org/10.1515/iral.39.1.19

19.

Hawk

S. T.

van Kleef

G. A.

Fischer

A. H.

van der Schalk

(2009). “Worth a thousand words”: Absolute and relative decoding of nonlinguistic vocalizations. Emotion, 9, 293–305. http://doi.org/10.1037/a0015178

20.

Hofstede

(2001). Culture’s consequences: Comparing values, behaviors, institutions, and organizations across nations (2nd ed.). Thousand Oaks, CA: SAGE.

21.

*Huang

C.-F.

Erickson

Akagi

(2008). Comparison of Japanese expressive speech perception by Japanese and Taiwanese listeners. In Proceedings of Acoustics ‘08 Paris (pp. 2317–2322). Paris, France: Société Française d’Acoustique.

22.

Izard

C. E.

(1971). The face of emotions. New York, NY: Appleton-Century-Crofts.

23.

*Jiang

Paulmann

Robin

Pell

M. D.

(2015). More than accuracy: Nonverbal dialects modulate the time course of vocal emotion recognition across cultures. Journal of Experimental Psychology: Human Perception and Performance, 41, 597–612. http://doi.org/10.1037/xhp0000043

24.

*Jürgens

Drolet

Pirow

Scheiner

Fischer

(2013). Encoding conditions affect recognition of vocally expressed emotions across cultures. Frontiers in Psychology, 4. http://doi.org/10.3389/fpsyg.2013.00111

25.

Juslin

P. N.

Laukka

(2003). Communication of emotions in vocal expression and music performance: Different channels, same code? Psychological Bulletin, 129, 770–814. http://doi.org/10.1037/0033-2909.129.5.770

26.

Juslin

P. N.

Laukka

Bänziger

(2018). The mirror to our soul? Comparisons of spontaneous and posed vocal expression of emotion. Journal of Nonverbal Behavior, 42, 1–40. http://doi.org/10.1007/s10919-017-0268-x

27.

Koeda

Belin

Hama

Masuda

Matsuura

Okubo

(2013). Cross-cultural differences in the processing of non-verbal affective vocalizations by Japanese and Canadian listeners. Frontiers in Psychology, 4. http://doi.org/10.3389/fpsyg.2013.00105

28.

*Kramer

(1964). Elimination of verbal cues in judgments of emotion from voice. Journal of Abnormal and Social Psychology, 68, 390–396. http://doi.org/10.1037/h0042473

29.

Laukka

Elfenbein

H. A.

Söder

Nordström

Althoff

Chui

. . .Thingujam

N. S.

(2013). Cross-cultural decoding of positive and negative non-linguistic emotion vocalizations. Frontiers in Psychology, 4. http://doi.org/10.3389/fpsyg.2013.00353

30.

*Laukka

Elfenbein

H. A.

Thingujam

N. S.

Rockstuhl

Iraki

F. K.

Chui

Althoff

(2016). The expression and recognition of emotions in the voice across five nations: A lens model analysis based on acoustic features. Journal of Personality and Social Psychology, 111, 686–705. http://doi.org/10.1037/pspi0000066

31.

Laukka

Neiberg

Elfenbein

H. A.

(2014). Evidence for cultural dialects in vocal emotion expression: Acoustic classification within and across five nations. Emotion, 14, 445–449. http://doi.org/10.1037/a0036048

32.

*Mandal

M. K.

(2008). Cultural in-group advantage in accuracy at recognizing vocal expressions of emotion. Psychological Studies, 53, 126–132.

33.

Matsumoto

(2002). Methodological requirements to test a possible in-group advantage in judging emotions across cultures: Comment on Elfenbein and Ambady (2002) and evidence. Psychological Bulletin, 128, 236–243. http://doi.org/10.1037/0033-2909.128.2.236

34.

*McCluskey

K. W.

Albas

D. C.

(1981). Perception of the emotional content of speech by Canadian and Mexican children, adolescents, and adults. International Journal of Psychology, 16, 119–132. http://doi.org/10.1080/00207598108247409

35.

*Mejvaldová

(2000). Expressions prosodiques de certaines attitudes en tchèque et en français [Prosodic expressions of certain attitudes in Czech and French]. In Bailly

Badin

(Eds.), Actes des 23èmes Journées d’Etudes sur la Parole [Proceedings of JEP 2000] (pp. 253–256). Aussois, France: Association Francophone de la Communication Parlée.

36.

*Nakamichi

Jogan

Usami

Erickson

(2003). Perception by native and non-native listeners of vocal emotion in a bilingual movie. Bulletin of Gifu City Women’s College, 52, 87–91.

37.

Nordström

Laukka

Thingujam

N. S.

Schubert

Elfenbein

H. A.

(2017). Appraisal dimensions inferred from vocal emotion expressions are consistent across cultures: A comparison between Australia and India. Royal Society Open Science, 4. http://doi.org/10.1098/rsos.170912

38.

O’Grady

Archibald

Aronoff

Rees-Miller

(2001). Contemporary linguistics (4th ed.). Boston, MA: Bedford/St. Martin’s.

39.

*Paulmann

Uskul

A. K.

(2014). Cross-cultural emotional prosody recognition: Evidence from Chinese and British listeners. Cognition and Emotion, 28, 230–244. http://doi.org/10.1080/02699931.2013.812033

40.

