The Same with Age: Evidence for Age-related Similarities in Interpersonal Accuracy

Abstract

Interpersonal accuracy refers to the ability to make accurate perceptions about others’ social and emotional qualities. Despite this broad definition, the measurement of interpersonal accuracy remains narrow, as most studies focus on the accurate perception of others’ emotional states. Moreover, previous research has relied primarily upon traditional tasks consisting of posed, prototypic expressions and behaviors as stimuli. These methodological limitations may constrain our understanding of how different interpersonal perception skills change in adulthood. The present study investigated the extent to which various interpersonal perception skills are worse, better, or remain the same with age using both traditional and nontraditional interpersonal accuracy tasks. One hundred fifty-one adults from three age groups (young, middle age, and older) completed a battery of interpersonal accuracy tasks that assessed eight different emotion perception skills and six different social perception skills. Analyses revealed age-related differences in accuracy for five interpersonal perception skills; differences were typically observed between younger and older adults on emotion perception accuracy and between younger and middle-age adults on social perception accuracy. In contrast, almost all remaining interpersonal perception skills—both emotional and social—revealed greater evidence for age-related similarities than differences in Bayesian analyses. Additional exploratory analyses indicated that the observed age differences in interpersonal accuracy may be due to individual differences in cognitive ability rather than age. Results provide a nuanced picture of how interpersonal perception skills change in adulthood and provide new methodological tools for a more complete and comprehensive assessment of interpersonal accuracy.

Interpersonal accuracy is a broad umbrella term used to refer to the ability to make accurate social and emotional judgments about others; this definition thus includes all the different types of interpersonal perception skills an individual might use when making judgments about others during an interaction. For example, when interacting with a social partner, we might try to accurately identify how positive or negative they feel from their verbal or nonverbal expressions (Hareli & Rafaeli, 2008; Keltner & Haidt, 1999; Riggio, & Reichard, 2008). We might also make judgments about the genuineness of these expressions (Gosselin, Beaupré, & Boissonneault, 2002, Porter & ten Brinke, 2008) or try to determine if our social partner is trying to deceive us (deTurck, Harszlak, Bodhorn, & Texter, 1990; Talwar, Crossman, Williams, & Muir, 2011; ten Brinke & Porter, 2012). We can even try to identify other non-emotional qualities about our social partner, including the extent to which he or she displays different personality traits (like extraversion or neuroticism) or is healthy, powerful, or competent (Cuddy, Glick, & Beninger, 2011; Kalick, Zebrowitz, Langlois, & Johnson, 1998; Mast & Hall, 2004; Naumann, Vazire, Rentfrow, & Gosling, 2009; Roter, Hall, & Katz, 1987; Scherer, 1978). These examples—drawn from diverse domains in psychology—serve to illustrate the corpora of perceptions one might make about others in everyday interactions.

With these examples in mind, it is not difficult to imagine the utility of interpersonal accuracy for social functioning and wellbeing. Numerous studies have shown that individuals with greater interpersonal perception skill also exhibit enhanced functioning in school, relational, and work settings (e.g., Bechtoldt, Rohrmann, De Pater, & Beersma, 2011; Doerwald, Scheibe, Zacher, & Van Yperen, 2016; Hall, Andrzejewski, & Yopchick, 2009; Izard et al., 2001). Moreover, an inability to accurately judge others’ feelings, thoughts, and behaviors is often cited as a symptom of serious psychological disorders, including autism (Baron-Cohen, O’Riordan, Stone, Jones, & Plaisted, 1999), schizophrenia (Mueser et al., 1996), borderline personality disorder (Domes, Schulze, & Herpertz, 2009), and Alzheimer’s Disease (Phillips, Scott, Henry, Mowat, & Bell, 2010; Weiss et al., 2008). Interpersonal accuracy is thus likely to remain an important competence throughout the lifespan, as individuals continue to engage in interactions that require skills in making accurate social and emotional judgments about others.

Consequently, it is both surprising and worrisome that aging is negatively associated with interpersonal accuracy. Specifically, across diverse samples and methods, older adults are significantly less accurate at identifying how other people feel than younger adults (e.g., Circelli, Clark, & Cronin-Golomb, 2013; Ebner & Johnson, 2009; Isaacowitz et al., 2007; Lambrecht, Kreifelts, & Wildgruber, 2012; Laukka & Juslin, 2007; Lima, Alves, Scott, & Castro, 2014; Mill, Allik, Realo, & Valk, 2009; Mitchell, Kingston, & Barbosa Bouças, 2011; Paulmann, Pell, & Kotz, 2008; Spencer, Sekuler, Bennett, Giese, & Pilz, 2016; Sullivan & Ruffman, 2004). This age difference is typically moderate in size (i.e., .2 to .7 Cohen’s d; Ruffman, Henry, Livingstone, & Phillips, 2008), underscoring the robustness of the effect. Moreover, this effect may be linear (Calder et al., 2003), with some evidence of similar differences in interpersonal accuracy between middle-age and younger adults (e.g., Ekman & O’Sullivan, 1991; Isaacowitz et al., 2007; Malatesta, Izard, Culver, & Nicolich, 1987; Mill et al., 2009).

If knowing how others feel is an integral component of social functioning and wellbeing across the lifespan, then how are older adults able to function socially and emotionally alongside declines in this skill? Findings from many studies provide evidence that older adults maintain functioning in select socioemotional domains, including empathy (Grühn, Rebucal, Diehl, Lumley, & Labouvie-Vief, 2008), emotion regulation (Gross et al., 1997), and well-being (Diener, Suh, Lucas, & Smith, 1999; Smith, Fleeson, Geiselmann, Settersten, & Kunzmann, 1999). Consequently, one possibility is that older adults maintain interpersonal perception skills, albeit not those typically studied in the laboratory.

Methodological Limitations of Traditional Interpersonal Accuracy Tasks

The study of interpersonal accuracy is rooted in social psychology. Thus, much of what we know about interpersonal perception skills—including if and how skills changes with age—is based on methods derived from social psychology research. For example, a traditional laboratory task involves presenting participants with a standard set of stimuli (often photographs) depicting actors who were instructed to pose discrete emotional behaviors or cues in their faces, voices, and/or bodies (e.g., Hess, Blairy, & Kleck, 1997; Matsumoto et al., 2000; Nowicki & Duke, 1994; Scherer & Scherer, 2011; Schlegel, Grandjean, & Scherer, 2014; Tottenham et al., 2009). Participants are then typically asked to select from a predetermined list of emotion labels the label that best matches how the actor feels (or rather, was told to show), and responses are scored and summed for accuracy (Hall & Bernieri, 2001).

Such tasks, although common, are limited in the type of content being judged, both with regard to the nature of the stimuli depicting this content and in the scope of the content depicted. For example, traditional tasks rely primarily on posed, prototypic behaviors despite accumulating evidence that such behaviors are more spontaneous (i.e., not posed or scripted) and less prototypic in everyday life. To illustrate, most studies that measure the accurate perception of emotion utilize stimuli that depict posed configurations of facial movements thought to convey discrete emotional states (i.e., prototypic emotional facial expressions; Ekman, Friesen, Hager, 2002). Yet, many findings suggest that such configurations are not frequently produced in real life situations, including situations that elicit highly intense emotions (e.g., bull fighting or sexual intercourse; García-Higuera, Crivelli, & Fernández-Dols, 2015; Fernández-Dols, Carrera, & Crivelli, 2011) and situations that are more emotionally subtle (e.g., a conflict discussion between family members; Castro, Camras, Halberstadt, & Shuster, 2017).

Additionally, traditional tasks tend to focus narrowly on emotion perception skills: The vast majority of the literature is comprised of studies that measure individual differences in the ability to identify others’ emotions, with fewer studies investigating the ability to identify other non-emotional qualities. Yet, as illustrated in the first paragraph, knowing how others feel is only one among countless types of interpersonal perception skills an individual may employ during daily life. The extent to which different types of interpersonal perception skills are associated within individuals is modest at best and seems to depend in part upon the methodology, including the nature of the stimuli (Schlegel, Boone, & Hall, 2017), further suggesting a need to widen the scope of interpersonal accuracy measurement in the laboratory.

A third limitation surrounds the cognitive skills implicated in traditional interpersonal accuracy tasks. Although recent efforts have been made to include more dynamic stimuli in traditional laboratory tasks (e.g., Schlegel et al., 2014), most studies in interpersonal perception continue to rely on static images that are presented for brief amounts of time (i.e., a few seconds). Even videos, which by nature include more information than photos, are often kept similarly brief so as to reduce participant burden. Because interpersonal accuracy can be influenced by the length of stimulus exposure (Carney, Colvin, & Hall, 2007; Matsumoto et al., 2000), some types of judgments may simply require more information (and thus, more exposure time) to achieve adequate levels of accuracy than others. Moreover, brief (i.e., a few seconds) assessments of behavior may place high cognitive demands on perceivers, who must process relevant information quickly as it is presented (and before it disappears) to facilitate a correct response. This process may be especially problematic for judgments of behaviors or qualities that do require greater exposure for accuracy, as perceivers may have little relevant information from which to make accurate judgments. Consequently, perceptions that are made from static behaviors and/or brief (i.e., a few seconds) stimuli presentations may rely more on attentional and information processing skills than on skills relating to the accurate perception and interpretation of socioemotional behavior. Indeed, recent studies suggest a positive link between basic information processing skills and accuracy on various interpersonal perception tasks (e.g., Castro & Boone, 2015; Schlegel, Witmer, & Rammsayer, 2017; for meta-analysis, see Murphy & Hall, 2011).

Relevance of Limitations for Research on Aging

Despite these methodological limitations, traditional tasks remain the norm in the interpersonal accuracy literature. However, these limitations are particularly problematic for studies that seek to measure interpersonal accuracy across the adult lifespan.

Although age differences in interpersonal accuracy appear to be robust, evidence from the cognitive and emotional aging literatures suggests that older adults perform better on laboratory tasks when stimuli are more ecologically valid and representative of everyday life than when stimuli lack everyday relevance and context (e.g., Hess, Rosenberg, & Waters, 2001; Stanley & Isaacowitz, 2015; Sze, Goodkind, Gyurak, & Levenson, 2012). Consequently, it is possible that traditional tasks exacerbate age differences in interpersonal accuracy by suppressing older adults’ skills. Methods that utilize spontaneous and dynamic rather than posed and static stimuli are likely to yield a more accurate assessment of aging individuals’ interpersonal perception skills (Phillips & Slessor, 2011). There is some preliminary evidence that age differences in interpersonal accuracy are reduced (albeit, not fully eliminated) when stimuli are spontaneous and/or dynamic (e.g., Holland, Ebner, Lin, & Samanez-Larkin, 2018; Krendl & Ambady, 2010; Murphy, Lehrfeld, & Isaacowitz, 2010; Sze et al. 2012); however, other findings challenge this pattern and continue to find age differences in the perception of spontaneous stimuli (e.g., Grainger, Henry, Phillips, Vanman, & Allen, 2015; Malatesta et al., 1987; Phillips, Allen, Bull, Hering, Kliegel, & Channon, 2015), suggesting a need for more research.

Moreover, most studies that focus on emotion perception accuracy—which as noted above constitute the majority of the literature—utilize a “basic emotions” approach whereby joy is the single positive emotion examined alongside several negative emotions (i.e., anger, sadness, fear, and disgust). Although individuals do make judgments about others’ negative emotions, recent evidence suggests that judgments about positive emotions may be more relevant—for example, more important and frequent in everyday life—for adults of all ages (Castro & Isaacowitz, 2018). Importantly, age differences in the perception of positive emotion (typically examined as the perception of joy/happiness) are often nonsignificant and very small in size (i.e., close to zero) (e.g., Blanke, Rauers, & Riediger, 2014; Di Domenico, Palumbo, Mammarella, & Fairfield, 2015; Isaacowitz et al., 2007; Ruffman et al., 2008), supporting the need to consider a wider range of positive emotions in interpersonal accuracy and aging research. Given evidence that stimulus relevance is associated with older adults’ social and emotional functioning in the laboratory (e.g., Kunzmann & Grühn, 2005; Richter & Kunzmann, 2011; Streubel & Kunzmann, 2011), it is especially important that laboratory tasks be adapted to include more relevant interpersonal perception skills.

In addition, because traditional tasks tend to focus so narrowly on emotion perception accuracy, and because different interpersonal perception skills are only weakly correlated with each other (Schlegel, Boone et al., 2017), the degree to which age differences in emotion perception accuracy extend to other types of interpersonal perception skills remains a critical area for inquiry in psychological research. A small but growing number of studies have found patterns of age-related similarities and even improvements in the perception of some non-emotional qualities and traits (e.g., dominance, deceit, competence, health, rapport; Bond, Thompson, & Malloy, 2005; Murry & Isaacowitz, 2018; Vicaria, Bernieri, & Isaacowitz, 2015; Zebrowitz et al., 2014), prompting the conclusion that perceiver age does not appear to have a strong negative effect on social judgment accuracy (Freund & Isaacowitz, 2018). Moreover, as noted briefly above, older adults maintain a variety of socioemotional skills that likely rely on the ability to make accurate social judgments (e.g., Diener et al., 1999; Grühn et al., 2008; Smith et al., 1999), further supporting the idea that social perception skills may be maintained with age. However, these studies represent a minority of findings in the literature, and thus, require further replication.

Finally, given that aging is related to robust declines in attentional and working memory processes (e.g., Loaiza & McCabe, 2013; Luszcz, 2011), traditional tasks that are thought to be linked to attention and information processing skills (Castro & Boone, 2015; Schlegel, Witmer et al., 2017) may conflate age-related declines in interpersonal accuracy with age-related declines in cognition. This possibility is further supported by significant associations between older adults’ interpersonal perception skills and cognitive abilities (Krendl & Ambady, 2010; Krendl, Rule, & Ambady, 2014; Sarabia-Cobo, Garcia-Rodriguez, Navas, & Ellgring, 2015). Methodological approaches that are less reliant on cognitive abilities—for example, that utilize longer presentations of behavior (i.e., more than 1–3 seconds) and that include contextual information —are thus needed to provide better unique assessments of aging individuals’ interpersonal perception skills.

In sum, the methodological limitations of traditional interpersonal accuracy tasks constrain our understanding of interpersonal perception skills; it is unclear to what extent individual differences in interpersonal accuracy generalize across stimuli and skills. These gaps in our understanding become especially problematic within the context of aging, as task limitations likely constrain the assessment of older adults’ full interpersonal accuracy capacity. In order to determine the extent to which interpersonal accuracy changes with age, including the nature of this change—what changes, when, and why—a new methodological approach must be adopted that accounts for task relevance and representativeness as a way to achieve conceptual continuity for aging samples. It is likely that making tasks more relevant—for example, by assessing a wider variety of relevant perceptions individuals make about others in everyday life—and more representative of everyday life—for example, by including real people experiencing real emotions as they naturally unfold over time in real situations—will reveal a more nuanced picture of how interpersonal accuracy changes with age. Such information is critical not only for aging research but also for a global understanding of the phenomena of interpersonal accuracy: Such an approach will allow researchers to investigate a greater diversity of interpersonal perception skills that better reflect the socioemotional milieu of everyday life for individuals of all ages.

The Present Study

In this study, we addressed past methodological limitations to advance our understanding of interpersonal accuracy across the adult lifespan. Specifically, we examined younger, middle-age, and older adults’ interpersonal perception skills using both traditional and nontraditional interpersonal perception tasks. The main objective was to identify the degree to which age group differences in interpersonal accuracy extend to a variety of interpersonal perception skills. We further examined whether the data provided greater evidence for age-related differences or age-related similarities in interpersonal accuracy using Bayesian statistics; this statistical approach is necessary given mixed evidence of age-related differences in making different types of interpersonal judgments. Moreover, a Bayesian approach affords the unique ability to quantify support for both the alternative hypothesis—that age groups differ in interpersonal accuracy—and the null hypothesis—that age groups do not differ in interpersonal accuracy; traditional null hypothesis significance testing is unable to quantify support for age-related similarities.

The overarching goal, thus, was to clarify the extent to which interpersonal accuracy becomes worse, better, or remains the same with age using a more diverse and ecologically-valid methodological approach and advanced statistical technique. Additionally, this goal has broad implications for the field of interpersonal accuracy by highlighting new methodological and statistical approaches to the study of individual differences in interpersonal perception skill.

In line with past research (e.g., Grainger et al., 2015; Isaacowitz et al., 2007; Malatesta et al., 1987; Paulmann et al., 2008; Spencer et al., 2016; Sullivan & Ruffman, 2004), we hypothesized that older and middle-age adults would be significantly less accurate than younger adults at identifying others’ posed and spontaneous emotional expressions. We expected these age differences to be greater for posed emotional expressions than spontaneous emotional expressions, given evidence that spontaneous stimuli can reduce—though not necessarily eliminate—age differences in emotion perception accuracy (e.g., Murphy et al., 2010; Sze et al. 2012). For non-emotion perception skills, however, we predicted age-related similarities in accuracy across the three age groups, following preliminary evidence from recent studies (e.g., Bond et al., 2005; Murry & Isaacowitz, 2018; Vicaria et al., 2015; Zebrowitz et al., 2014).

Method

Participants

Perceivers.

One hundred and fifty-one adults participated as perceivers in the study. Young adults (N = 51; 63% female) were on average 20.45 years old (SD = 4.22; range: 18 to 39) and were recruited primarily from the psychology introductory course participant pool at Northeastern University and from advertisements posted on and around campus. Middle-age adults (N = 49; 47% female) were on average 54.27 years old (SD = 7.81; range: 40 to 64) and were recruited from Boston area through online and community advertisements. Older adults (N = 51; 51% female) were on average 69.43 years old (SD = 3.22; range: 65 to 76) and were recruited from existing laboratory databases and through online advertisements and flyers posted at various community sites.

Sample size was determined through an a priori power analysis conducted using the program G*Power. Based on a two-tailed alpha value of 0.05, a power value of 0.80, Cohen’s f² value (F-test effect size) of .31 (effect size taken from Grainger et al., 2015; consistent with other published effect sizes in the literature, see Ruffman et al., 2008), and three groups, the recommended sample size was 102, or 34 perceivers per age group. Because the study design required perceivers to complete two laboratory visits (see Procedure section below), we anticipated some missing data. Thus, we oversampled perceivers to ensure sufficient minimum sample sizes for each age group. A sensitivity analysis conducted post hoc in G*Power indicated that the study was able to detect (with 80% power and alpha = .006) moderate sized age group differences (i.e., Cohen’s f = .32), providing further support for our sample size. Moreover, a power analysis for Bayesian analysis of variance conducted in R version 3.4.0 (R core team, 2017) using the package BayesFactor (Morey, Rouder, & Jamil, 2015), with a true effect d = 0.00, r scale = 0.5, group sample size of 50, and 100,000 simulations indicated a 98.5% probability of finding support for the null hypotheses (BF10 < 0.30) if the true effect size is 0. A second Bayesian power analysis, with a true effect d = 0.62 (transformed from f = .31), r scale = 0.5, group sample size of 50, and 100,000 simulations indicated a 76% probability of finding support for the alternative hypotheses (BF10 > 3.0) if the true effect size is 0.62. Collectively, these power analyses indicate adequate power to detect age differences and similarities in interpersonal accuracy in the present sample.

Perceivers represented diverse racial-ethnic backgrounds (58.4% White, 18.1% Black, 6.0% Hispanic/Latino, 5.4% Asian, 5.4% Multiracial, 6.7% Other). Most perceivers completed some college education (456%), with fewer perceivers indicating their highest level of education as a graduate degree (22.4%), Bachelor’s degree (22.4%), and high school diploma (9.5%). Prior to participation, middle-age and older adult perceivers were screened for cognitive impairment using the adapted telephone version of the Mini Mental State Examination (Newkirk et al., 2004); adults scoring above 21 were considered for enrollment.

Targets.

As described in detail below, a separate stimulus development study was conducted to create the stimuli for the new nontraditional interpersonal accuracy tasks presented to perceivers. Specifically, twenty-four adults from three age groups (young, middle age, and older) participated as targets in the stimuli development study. Young targets (N = 8; 37.5% female) were on average 19.63 years old (SD =0.92; range: 18–21 years), middle age targets (N = 8; 50% female) were on average 55.75 years old (SD = 7.21; range: 47–64 years), and older targets (N = 8; 62.5% female) were on average 71.13 years old (SD = 4.05; range: 65–76).¹ Targets were recruited using the same methods and procedures as perceivers; no targets served as perceivers.

Targets also represented a range of racial-ethnic backgrounds (33.3% White, 25.0% Black, 12.5% Hispanic/Latino, 12.5% Asian, 16.7% Multiracial) and education levels (4.2% less than high school, 16.7% high school or equivalency, 37.5% some college, 4.2% Associate degree, 16.7% Bachelor’s degree, 20.8% graduate school). As in the perceiver sample, middle age and older targets were screened for cognitive impairments prior to enrollment.

Interpersonal Accuracy Tasks

Perceivers completed a battery of interpersonal accuracy tasks that measured either emotion perception skills (identifying how others feel) or non-emotion perception skills (identifying other qualities in others). Tasks are summarized in Table 1. The majority of these tasks—herein referred to as nontraditional tasks—were developed specifically for the present study utilizing paradigms that elicited nontraditional stimuli depicting spontaneous (i.e., not posed or scripted) expressions and behaviors.

Table 1.

Summary of Interpersonal Accuracy Tasks and Interpersonal Perception Skills

Task	Type of Task	Type of Quality	Task Description	Skills Assessed
GERT-S	Traditional	Emotion	Assesses the ability to identify posed facial and vocal expressions of discrete emotions from 42 1-to-3-second video clips. Participants watch each video clip and then identify which emotion out of 14 is being displayed in the video clip. Responses are scored and averaged across video clips for accuracy.	posed emotion perception
Emotional Event Judgment	Non- traditional	Emotion	Assesses the ability to identify spontaneous global and discrete emotional expressions from 24 30-second video clips of an event discussion. Participants watch each video clip and then judge whether the target was discussing a positive or negative event and rate the targets’ satisfaction and tense feelings on 0–100 scales. Event valence responses are scored and averaged across videos for accuracy; participants’ satisfaction and tense ratings are correlated with targets’ satisfaction and tense ratings for accuracy.	event valence perception; satisfaction perception; tense perception
Disgust Judgment	Non- traditional	Emotion	Assesses the ability to identify spontaneous disgust expressions from 48 5-second video clips of someone responding to a disgusting image. Participants watch each video clip and then rate the targets’ disgust feelings on 0– 100 scale. Participants’ disgust ratings are correlated with targets’ disgust ratings for accuracy.	disgust perception
Amusement Judgment	Non- traditional	Emotion	Assesses the ability to identify spontaneous positive discrete emotional expressions from 48 10-second video clips of someone responding to a funny video. Participants watch each video clip and then rate the targets’ amusement, interest, and enjoyment on 0–100 scales. Participants’ amusement, interest, and enjoyment ratings are correlated with targets’ amusement, interest, and enjoyment ratings for accuracy.	amusement perception; interest perception; enjoyment perception
IPT-15	Non- traditional	Non- emotion	Assesses the ability to identify spontaneous displays of five types of social interactions: status, intimacy, kinship, competition, and deception. Participants watch 15 video- recorded scenes ranging from 30 to 120 seconds in length and are asked to identify what is going on in each scene by responding to a multiple choice question. For example, participants watch a scene depicting two basketball players discussing a recent match and are asked to identify which player won the basketball match. Responses are scored and averaged across scenes for accuracy.	social perception
Personal Qualities Judgment	Non- traditional	Non- emotion	Assesses the ability to identify spontaneous displays of personal qualities and traits from 24 30-second video clips of people discussing their life stories. Participants watch each video clip and then rate the targets’ education, health, and personality traits (for specific rating scales, see S1 in Supplemental Materials). Participants’ education, health, and personality ratings are correlated with targets’ education, health, and personality ratings for accuracy.	education perception; health perception; extraversion perception; agreeableness perception; conscientiousness perception; neuroticism perception; openness perception

For comparative purposes, a traditional task was also included that measured perceptions based on posed, prototypic expressions and behaviors. These methodological details and distinctions are noted below and in Table 1. Descriptive statistics for each interpersonal perception skill (as assessed by each task) for each age group are presented in Table 2.

Table 2.

Reliability Estimates and Descriptive Statistics for Interpersonal Perception Skills

	Reliability	Young		Middle-age		Older		Total Sample
	KR-20 or ra	N	M (SD)	N	M (SD)	N	M (SD)	N	M (SD)
Emotion Perception Skills
Posed Emotion	.79	51	.63 (.10)	47	.47 (.17)	49	.52 (.14)	147	.54 (.15)
Event Valence1	.70	27	.97 (.05)	23	.95 (.07)	20	.94 (.06)	70	.96 (.06)
Event Valence2	.70	24	.73 (.12)	26	.69 (.12)	31	.67 (.10)	70	.69 (.12)
Tense1	.03	27	.21 (.16)	23	.13 (.20)	20	.05 (.16)	70	.14 (.18)
Tense2	.30	24	−.08 (.17)	26	−.08 (.26)	31	−.06 (.18)	81	−.08 (.20)
Satisfaction1	−.16	27	.21 (.13)	23	.18 (.11)	20	.24 (.19)	81	.21 (.14)
Satisfaction2	−.41	24	.19 (.11)	26	.13 (.12)	31	.13 (.12)	81	.15 (.12)
Disgust	.11	51	.14 (.13)	48	.13 (.16)	51	.12 (.18)	150	.13 (.16)
Amusement	.56	51	.38 (.17)	47	.41 (.15)	46	.38 (.13)	144	.39 (.15)
Interest	.38	51	.40 (.13)	47	.39 (.17)	46	.38 (.15)	144	.39 (.15)
Enjoyment	.02	51	.50 (.14)	47	.47 (.18)	46	.49 (.13)	144	.49 (.15)
Non-emotion Perception Skills
Social Perception	.31	50	.67 (.11)	49	.59 (.16)	51	.63 (.13)	150	.63 (.14)
Education	.46	50	.56 (.22)	49	.40 (.24)	49	.45 (.23)	148	.47 (.24)
Health	−.17	50	.28 (.17)	49	.22 (.18)	49	.21 (.13)	148	.24 (.16)
Extraversion	.13	50	.33 (.19)	49	.16 (.22)	48	.26 (.21)	147	.25 (.22)
Agreeableness	−.25	49	.14 (.21)	47	.10 (.19)	48	.16 (.20)	144	.13 (.20)
Openness	−.01	49	.08 (.21)	47	.04 (.19)	47	.05 (.19)	144	.06 (.20)

Note:

¹= Version 1;

²= Version 2;

^a= split-half correlation with Spearman-Brown adjustment.

Emotion perception tasks.

Emotion perception skill was measured using both traditional and nontraditional tasks.

The Geneva Emotion Recognition Test.

Emotion perception accuracy was measured traditionally using the short version of the Geneva Emotion Recognition Test (GERT-S; Schlegel & Scherer, 2016). The GERT-S assesses the ability to identify posed facial and vocal expressions of discrete emotions from video clips. Perceivers were shown 42 1–3-second video clips of an actor saying a nonsensical sentence with a target emotional vocal and facial expression (e.g., an angry voice and an angry face). After each video clip, perceivers were presented with a list of 14 discrete emotion labels and asked to select the emotion word that best corresponds to the emotion that the actor in the video was trying to express. Accuracy was scored for each video clip (accurate or not) and averaged across clips to create a posed emotion perception accuracy score ranging from 0 to 1. Reliability for the GERT-S was high (KR-20 = .79), and validity has been demonstrated through significant associations with other interpersonal tasks (Schlegel, Boone et al., 2017) and measures of emotion understanding and emotion regulation (Schlegel & Scherer, 2016). The GERT-S has been used with aging samples (Schlegel, Vicaria, Isaacowitz, & Hall, 2017).

Three nontraditional emotion perception tasks were developed for the present study that collectively assessed seven additional emotion perception skills. As noted briefly above, these new tasks were developed as part of an independent study examining individuals’ emotion-related experiences in the laboratory; the single goal of this study thus was to develop the stimuli for the new nontraditional interpersonal accuracy tasks. Twenty-four adults (see Targets section above) visited the laboratory and were video-recorded as they engaged in various emotion-eliciting tasks. These video-recordings were used as stimuli for the new emotion perception accuracy tasks, as described below. Thus, in each task, the same 24 targets (i.e., the adults in the video-recordings) were presented to perceivers, resulting in a task battery that measures a variety of emotion perception skills based on the same targets. To make these tasks nontraditional, targets were video-recorded as they behaved naturally in response to various emotion-eliciting situations in the laboratory (e.g., Buck, Powers, & Hull, 2017; Dunlap, 1927; North, Todorov, & Osherson, 2012), creating stimuli that are unscripted, spontaneous, and representative of a range of naturally occurring expressivity levels. As described below, targets provided ratings on their emotional experiences which served as the criteria for accuracy in the new nontraditional emotion perception tasks (e.g., Ickes, Robertson, Tooke, & Teng, 1986; Kraus, Côté, & Keltner, 2010; Snodgrass, Hecht, & Ploutz-Snyder, 1998).

Emotional event judgment task.

This task consisted of 24 30-second² video clips depicting targets discussing a recent positive or negative interpersonal event (e.g., a party, an argument). Specific events were identified by targets and thus could vary across targets. After discussing each type of event, targets rated the degree to which they were feeling tense and satisfied on independent scales ranging from 0 (not at all) to 100 (extremely); these ratings served as criterion measures for the Emotional Event Judgment Task. Because each target discussed two types of events, positive and negative events for each target were randomized and balanced across two parallel versions of the task (i.e., one version depicting Target A discussing a positive event, and another version depicting Target A discussing a negative event).

Perceivers were randomly assigned to a task version.³ In each version, perceivers viewed each video clip and were asked to indicate what type of event the target in the video was talking about (positive or negative) and to rate the degree to which the target was feeling tense and satisfied immediately after speaking about the event using the same 0–100 scales that targets completed.

Three emotion perception accuracy scores were generated from this task: event valence perception accuracy, which was the accuracy in identifying the valence of the event the target was talking about averaged across targets, ranging from 0 to 1; tense perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ tense ratings, ranging from −1.00 to 1.00; and satisfaction perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ satisfaction ratings, ranging from −1 to 1. Following common standards in the field of interpersonal accuracy (e.g., Ambady, Hallahan, & Rosenthal, 1995; Snodgrass & Rosenthal, 1985), tense and satisfaction perception accuracy scores were then Fisher z-transformed to normalize the distributions. Analyses were thus conducted on these transformed correlations.

Disgust judgment task.

This task consisted of 48 5-second⁴ video clips depicting targets reacting to disgusting images that were presented to them on a computer screen. Targets were depicted twice in the task, reacting to two different images, resulting in two video clips per target. The images used to elicit target reactions were selected from the International Affective Picture System (IAPS; Lang, Bradley, & Cuthbert, 1997); images were selected to elicit varying levels of disgust. Specifically, we utilized published norms to select images that were equally evocative for young and older adults (Grühn & Scheibe, 2008). To ensure sufficient disgust elicitation, IAPS images were supplemented with similar published images (Ngo & Isaacowitz, 2015). After viewing each image, targets rated the extent to which they felt disgusted on a scale ranging from 0 (not at all) to 100 (extremely); these ratings served as the criterion measures for the Disgust Judgment Task.

Perceivers viewed targets as they reacted to disgusting images (e.g., a dirty bathroom; a decaying corpse) and then rated the degree to which the targets felt disgusted on the same 0–100 scale that targets completed. A disgust perception accuracy score was calculated by correlating perceiver and target ratings within each perceiver; scores thus ranged from −1 to 1. As in the previous task, correlations were Fisher r-to-z transformed prior to data analysis.

Amusement judgment task.

This task consisted of 48 10-second⁵ video clips depicting targets as they watched amusing segments from popular TV sitcoms (i.e., Curb Your Enthusiasm, Golden Girls, Mr. Bean, The Office). Segments were selected from a previous study that used the segments to successfully elicit amusement from young, middle-age, and older adults (Stanley, Lohani, & Isaacowitz, 2014). After viewing each segment, targets rated the extent to which they felt amused by and interested in the segment on independent scales ranging from 0 (not at all) to 100 (extremely). Targets also rated the extent to which they enjoyed watching each segment on a scale ranging from 0 (did not enjoy at all) to 100 (enjoyed very much); these ratings served as the criterion measures for the Amusement Judgment Task.

Perceivers viewed each video clip and were asked to rate the degree to which the target in the video felt amused and interested by what they saw as well as the degree to which the target enjoyed the TV segment; perceiver ratings were made on the same 0–100 scales as completed by targets. Perceiver and target ratings were then correlated within perceivers and Fisher r-to-z transformed to create three accuracy scores: amusement perception accuracy, interest perception accuracy, and enjoyment perception accuracy.

Non-emotion perception tasks.

All non-emotion perception tasks consisted of nontraditional stimuli (i.e., targets depicting spontaneous, unscripted expressions and behavior).

The Interpersonal Perception Test.

A short version of the Interpersonal Perception Test (IPT-15; Costanzo & Archer, 1993) was used to measure social perception accuracy. The IPT-15 consists of 15 video-recorded scenes varying in length from 28-to-122 seconds. Each scene depicts one of five types of social interactions: status, intimacy, kinship, competition, and deception. Each of the 15 video-recorded scenes depicted spontaneous behavior displayed by one to five people. For example, one scene (depicting competition) displays two adult males following a basketball match where one player won; another scene (depicting status) displays an adult male and an adult female discussing a work-related interaction, and one of the adults is the other’s boss. The task was originally created to provide a more ecologically valid assessment of interpersonal accuracy, as scenes depicted the behaviors of real people in a variety of real life situations (Costanzo & Archer, 1993).

Prior to viewing each scene, perceivers were presented with a target question (e.g., “Who won the game of one-on-one basketball?” or “Who is the higher status person?”); each scene thus has an objectively correct answer (e.g., one of the males did win the basketball match; one of the adults is the others’ boss). Perceivers then watched the scene and were again presented with the target question and two to three multiple choice responses. Perceivers were instructed to select the response that best answered the target question and to speak their response selection out loud. A trained research assistant recorded perceivers’ responses from an experimental control room. Accuracy was scored for each scene and averaged to create a total social perception accuracy score ranging from 0 to 1.

Reliability for the IPT-15 (Cronbach’s α = .31) though low by traditional psychometric standards was adequate for this type of task assessing multiple domains (Bollen & Lennox, 1991) and consistent with past research utilizing the task (e.g., Costanzo & Archer, 1993). Importantly, the IPT has demonstrated good test-retest reliability (Costanzo & Archer, 1993) and validity through associations with peer ratings of interpersonal sensitivity (Costanzo & Archer, 1993) and other social perception tasks in aging samples (Murry & Isaacowitz, 2018).

Personal qualities judgment task.

This task consists of video-recordings from the same target stimulus development study used to develop the nontraditional emotion perception tasks. Thus, the 24 targets in the Personal Qualities Judgment Task are the same 24 targets in the Emotional Event Judgment Task, Disgust Judgment Task, and Amusement Judgment Task.

The Personal Qualities Judgment Task consists of 24 30-second⁶ video clips depicting targets sharing their life stories. Specifically, targets were given a prompt to freely discuss who they are, what it was like growing up in their family, what their current interests and hobbies are, what their goals for the future are, and so on. Targets could share as little or as much as they wanted about any of these areas of their lives. Criterion measures for the Personal Qualities Judgment Task were provided by target ratings on various personal qualities and traits. Specifically, targets indicated their highest level of education from a list of less than high school, high school or equivalency, some college, Associate degree, Bachelor’s degree, and graduate school or more. Targets then rated their global health status on a 5-point scale ranging from excellent to poor (Hays, Sherbourne, & Mazel, 1993). Finally, targets reported on their extraversion, agreeableness, conscientiousness, neuroticism, and openness to experience using the 10-item Big Five Inventory (BFI-10; Rammstedt & John, 2007).

After viewing each video clip in the Personal Qualities Judgment Task, perceivers were asked to rate the same personal qualities and traits about each target. Specifically, perceivers were asked to indicate the highest level of education the target completed, how healthy the target was, and how extraverted, agreeable, conscientious, neurotic, and open to experience the target was. For specific rating scales, see S1 in Supplemental Materials.

Seven correlation-based accuracy scores were calculated from these ratings (all ranging from −1 to 1): education perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ education ratings, with education ratings treated as ordinal data to enable correlation calculations; health perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ health ratings; extraversion perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ extraversion ratings; agreeableness perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ agreeableness ratings; conscientiousness perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ conscientiousness ratings; neuroticism perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ neuroticism ratings; and openness perception accuracy, which was the within-perceiver correlation between perceivers’ and targets’ openness to experience ratings. As with the other correlation-based accuracy scores, correlations were Fisher z-transformed prior to data analysis.

Other Measures

Because we created many new interpersonal accuracy tasks, we also measured additional constructs that are known correlates of interpersonal perception skill; specifically, fluid and crystallized cognition and the Big Five personality traits. Descriptive statistics for each measure for each age group are presented in S2 of Supplemental Materials.

Fluid and crystallized cognition.

Perceivers completed a battery assessing their fluid and crystallized cognitive abilities. To measure fluid cognition, perceivers completed the Digit Span (Forward and Backward) and Digit Symbol (Digit and Symbol Substitution) tasks from the Wechsler Adult Intelligence Scale-Revised (WAIS-R; Wechsler, 1981). To measure crystallized cognition, perceivers completed the Shipley Vocabulary Test (Shipley, 1940) and two measures of verbal fluency (animals and letters; Rosen, 1980; Spreen & Benton, 1977). In all cases, higher scores indicate greater cognitive ability. Measures of cognition such as these have been shown to relate positively to various interpersonal perception skills, including the ability to make accurate emotional and non-emotional judgments (Murphy & Hall, 2011; Schlegel et al., 2017).

Personality inventory.

Perceivers reported on their personality traits using the BFI-10 (Rammstedt & John, 2007). Specifically, perceivers rated the extent to which they saw themselves as the type of person who displays different behaviors or tendencies on a 5-point Likert scale (Disagree Strongly to Agree Strongly). Two items are presented for each of five traits: Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Openness to Experience. Ratings were averaged across trait items. Meta-analytic results suggest that interpersonal perception skill is positively linked to extraversion, conscientiousness, and openness to experience and negatively linked to neuroticism, although these effects are small (Hall et al., 2009).

Procedure

To reduce participant burden and maximize the potential for all interpersonal accuracy tasks to be completed, the study was divided into 2 two-hour laboratory visits; all visits were run individually. Tasks were randomized across and within visits to control for any potential order effects. Visits were completed one-to-two weeks apart, typically on the same day of the week and at the same time of day. All study procedures received approval from the [omitted for blind review] Institutional Review Board (#XX).

Perceivers completed informed consent at the beginning of the first visit. They then were administered the cognitive battery by a trained research assistant; perceivers began with forward and backward digit span task, followed by the vocabulary task, verbal fluency tasks, and digit symbol substitution tasks. The same order of cognitive tasks was presented across perceivers. Next, perceivers completed half of the interpersonal accuracy tasks in a randomized order, with a 5-min break offered as needed at the half-way point (e.g., after one hour). At the second visit, perceivers completed the remaining interpersonal accuracy tasks in a randomized order and were debriefed and compensated. A demographic questionnaire (assessing perceivers’ age, gender, ethnicity, and education level) and personality inventory were typically administered at the end of the second visit or at the end of the first visit if time allowed (i.e., if participants completed tasks early).

All interpersonal accuracy tasks consisted of audiovisual information that was presented to perceivers on a computer in the laboratory. Perceivers used headphones to listen to stimuli and could adjust the volume as needed. The new tasks created for the study were administered via surveys hosted on Qualtrics which allowed perceivers’ ratings and responses to be recorded directly on the computer. Similarly, the GERT-S was presented and recorded via a Qualtrics survey. The IPT-15 was presented to perceivers using Microsoft PowerPoint software, and responses were recorded by a research assistant in a separate control room using paper and pencil following past procedures with aging samples to ensure that all perceivers understood the question and response options for each scene (Murray & Isaacowitz, 2017).

Results

Data Processing

As has been noted elsewhere (e.g., Hall, 2001; Schlegel, Boone et al., 2017), interpersonal accuracy tasks often have low internal consistency reliability, perhaps reflecting the relatively small number of items used to construct such tasks and the fact that items may measure different facets of the same underlying general ability (for more discussion, see Bollen & Lennox, 1991). Nonetheless, it is also known that low reliability constrains the predictive potential of interpersonal perception tasks (Guilford, 1954), and thus, null effects (such as those relating to age) might reflect measurement error rather than true null effects.

To address this concern, we engaged in three psychometric data processing steps. First, we conducted analyses aimed at identifying and improving the internal consistency reliability estimates for the newly created interpersonal accuracy tasks. To do so, we utilized a data-driven procedure. For event valence perception accuracy, we first calculated the KR-20 coefficient. We then identified items that were negatively correlated with the average item-correlation (without that item); items with the largest negative correlation were removed one at a time until KR-20 coefficient was ≥ .70, indicating sufficient reliability. This process resulted in the dropping of 10 items; the final event valence perception accuracy score thus consisted of 14 items. For the remaining tasks that were scored using within-perceiver correlations (i.e., tense perception accuracy, satisfaction perception accuracy, disgust perception accuracy, amusement perception accuracy, interest perception accuracy, enjoyment perception accuracy, education perception accuracy, health perception accuracy, extraversion perception accuracy, agreeableness perception accuracy, conscientiousness perception accuracy, neuroticism perception accuracy), reliability was calculated using split-half correlations with a Spearman-Brown adjustment; halves were randomly generated for each task prior to correlation.⁷

Next, we examined whether all mean accuracy scores were significantly different than chance using one sample t tests. Chance accuracy was identified for each task as the proportion of correctly identified items due to random guessing given the number of response options and the way in which accuracy was calculated; for example, for posed emotion perception accuracy (measured by the GERT-S, which has 14 response options one of which is correct for each item), chance accuracy is 1/14 or 0.0714. In contrast, for correlation-based accuracy scores, such as education perception accuracy, chance accuracy is 0. All tasks except for conscientious perception accuracy demonstrated mean accuracy levels that varied significantly from chance (ps < .05).⁸ Consequently, conscientiousness perception accuracy was dropped from subsequent analyses.

Finally, because we know of no psychometric method to improve the reliability of correlation-based accuracy tasks, we examined the predictive validity of these tasks by calculating the intercorrelations among all interpersonal accuracy scores (see S3 in Supplemental Materials). This step provided additional information regarding the psychometrics of our newly developed tasks: Tasks with low reliability were likely to be unrelated to other tasks, indicating that they are not good measures of individual differences in interpersonal accuracy, whereas tasks with moderate reliability may demonstrate good predictive validity through significant positive associations with other tasks.

All accuracy scores were positively and significantly correlated with at least one other accuracy score except for neuroticism perception accuracy, which was negatively and significantly correlated with health perception accuracy, extraversion perception accuracy, event valence perception accuracy, disgust perception accuracy, posed emotion perception accuracy, and social perception accuracy. Consequently, we dropped neuroticism perception accuracy from subsequent analyses. All other accuracy scores demonstrated sufficient construct validity, indicating that any measurement error related to low reliability did not interfere with these tasks’ abilities to predict individual differences in interpersonal perception skill. Reliability estimates and descriptive statistics for the final set of 14 interpersonal perception skills are presented in Table 2.⁹

As further demonstration of predictive validity, we correlated these 14 interpersonal perception skills with perceivers’ fluid and crystallized cognitive skills and personality traits (see Table 3). Interpersonal perception skills were generally positively correlated with fluid and crystallized cognition, replicating past research (i.e., Murphy & Hall, 2011; Schlegel et al., 2017). However, in contrast to previous findings (Hall et al., 2009), interpersonal perception skills were generally unrelated to perceivers’ personality traits and in some cases were inversely related to some traits. For example, conscientiousness was negatively associated with the ability to accurately perceive others’ posed emotional expressions and extraversion. Similarly, neuroticism was negatively associated the ability to accurately perceive others’ posed emotional expressions, satisfaction, health, and openness to experience. Although these latter associations stand in contrast to past meta-analytic results (Hall et al., 2009), recent research has found null correlations between interpersonal perception skill and conscientiousness and negative associations with neuroticism (Schlegel, Fontaine, & Scherer, 2017), supporting our results.

Table 3.

Correlations between Interpersonal Perception Skills, Cognitive Abilities, and Personality Traits

	Fluid Cognition				Crystallized Cognition
	Forward Digit Span	Backward Digit Span	Digit Substitution	Symbol Substitution	Vocabulary	Letter Fluency	Animal Fluency
Emotion Perception Skills
Posed Emotion	.39***	.43***	.40***	.37***	.38***	.30***	.39***
Event Valence1	.30*	−.03	.14	.19	.22†	.02	.04
Event Valence2	.24*	.11	.15	.12	−.08	−.08	.02
Tense1	.17	−.11	.35**	.33**	.03	−.11	−.05
Tense2	.21†	−.01	−.03	−.12	−.09	−.12	−.14
Satisfaction1	−.16	−.08	−.02	.06	.03	−.06	−.07
Satisfaction2	.18	.10	.18	.02	.02	.13	.13
Disgust	.17*	.21**	.10	.11	.33***	.17*	.20*
Amusement	.07	.01	.02	−.10	.10	.06	.11
Interest	.02	.11	.09	.02	.08	.09	.17*
Enjoyment	−.01	−.01	.01	−.10	−.02	−.02	.05
Non-emotion Perception Skills
Social Perception	.22**	.27**	.21*	.25**	.21*	.11	.36***
Education	.16†	.04	.22**	.28**	.12	.12	.20*
Health	.15†	.21*	.14†	.10	.14†	.05	.12
Extraversion	.23**	.11	.23**	.29***	.16*	.04	.13
Agreeableness	.09	−.03	.02	.05	−.03	.00	−.17*
Openness	.24**	.22**	.13	.14	.04	.18*	.14†
	Personality Traits
	Extraversion	Agreeableness	Conscientiousness		Neuroticism	Openness to Experience
Emotion Perception Skills
Posed Emotion	.03	−.24**	−.17*		.28**	.09
Event Valence1	−.18	−.07	−.19		.14	−.00
Event Valence2	.02	.10	.00		.20†	.04
Tense1	−.01	−.10	.01		−.09	.16
Tense2	−.06	−.08	.02		−.10	−.12
Satisfaction1	.09	.08	.10		−.19	−.05
Satisfaction2	−.09	−.06	.01		.23*	.02
Disgust	−.09	−.18*	−.07		.10	.04
Amusement	.10	−.02	.03		.03	−.01
Interest	.12	−.10	.05		.04	.01
Enjoyment	.08	.04	.03		.01	.01
Non-emotion Perception Skills
Social Perception	−.06	−.10	.02		.02	−.07
Education	.11	−.00	−.07		.04	.07
Health	.03	.02	.01		.21*	.02
Extraversion	−.07	−.18*	−.19*		.07	.00
Agreeableness	.19*	.06	.10		.00	.05
Openness	−.07	−.03	−.01		.30***	.10

Note:

¹= Version 1

²= Version 2.

^†p < .10.

^*p < .05.

^**p < .01.

^***p < .001.

Collectively, these findings support the reliability and validity of our interpersonal accuracy measures.

Testing Age Differences and Similarities in Interpersonal Accuracy

To test our primary hypotheses that (1) older and middle-age adults would be significantly less accurate than younger adults at identifying others’ emotional expressions and (2) younger, middle-age, and older adults would perform similarly when identifying others’ non-emotional qualities and expressions, we conducted a series of Analyses of Variance (ANOVA) models using both Null Hypothesis Significance Testing (NHST) and Bayesian statistics. NHST quantifies support for the alternative hypothesis—that age groups vary in interpersonal accuracy—but it does not quantify support for the null hypothesis—that age groups do not vary in interpersonal accuracy. A nonsignificant finding is thus uninformative within the framework of NHST; one cannot determine whether there are no true age differences, for example, or if age differences were simply too small or difficult to detect. In contrast, Bayesian statistics provide a method to make sense of seemingly uninformative null effects; that is, Bayesian statistics can quantify support for the null hypothesis relative to the alternative hypothesis (and vice versa) given the data and some assumed prior. Consequently, our use of both NHST and Bayesian statistics is appropriate given our interests in testing both age-related differences and similarities in interpersonal accuracy.

Traditional ANOVAs were conducted in SPSS; for these models, age group was entered as a between-subjects factor with three levels. In the cases of event valence perception accuracy, tense perception accuracy, and satisfaction perception accuracy, task version was entered as a second between-subjects factor to account for version differences in accuracy. To control for increased Type II errors resulting from multiple models testing the same hypothesis, we adjusted the omnibus critical p-value to .006 for comparisons among emotion perception skills and .008 for comparisons among non-emotion perception skills. Significant age group differences were examined using Bonferroni or Games-Howell post hoc comparisons with the critical alpha adjusted to .002 to account for the multiple pairwise comparisons among age groups across models (i.e., 21 comparisons for emotional judgments and 18 comparisons for non-emotional adjustments). Results are presented in Table 4.

Table 4.

Traditional ANOVA Results Comparing Age Groups on Interpersonal Perception Skills

	Age Group			Task Version			Age X Version
	F	η2p	p	F	η2p	p	F	η2p	p
Emotion Perception Skills
Posed Emotion1	18.27	.18	.001
Event Valence	3.16	.04	.045	286.93	.66	.001	0.50	.01	.606
Tense	1.92	.03	.151	43.23	.23	.001	2.98	.04	.054
Satisfaction	1.27	.02	.284	9.02	.06	.003	1.48	.02	.231
Disgust	0.20	.00	.819
Amusement	0.53	.01	.588
Interest	0.18	.00	.835
Enjoyment	0.31	.00	.737
Non-emotion Perception Skills
Social Perception1	4.00	.06	.021
Education	6.14	.08	.003
Health	2.72	.04	.069
Extraversion	8.33	.10	.001
Agreeableness	0.94	.01	.392
Openness	0.65	.01	.523

Note:

¹Levene’s statistic was significant, indicating that the assumption of homogeneity of group variance was violated. Thus, the Welch statistic is reported for the robust test of quality of means.

Significance criteria adjusted for multiple comparisons to .006 alpha for emotional judgments and .008 for non-emotional judgments.

Bayesian ANOVAs were conducted in JASP (JASP Team, 2017) following the recommendations of Wagenmakers and colleagues (2017). For these models, we applied the default Cauchy prior width of 0.5, which assigns the most mass to effect sizes that are closer to 0 than to 1, and thus, may be thought to sufficiently represent small-to-moderately sized effects (for further discussion, see Wagenmakers et al., 2017). We examined the BF₁₀ values for age group which quantifies support for the alternative hypothesis—that the age groups differ in interpersonal accuracy—relative to the null hypothesis—that the age groups do not differ in interpersonal accuracy—given the prior and the data; BF₁₀ > 3.0 indicates evidence for the alternative hypothesis, whereas BF₁₀ < 0.3 indicates evidence for the null hypothesis (Wagenmakers, Wetzels, Borsboom, & van der Maas, 2011). Bayesian post hoc comparisons were similarly examined for specific age group pairwise comparisons. In the cases of event valence perception accuracy, tense perception accuracy, and satisfaction perception accuracy, we first examined whether the age group by task version interaction model outperformed the other models; in all cases, the interaction effect resulted in lower posterior odds (all BF_inclusion < 1.0), arguing against the existence of interaction effects. We then examined the BF₁₀ values for age group (both the omnibus effect and post hoc comparisons) and task version using the same criteria as above.

Emotion perception skills.

As shown in the top half of Table 4, of the eight emotion perception skills examined, only one—posed emotion perception accuracy—demonstrated an omnibus age group difference (p < .006). Indeed, there was substantial support for the hypothesis that age groups differed in their posed emotion perception accuracy relative to the null hypothesis, given the prior and the data, BF₁₀ = 25896.63. Games-Howell post hoc comparisons indicated that both middle-age and older adults were significantly less accurate than younger adults in identifying others’ posed discrete emotional expressions (ps < .001); these results were further supported by Bayesian analyses, with BF₁₀ = 53094.68 and 314.31 for middle-age and older adult comparisons to young adults. Although middle-age adults did not vary significantly from older adults on posed emotion perception accuracy (p = .244), there was no conclusive evidence for differences or similarities in accuracy for these age groups given the prior and the data, BF₁₀ = 0.68.

Although the omnibus age effect for event valence perception accuracy was not significant in the traditional ANOVA (p > .006) and inconclusive in the Bayesian analyses (BF₁₀ = 1.21), Bayesian post hoc comparisons indicated greater support for age differences than similarities between younger and older adults specifically, BF₁₀ = 4.72, suggesting that the data provided some evidence for a perceptual advantage for younger adults over older adults on judgments of event valence. Indeed, post hoc comparisons in the traditional ANOVA also indicated significant differences in event valence perception accuracy between younger and older adults, p < .001. The model that included age group and version was 10.6 times more likely than the model that included only version (BF₁₀ = ~ 30210/2845 = 10.6) and was 5.95 times more likely than the model that included age group, version, and their interaction (BF₁₀ = ~ 30210/5078 = 5.95), indicating that this age group difference was robust and not dependent upon task version differences. No other age group differences were supported by the data: comparisons between younger and middle-age adults were inconclusive (p = .042, BF₁₀ = 0.55) as were comparisons between middle-age and older adults (p = .162, BF₁₀ = 0.36).

For the other emotion perception skills, all of which assessed accuracy in identifying the degree to which a target displayed a specific spontaneous emotion, there was generally greater evidence of age-related similarities than differences in emotion perception accuracy given the prior and the data. Specifically, null age effects were supported for five of the six remaining emotion perception skills: satisfaction perception accuracy, disgust perception accuracy, amusement perception accuracy, interest perception accuracy, and enjoyment perception accuracy (across models, age group BF₁₀ < 0.20). Bayesian post hoc comparisons revealed the most evidence for similarities between younger and older adults, and to lesser extents between middle-age and older adults and between younger and middle-age adults. For example, for satisfaction perception accuracy, older adults performed similarly to younger (p = 1.00, BF₁₀ = 0.30) and middle-age adults (p = .305, BF₁₀ = 0.25).¹⁰ Older adults also performed similarly to younger (p = 1.00, BF₁₀ = 0.25) and middle-age adults (p = 1.00, BF₁₀ = 0.22) for disgust perception accuracy, as did middle-age and younger adults (p = 1.00, BF₁₀ = 0.23). For amusement perception accuracy, older and younger adults performed similarly (p = 1.00, BF₁₀ = 0.25) as did younger and middle-age adults (p = 1.00, BF₁₀ = 0.23). For interest perception accuracy, older adults performed similarly to younger (p = 1.00, BF₁₀ = 0.26) and middle-age adults (p = 1.00, BF₁₀ = 0.23) and middle-age adults performed similarly to younger adults (p = 1.00, BF₁₀ = 0.22). Finally, for enjoyment perception accuracy, older and younger adults performed similarly (p = 1.00, BF₁₀ = 0.22), as did older and middle-age adults (p = 1.00, BF₁₀ = 0.25) and younger and middle-age adults (p = 1.00, BF₁₀ = 0.26). For tense perception accuracy, there was inconclusive evidence for age-related differences or similarities between age groups overall (BF₁₀ = .55) and in comparisons between younger and older adults (p = .038, BF₁₀ = 2.68), between younger and middle-age adults (p = .476, BF₁₀ = 0.37), and between middle-age and older adults (p = .844, BF₁₀ = 0.31).¹¹

Taken together, these results provide only partial support for our first hypothesis. As predicted, younger adults outperformed older and middle-age adults on a traditional laboratory task that measured the perception of posed, prototypic emotional expressions. Younger adults also performed better than older adults on a nontraditional task that measured the perception of the global valence of an event discussion from spontaneous expressions. In contrast to our prediction, however, age-related similarities were supported for five other emotion perception skills, all assessed using nontraditional tasks, and one skill received inconclusive support for either age-related differences or similarities.

Non-emotion perception skills.

As shown in the bottom half of Table 4, two of the six non-emotion perception skills demonstrated omnibus age group differences (education perception accuracy and extraversion perception accuracy; ps < .007). Bayesian analyses further supported age differences in education perception accuracy (BF₁₀ = 11.48) and extraversion perception accuracy (BF₁₀ = 69.04). Specifically, younger adults outperformed middle-age adults when perceiving others’ education (p = .002, BF₁₀ = 36.96) and when perceiving others’ extraversion (p < .001, BF₁₀ = 259.65). Comparisons on these skills between younger and older adults (p = .076, BF₁₀ = 2.14 for education perception accuracy; p = .333, BF₁₀ = 0.73 for extraversion perception accuracy) and between middle-age and older adults (p = .720, BF₁₀ = 0.38 for education perception accuracy; p = .050, BF₁₀ = 2.15 for extraversion perception accuracy), however, were inconclusive.

Although the omnibus age effect for social perception accuracy was not significant in the traditional ANOVA (p > .008) and inconclusive in the Bayesian analyses (BF₁₀ = 2.48), Bayesian post hoc comparisons indicated greater support for age differences than similarities between younger and middle-age adults (p = .016, BF₁₀ = 6.80). Comparisons between older and younger adults (p = .353, BF₁₀ = 0.49) and between middle-age and older adults (p = .25, BF₁₀ = 0.65) on social perception accuracy were inconclusive.

For the remaining three non-emotion perception skills, there was some evidence for age-related similarities. Although the omnibus age effects were inconclusive for health perception accuracy (BF₁₀ = 0.66) and agreeableness perception accuracy (BF₁₀ = 0.38), post hoc comparisons revealed that older adults performed similarly to middle-age adults when perceiving others’ health (p = 1.00, BF₁₀ = 0.22) and agreeableness (p = .558, BF₁₀ = 0.21). Moreover, age-related similarities were supported for openness perception accuracy (BF₁₀ = 0.12), specifically between younger and older adults (p = 1.00, BF₁₀ = 0.27) and between middle-age and older adults (p = 1.00, BF₁₀ = 0.23). Comparisons among young and middle-age adults were inconclusive for both agreeableness perception accuracy (p = .996, BF₁₀ = 0.79) and openness perception accuracy (p = .792, BF₁₀ = 0.37) as was the comparison between younger and older adults’ perception of agreeableness (p = 1.00, BF₁₀ = 0.85). Similarly, for health perception accuracy, there was inconclusive evidence of similarities or differences between younger and middle-age adults (p = .196, BF₁₀ = 0.79) and between younger and older adults (p = .101, BF₁₀ = 2.05).

Taken together, these results provide partial support for our second hypothesis: As predicted, some types of non-emotion perception skills demonstrated age-related similarities whereas other types of non-emotion perception skills either varied with age or received inconclusive support for differences or similarities across age groups. When differences were supported, however, middle-age adults were the lowest performing group.

Exploratory Analyses

Because age group was significantly associated with participant education, χ² (6) = 92.32, p < .001, and ethnicity, χ² (10) = 52.79, p < .001, we explored whether education and ethnicity accounted for the five significant age group differences reported above. Moreover, given numerous positive correlations between perceivers’ fluid and crystallized cognition and their interpersonal perception skills (see Table 3), we also explored whether significant age group differences in interpersonal accuracy were independent of individual differences in cognition. Because we did not plan for these analyses a priori, and thus, did not account for these tests in our initial power analysis, we addressed these possibilities using exploratory Bayesian analyses. However, we note that these exploratory analyses are also explanatory in that they examined the extent to which age differences in interpersonal perception accuracy are potentially due to other perceiver-level characteristics, including education, ethnicity, and cognitive ability.

Specifically, we conducted a series of two-way Bayesian ANOVAs¹² with age group and education/ethnicity as between-subject factors. We examined the BF₁₀ for each model (i.e., single main effects, main effects together, main effects + interaction effects) to identify the best fitting model; we were interested in whether the models that included education/ethnicity only or the interaction models would be more competitive than the models that included age group only or age group and education/ethnicity main effects. We then examined the BF_inclusion for each effect to identify the variable that resulted in the greatest change from prior to posterior odds given the data. Because fluid and crystallized cognition were measured as continuous rather than categorical variables, Bayesian multiple regression models were conducted. In the cognition models, we examined the BF_inclusion for each effect to determine whether age group resulted in the greatest change from prior to posterior odds given the data.

Education.

Across models, age differences in interpersonal accuracy appeared to be robust to education effects; in some cases, education resulted in lower model probability, arguing against the inclusion of education effects alongside age group effects. Consequently, age differences in posed emotion perception accuracy, event valence perception accuracy, education perception accuracy, extraversion perception accuracy, and social perception accuracy were not due to individual differences in perceivers’ education levels. For full education exploratory analyses, see S4 in Supplemental Materials.

Ethnicity.

For posed emotion perception accuracy, the best-fitting model was the model that included age group, ethnicity, and their interaction, but this model was not more likely than the model that included age group and ethnicity only (BF₁₀ = 522760/257774 = 2.0). However, the model that included age group and ethnicity main effects was 17.0 times more likely than the model that included age group only (BF₁₀ = 257774/15154 = 17.0) and 1115.9 times more likely than the model that included ethnicity only (BF₁₀ = 257774/231 = 1115.9), suggesting the presence of unique age group and ethnicity effects. Moreover, age group resulted in the greatest change from prior to posterior odds (BF_inclusion = 2277.13), with ethnicity resulting in a smaller improvement to the model (BF_inclusion = 34.35) and their interaction resulting in less improvement (BF_inclusion = 7.66). These results suggest that age differences in posed emotion perception accuracy were robust to effects of ethnicity.

For event valence perception accuracy, the model that included age group, ethnicity, and task version main effects was the best fitting model but was not more likely than the models that contained the age group main effect only and the age group and task version main effects (1 > BF₁₀ > 3), with age group being the common factor across these models. Moreover, task version was the only effect that resulted in model improvement, BF_inclusion > 1,000,000. Together, these results argue against the inclusion of ethnicity effects in predicting event valence perception accuracy.

For education perception accuracy, the model that included age group only was the best fitting model and was 53.5 times more likely than the model that included ethnicity only (BF₁₀ = 9.63/.18 = 53.5), 10.82 times more likely than the model that included age group and ethnicity (BF₁₀ = 9.63/.89 = 10.82), and 107 times more likely than the model that included age group, ethnicity, and their interaction (BF₁₀ = 9.63/.09 = 107). Moreover, age group was the only effect that resulted in model improvement, BF_inclusion = 6.02; all other effects resulted in lower posterior odds. These results argue against the modeling of ethnicity effects and indicate that age differences in education perception accuracy were not due to ethnicity.

For extraversion perception accuracy, the model that included age group only was the best fitting model and was 35.9 times more likely than the model that included ethnicity only (BF₁₀ = 39.48/1.10 = 35.9), 3 times more likely than the model that included age group and ethnicity (BF₁₀ = 39.48/13.14 = 3.0), and 25.5 times more likely than the model that included age group, ethnicity, and their interaction (BF₁₀ = 39.48/1.55 = 25.5). Moreover, age group was the only effect that resulted in greater posterior odds, BF_inclusion = 17.18. Together, these results suggest that age differences in extraversion perception accuracy were not due to ethnicity, as ethnicity effects did not improve model probability.

For social perception accuracy, the model that included age group, ethnicity, and their interaction was the best fitting model but was not a better fit than the models that included ethnicity only and age group and ethnicity (all 1 < BF₁₀ < 3); in these models, ethnicity is the common factor. Moreover, only ethnicity resulted in improved model probability, BF_inclusion = 3.12., suggesting that ethnicity effects were more robust than age effects in predicting social perception accuracy. Thus, it is likely that the age group differences in social perception accuracy were due to effects of ethnicity, as age group was no longer predictive once ethnicity was modeled. Results indicated that Black perceivers were less accurate than White (BF₁₀= 19.85) and Asian perceivers (BF₁₀= 3.71).

Cognition.

For posed emotion perception accuracy, vocabulary resulted in the greatest model improvement (BF_inclusion = 2495.88), with age group (BF_inclusion = 988.03) and backward digit span (BF_inclusion = 25.00) also resulting in model improvement, albeit to lesser extents. All other cognitive variables resulted in lower posterior odds suggesting that they need not be included in the model. Consequently, we reran the model with only age group, vocabulary, and backward digit span. The model that included age group, vocabulary, and backward digit span was the best fitting model (BF₁₀ > 9325), and age group resulted in the greatest model improvement, BF_inclusion > 900,000, suggesting that age group differences in posed emotion perception accuracy were robust to effects of vocabulary and backward digit span.

For event valence perception accuracy, task version resulted in the greatest improvement to the model, BF_inclusion > 1,000,000, with forward digit span resulting in a smaller improvement to the model, BF_inclusion = 10.62. All other effects—including age group—resulted in lower posterior odds. We reran the model with age group, task version, and forward digit span. Although the model that included age group, task version, and forward digit span was the best fitting model, it was not more likely than the model that included task version and forward digit span only (BF₁₀ = ~ 4052/1968 = 2.06) suggesting that age group differences in event valence perception accuracy may have been due to individual differences in fluid cognition, specifically forward digit span.

For education perception accuracy, symbol substitution was the only effect that resulted in model improvement, BF_inclusion = 5.92; all other effects resulted in lower posterior odds. We reran the model with age group and symbol substation. The model with symbol substation only was the best fitting model and was 3.7 times more likely than the model with age group and symbol substation (BF₁₀ = 47.82/12.77 = 3.7). Moreover, age group did not result in model improvement BF_inclusion = 0.30. Together, these results suggest that age group differences in education perception accuracy may have been due to individual differences in fluid cognition, specifically symbol substitution.

For extraversion perception accuracy, symbol substitution was the only effect that resulted in model improvement, BF_inclusion = 9.53; all other effects resulted in lower posterior odds. We reran the model with age group and symbol substation. The model with symbol substation only was the best fitting model and was 4.5 times more likely than the model with age group and symbol substation (BF₁₀ = 55.21/12.22 = 4.5). Moreover, age group did not result in model improvement BF_inclusion = 0.23. Together, these results suggest that age group differences in extraversion perception accuracy may have been due to individual differences in fluid cognition, specifically symbol substitution.

For social perception accuracy, animal fluency was the only effect that resulted in model improvement, BF_inclusion = 28.64; all other effects resulted in lower posterior odds. We reran the model with age group and animal fluency. The model with animal fluency only was the best fitting model and was 3.3 times more likely than the model with age group and animal fluency (BF₁₀ = 2463.56/742.40 = 3.3). Moreover, age group did not result in model improvement BF_inclusion = 0.30. Together, these results suggest that age group differences in social perception accuracy may have been due to individual differences in crystallized cognition, specifically animal fluency.

Exploratory results summary.

To summarize the exploratory analyses, age group differences in social perception accuracy may have been due to ethnicity effects, as age group did not improve model probability whereas ethnicity did when both variables were included in the model. However, education and ethnicity did not appear to account for any of the other observed age differences in interpersonal accuracy. Importantly, except for posed emotion perception accuracy, the modeling of fluid and crystallized cognition resulted in age group effects having no improvement on the model. Instead, cognitive ability outperformed age group for event valence perception accuracy, education perception accuracy, extraversion perception accuracy, and social perception accuracy, suggesting that these interpersonal perception skills are better explained by individual differences in cognition than age.

Discussion

The main objective of the present study was to investigate the extent to which different interpersonal perception skills vary across adulthood. We extended past research on interpersonal accuracy by assessing a wider variety of interpersonal perception skills, including both emotion and non-emotion perception skills, within a single adult lifespan sample. Most of these skills were assessed using nontraditional laboratory tasks that elicited spontaneous expressions and behaviors in the same targets, thus overcoming methodological limitations associated with the use of posed, prototypic expressions and behaviors that may lack everyday relevance. Moreover, we utilized Bayesian analyses to quantify support for two competing hypotheses—that interpersonal perception skills vary significantly with age (the alternative hypothesis) and that interpersonal perception skills do not vary significantly with age (the null hypothesis)—further advancing past research on interpersonal accuracy.

As predicted, and in line with previous findings (e.g., Grainger et al., 2015; Isaacowitz et al., 2007; Malatesta et al., 1987; Paulmann et al., 2008; Spencer et al., 2016; Sullivan & Ruffman, 2004), older and middle-age adults were significantly less accurate than younger adults at identifying others’ posed emotional expressions; older adults were also less accurate than younger adults when identifying spontaneous emotional expressions, but only when judging the valence of an emotional event discussion. For other types of emotion perception skills based on spontaneous expressions and behaviors, age differences were not significant and instead there was greater support for age-related similarities than differences as indicated by the Bayesian analyses. Consequently, age differences in emotion perception accuracy were inconsistent: For some emotion perception skills (perceiving posed expressions; identifying global valence from spontaneous displays) age differences were observed, whereas for other emotion perception skills (perceiving disgust, satisfaction, amusement, interest, and enjoyment from spontaneous displays) age similarities were observed.

Similarly, our second hypothesis was partially supported. In line with recent findings on non-emotion perception skills (e.g., Bond et al., 2005; Vicaria et al., 2015; Zebrowitz et al., 2014), age-related similarities in accuracy were found for perceptions of health, agreeableness, and openness to experience, although the specific similarities (e.g., which perceiver age groups were similar) varied across perception skills. However, in contrast to our predictions, age differences were observed for the remaining non-emotion perception skills; specifically, middle-age adults were less accurate than younger adults at identifying others’ education and extraversion levels and in making judgments about different types of social interactions like kinship, intimacy, and dominance. Because the exploratory analyses indicated that some of these observed age differences in interpersonal accuracy were due to factors other than age (i.e., ethnicity, cognition), we discuss our findings below with regard to the effects of age, ethnicity, and cognition on interpersonal perception skill.

Interpersonal Accuracy: Evidence for Differences and Similarities across Adulthood

Both middle-age and older adults performed significantly worse than younger adults when identifying posed, prototypic expressions of emotion on a traditional laboratory task. This result was moderate in size and did not appear to be due to effects of education, ethnicity, or cognition, resulting in the conclusion that age differences in posed emotion perception accuracy are robust. This conclusion is consistent with past research: Previous studies that have found significant age differences in emotion perception accuracy have typically utilized traditional tasks that assess the perception of posed and prototypic emotional expressions (e.g., Ebner & Jonhnson, 2009; Isaacowitz et al., 2007; Sullivan & Ruffman, 2004).

Yet, our findings also provide evidence for age-related similarities in the ability to identify others’ spontaneous emotional expressions, particularly when skills are assessed using nontraditional tasks. To date, no research has tested the extent to which young, middle-age, and older adults’ emotion perception skills are statistically the same. Our study, thus, constitutes the first effort to do so. Specifically, we found evidence for age-related similarities in the ability to identify others’ spontaneous expressions of disgust, satisfaction, amusement, interest, and enjoyment; these perceptions were unrelated to age group as indicated by null hypothesis testing and provided greater support for the null hypothesis—that age groups are the same in accuracy—than the alternative—that age groups vary in accuracy—as indicated by Bayesian analyses. Consequently, we also conclude that not all emotion perception skills decline with age.

If some emotion perception skills worsen with age while others remain the same with age, what cumulative statement can be made about emotion perception accuracy and aging on a whole? Our results suggest that a single statement may not exist; instead, a more nuanced understanding of how emotion perception skills change with age is needed. For example, patterns of age-related differences vs. similarities may vary in their representation of interpersonal accuracy in everyday life. For example, despite theoretical claims that emotions are conveyed via specific configurations of facial movements (i.e., prototypic facial expressions; Ekman et al., 2002), such configurations are rarely spontaneously produced during naturalistic emotional situations, including those assessed in and outside of the laboratory (Castro et al., 2017; García-Higuera et al., 2015; Fernández-Dols et al., 2011). Given evidence that older adults perform better on laboratory tasks that include more ecologically valid and representative stimuli (e.g., Hess et al., 2001; Stanley & Isaacowitz, 2015; Sze et al., 2012), tasks that measure the perception of posed, prototypic emotional expressions likely exacerbate age differences by suppressing older adults’ skills.

Consequently, it may be more accurate to assess the perception of spontaneous emotional expressions, as these better reflect the nature of everyday emotion communication. However, our results suggest that doing so does not always result in an absence of age differences: That is, we found evidence for both age-related differences and similarities for emotion perception skills assessed using nontraditional tasks depicting spontaneous emotional expressions and behaviors. Thus, our findings are most consistent with the mixed evidence—both finding and failing to find significant age differences—in the literature on spontaneous emotion perception accuracy (e.g., Grainger et al., 2015; Krendl & Ambady, 2010). It remains unclear why these mixed patterns exist.

One possibility concerns the way in which individuals make accurate judgments about others’ emotions. For example, in the present study, age differences in emotion perception accuracy were found when perceivers were asked to make categorical judgments about others’ emotions—identifying the discrete emotion category being expressed in the GERT-S and identifying the valence of the event being discussed in the emotional event judgment task. In contrast, evidence of similarities in emotion perception accuracy were typically found when perceivers were asked to make dimensional judgments about others’ emotions—identifying the degree to which someone felt satisfied, disgusted, amused, interested, and enjoyment. These different judgment methods may vary in their relevance to everyday life. Although it is possible that real-life emotion perception accuracy involves identifying whether someone feels angry or sad, it is also possible that individuals engage in non-categorical judgments about others’ emotions, including judgments about the degree to which an individual is displaying a given feeling, thought, or behavior. For example, we might know that someone feels sad, but we may need to accurately identify how sad they feel so that we can respond in the most effective way—by quietly offering our condolences (if only a little sad) or by sitting with the person and providing extended support (if extremely sad). Indeed, recent research suggests that perceptions of others’ emotions are complex and multidimensional, involving the attribution of multiple emotions that are thought to be displayed at varying levels (e.g., Kim, Geren, & Knight, 2015; Riediger, Voelkle, Ebner, & Lindenberger, 2011). The extent to which categorical and dimensional perceptions of emotion vary in frequency and importance in everyday life may have implications for whether we observe patterns of age differences or similarities in emotion perception accuracy.

Alternatively, these different judgment methods may vary in their required effort, with age differences emerging for accuracy on more effortful judgments, such as identifying which of 14 different emotions is being expressed, and age similarities emerging for accuracy on less effortful judgments, such as identifying the extent to which a specific emotion is being expressed. Given that we did not systematically vary judgment type—categorical or dimensional—and stimuli type—posed and spontaneous—across our emotion perception accuracy tasks, we cannot test these possibilities directly in the present study. Nonetheless, future studies may extend our findings to investigate the extent to which these and other methodological features may contribute to different aging patterns in emotion perception accuracy.

We observed similar discrepancies in the aging patterns for non-emotion perception skills, all of which were assessed using nontraditional laboratory tasks (i.e., measuring the perception of various social and personal qualities from spontaneous, unscripted expressions and behaviors). In contrast to the emotion perception accuracy results, however, age differences in non-emotion perception skills were observed for both categorical judgments—for example, identifying someone’s education level—and dimensional judgments—for example, identifying someone’s extraversion. Moreover, we found evidence for age-related similarities in non-emotional dimensional judgments, including judging others’ health, agreeableness, and openness to experience. Collectively, these findings suggest that the methodological distinctions among categorical and dimensional judgments that may contribute to different aging patterns for emotion perception accuracy may not underlie different aging patterns for non-emotion perception accuracy.

How then can we explain the different aging patterns observed for non-emotion perception skills? One possibility lies in the specific patterns found: Where significant age effects were observed, differences in accuracy emerged between younger and middle-age adults only. These findings are somewhat consistent with past research on non-emotion perception skills and aging which has typically failed to find significant differences between younger and older adults’ skills (e.g., Bond et al., 2005; Vicaria et al., 2015; Zebrowitz et al., 2014). Our results additionally suggest that middle-age adults did not vary significantly from older adults in their abilities to perceive non-emotional expressions and behavior. Indeed, the majority of the evidence for age-related similarities in non-emotion perception skills were between middle-age and older adults. Most studies on interpersonal accuracy and aging do not include middle-age adults, resulting in more limited tests of aging patterns. If our study had included only young and older adult perceivers, we may have observed a more consistent pattern suggesting age-related similarities across non-emotion perception accuracy tasks. By including a more comprehensive adult lifespan sample, thus, our results augment previous findings to provide a more accurate picture of how non-emotion perception skills changes across adulthood. Because so few studies have examined age differences in the ability to perceive non-emotional qualities, future studies should aim to replicate our findings. Importantly, the age differences in interpersonal accuracy observed between younger and middle-age adults appear to be explained by other individual difference factors—cognition and ethnicity. We describe and interpret these effects below.

Why Age is Not Enough: Effects of Cognition and Ethnicity on Interpersonal Accuracy

Across the fourteen different interpersonal perception skills examined in the present study, we found evidence for age-related similarities in eight skills and evidence for significant age differences in only five skills; most of these differences emerged between younger and middle-age adults. If aging is in fact predictive of declines in interpersonal accuracy, there should be a linear negative relation between age and interpersonal accuracy. Consequently, we should have observed not only a greater number of age differences across interpersonal perception skills but also greater differences between older adults and the two younger age groups.

Our exploratory analyses suggest that other factors beyond age may be at play in predicting individual differences in interpersonal accuracy. Specifically, results suggested that four of the five significant age effects were better explained by individual differences in cognitive ability. That is, for event valence perception accuracy, education perception accuracy, extraversion perception accuracy, and social perception accuracy, the addition of cognitive factors resulted in age group no longer improving model probability; in these models, cognitive ability resulted in the largest change from prior to posterior odds whereas age group resulted in lower posterior odds, indicating that cognitive ability—and not age—predicted perceivers’ interpersonal perception skills. Importantly, our exploratory analyses did not indicate that age differences were qualified by cognitive ability, suggesting that the two factors did not interact to influence interpersonal perception skill; instead, the variance in interpersonal accuracy explained by age appeared to be accounted for by cognitive ability.

Our study is not the first to consider whether age differences in interpersonal accuracy are due to individual differences in cognition. Developmental models have long emphasized that effects of age are not due to age itself but rather to other factors that covary meaningfully with age (e.g., Baltes, Reese, & Nesselroade, 1977), such as cognitive ability (Ruffman et al., 2008). Indeed, many aging studies have found positive links between interpersonal perception skill and cognitive ability (e.g., Krendl & Ambady, 2010; Krendl et al., 2014; Sarabia-Cobo et al., 2015).

Our results add to these findings and suggest not only that age differences in interpersonal accuracy may be explained by individual differences in cognition but also that specific cognitive abilities are related to specific interpersonal perception skills. For example, we found evidence for the role of fluid cognitive abilities (measured by the forward digit span and symbol substitution tasks on the WAIS-R) on the ability to identify the valence of an event being discussed and in accurately identifying others’ education and extraversion levels. Moreover, we found evidence for the impact of crystallized cognition (measured by verbal fluency) on the ability to accurately identify different types of social interactions including status, intimacy, kinship, competition, and deception. Consequently, associations between cognition and interpersonal accuracy may reflect unique pathways, such that some cognitive abilities might be more (or less) relevant for different interpersonal perception skills. We note this as an especially promising area for future research, and hope that studies will examine both a diversity of interpersonal perception skills and cognitive abilities to better delineate specific associations between these two constructs.

In addition, we also explored whether age differences in interpersonal accuracy were due to demographic characteristics that correlated with age in our sample, specifically perceivers’ education and ethnicity. Although we found no evidence that age differences in interpersonal accuracy were due to or qualified by perceivers’ level of education, we found some evidence that age differences in interpersonal accuracy may be due to perceivers’ ethnicity. Specifically, ethnicity effects were more predictive of social perception accuracy than age group; indeed, ethnicity improved model probability whereas age group did not. This finding, coupled with our exploratory cognitive results, suggest that differences between younger and middle-age adults’ social perception skills are better explained by factors other than age, specifically ethnicity and/or verbal fluency. Because so few interpersonal accuracy studies include middle-age adults and diverse samples that allow for a lifespan comparison and the examination of ethnicity effects, it is important that future research replicate our findings; effects of age on interpersonal accuracy may be very well be explained by other psychological processes.

Broad Implications for Interpersonal Accuracy Research

Although our results clearly contribute to the literature on interpersonal accuracy and aging, they also hold implications for research on social and emotional perception skills more broadly. For example, the results from the many different interpersonal accuracy tasks created for the present study demonstrate that adults can accurately perceive a wide range of social and emotional qualities in others. Indeed, the final set of 14 interpersonal perception skills—which varied significantly from chance and demonstrated construct and predictive validity through associations between skills and with cognitive ability and personality traits—represents a more diverse array of skills than is typically examined in studies on interpersonal accuracy. Indeed, the skills assessed in the present study were deliberately exploratory and eclectic to provide a more diverse and comprehensive assessment of interpersonal accuracy than examined in previous research. These skills thus represent some but likely not all of the interpersonal perception skills of interest to researchers. We hope that our results will inspire researchers to consider the corpora of interpersonal perception skills individuals may employ in daily life and, when possible, to assess as many of these skills within a single study as doing so we will afford a more comprehensive assessment of the broad construct of interpersonal accuracy.

In addition, we believe that the tasks developed for the present study may prove useful tools for future studies that seek to measure individual differences in interpersonal accuracy. Importantly, the tasks were created using nontraditional stimuli that display spontaneous expressions and behaviors in real people. Such an approach is considered more ecologically valid than traditional tasks that rely on posed, prototypic expressions and behaviors (Costanzo & Archer, 1993; Isaacowitz & Stanley, 2011). Because spontaneous expressions tend to be more common than posed, prototypic expressions, at least in the emotion domain (e.g., Castro et al., 2017; García-Higuera et al., 2015; Fernández-Dols et al., 2011), results based on the new nontraditional tasks may better represent the nature of interpersonal accuracy in real life.

Finally, the new nontraditional tasks may be useful in overcoming challenges associated with ceiling effects in interpersonal accuracy. Nontraditional tasks typically include stimuli that are by nature more difficult to perceive—spontaneous expressions and behaviors that are more subtle, mixed, and fragmented and thus may lack the signal quality of more traditional stimuli (i.e., posed or prototypic expressions and behaviors). Indeed, the average accuracy levels for skills measured using nontraditional tasks (e.g., enjoyment perception accuracy, social perception, accuracy) were often (although not always) lower than the accuracy level for skills assessed using a traditional task (posed emotion perception accuracy measured with the GERT-S). Because nontraditional tasks may assess more challenging interpersonal perception skills, they may allow for a finer discrimination among individuals varying in interpersonal perception skill. For example, individuals who are highly skilled in interpersonal accuracy may perform similarly to individuals with average levels of interpersonal accuracy on traditional laboratory tasks that present posed, prototypic expressions, as such expressions are often very easy to perceive. In contrast, the use of spontaneous expressions may provide for finer discriminations among individuals who are expert perceivers and those who are good or average perceivers.

Limitations and Conclusions

Our study is not without limitations. As noted briefly above, our study was not designed to test the extent to which age differences and similarities in interpersonal accuracy are explained by specific task features—i.e., posed vs. spontaneous expressions, categorical vs. dimensional judgments. Thus, our interpretations of aging patterns as reflective of methodological context are tentative, requiring further replication and systematic investigation. However, we believe our study is an important first step in this direction, as no studies have examined such a wide array of interpersonal perception skills within the same sample.

As noted above, nontraditional tasks may measure more challenging interpersonal perception skills than traditional tasks, resulting in the potential for constrained variance (i.e., if the task is similarly difficult for everyone). However, there was variability in interpersonal perception skills measured using nontraditional tasks, and this variability was predictive as demonstrated through the associations with cognition and personality (see Table 3). Thus, null age effects on nontraditional interpersonal accuracy tasks were likely not due to issues of item difficulty and constrained variance.

In sum, findings from the present study help to clarify the extent to which interpersonal perception skills become worse, better, or remain the same with age. By using a more diverse and ecologically-valid methodological approach and advanced statistical technique, we found that aging patterns vary across interpersonal perception skills: Some skills appear to get worse with age whereas other skills appear to be maintained with age. The skills that appear to get worse with age may actually reflect other psychological processes, including cognition and ethnicity, suggesting a need for researchers to consider these factors and the role they may play in predicting interpersonal accuracy. At the very least, aging itself does not seem to be as strongly or uniformly associated with declines in interpersonal accuracy as past research suggests. Future research will prove useful in helping to disentangle different aging patterns from effects relating to methodology. We believe that such studies, coupled with our results, will provide a more accurate and nuanced understanding of interpersonal accuracy in adulthood.