*Pell

M. D.

Monetta

Paulmann

(2009). Recognizing emotions in a foreign language. Journal of Nonverbal Behavior, 33, 107–120. http://doi.org/10.1007/s10919-008-0065-7

41.

Pfitzinger

H. R.

Amir

Mixdorff

Bösel

(2011). Cross-language perception of Hebrew and German authentic emotional speech. In Lee

W.-S.

Zee

(Eds.), Proceedings of the 17th International Congress of Phonetic Sciences (pp. 1586–1589). Hong Kong: City University of Hong Kong.

42.

*Riviello

M. T.

Esposito

(2016). On the perception of dynamic emotional expressions: A cross-cultural comparison. Berlin, Germany: Springer.

43.

Rosenthal

Rubin

D. B.

(1989). Effect size estimation for one-sample multiple-choice-type data: Design, analysis, and meta-analysis. Psychological Bulletin, 106, 332–337. http://doi.org/10.1037/0033-2909.106.2.332

44.

Russell

J. A.

(1994). Is there universal recognition of emotion from facial expression? A review of the cross-cultural studies. Psychological Bulletin, 115, 102–141. http://doi.org/10.1037/0033-2909.115.1.102

45.

*Sauter

D. A.

(2013). The role of motivational and cultural dialects in the in-group advantage for emotional vocalizations. Frontiers in Psychology, 4. http://doi.org/10.3389/fpsyg.2013.00814

46.

*Sauter

D. A.

Eisner

Ekman

Scott

S. K.

(2010). Cross-cultural recognition of basic emotions through nonverbal emotional vocalizations. Proceedings of the National Academy of Sciences of the United States of America, 107, 2408–2412. http://doi.org/10.1073/pnas.0908239106

47.

Scherer

K. R.

(1986). Vocal affect expression: A review and a model for future research. Psychological Bulletin, 99, 143–165. http://doi.org/10.1037/0033-2909.99.2.143

48.

*Scherer

K. R.

Banse

Wallbott

H. G.

(2001). Emotion inferences from vocal expression correlate across languages and cultures. Journal of Cross-Cultural Psychology, 32, 76–98. http://doi.org/10.1177/0022022101032001009

49.

Scherer

K. R.

Clark-Polner

Mortillaro

(2011). In the eye of the beholder? Universality and cultural specificity in the expression and perception of emotion. International Journal of Psychology, 46, 401–435. http://doi.org/10.1080/00207594.2011.626049

50.

*Shigeno

(2009). Recognition of vocal and facial emotions: Comparison between Japanese and North Americans. In Izdebski

(Ed.), Emotions in the human voice. Vol II. Culture and perception (pp. 187–204). San Diego, CA: Plural.

51.

*Shigeno

(2013). Multimodal perception of emotion in Japanese and Chinese. Psychology Research, 3(9), 504–511. http://doi.org/10.17265/2159-5542/2013.09.003

52.

*Shochi

Rilliard

Aubergé

Erickson

(2009). Intercultural perception of English, French and Japanese social affective prosody. In Hancil

(Ed.), The role of prosody in affective speech (pp. 31–60). Bern, Switzerland: Peter Lang.

53.

Thibault

Bourgeois

Hess

(2006). The effect of group-identification on emotion recognition: The case of cats and basketball players. Journal of Experimental Social Psychology, 42, 676–683. http://doi.org/10.1016/j.jesp.2005.10.006

54.

*Thompson

W. F.

Balkwill

L.-L.

(2006). Decoding speech prosody in five languages. Semiotica, 158, 407–424. http://doi.org/10.1515/SEM.2006.017

55.

*Tickle

(2000). English and Japanese speakers’ emotion vocalization and recognition: A comparison highlighting vowel quality. In Cowie

Douglas-Cowie

Schröder

(Eds.), Proceedings of the ISCA Workshop on Speech and Emotion (pp. 104–109). Belfast, Northern Ireland: International Speech Communication Association.

56.

*Van Bezooijen

Otto

S. A.

Heenan

T. A

. (1983). Recognition of vocal expressions of emotion: A three-nation study to identify universal characteristics. Journal of Cross-Cultural Psychology, 14, 387–406. http://doi.org/10.1177/0022002183014004001

57.

*Waaramaa

(2015). Perception of emotional nonsense sentences in China, Egypt, Estonia, Finland, Russia, Sweden, and the USA. Logopedics Phoniatrics Vocology, 40, 129–135. http://doi.org/10.3109/14015439.2014.915982

58.

Wagner

H. L.

(1993). On measuring performance in category judgment studies of nonverbal behavior. Journal of Nonverbal Behavior, 17, 3–28. http://doi.org/10.1007/BF00987006

59.

*Wickline

V. B.

Bailey

Nowicki

(2009). Cultural in-group advantage: Emotion recognition in African American and European American faces and voices. Journal of Genetic Psychology, 170, 5–29. http://doi.org/10.3200/GNTP.170.1.5-30

60.

Yanushevskaya

Gobl

Ní Chasaide

(2018). Cross-language differences in how voice quality and f₀ contours map to affect. Journal of the Acoustical Society of America, 144, 2730–2750. http://doi.org/10.1121/1.5066448

61.

*Zhu

(2013). Which is the best listener group? Perception of Chinese emotional prosody by Chinese natives, naïve Dutch listeners and Dutch L2 learners of Chinese. Dutch Journal of Applied Linguistics, 2, 170–183. http://doi.org/10.1075/dujal.2.2.03zhu

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB