Robustness of the aging effect of smiling against vertical facial orientation

Introduction

The face is a valuable source of information for social communication, and humans have developed specific processing methods for others’ faces. For example, humans immediately lose the ability to recognize other's faces when the orientation or parts' position of the face becomes unnatural (Tanaka & Farah, 1993; Yin, 1969). These distortions in face processing highlight that the face is processed holistically in contrast to other visual stimuli. Therefore, we cannot identify others’ faces and judge facial expressions when holistic processing is inhibited (e.g., Rakover, 2013). This holistic processing of the face creates complex interactions between multiple factors, such as the interaction between emotion and gender (Atkinson et al., 2005).

Age is among the crucial information obtained from the face. We generally estimate a person’s age from their faces and accordingly change our attitude and manner of speaking (Ryan et al., 1986). Among the many information dimensions that can easily be extracted from a face, age is considered the primary dimension (George & Hole, 1998). Thus, accurate age identification is crucial in determining social roles and facilitating social interaction. However, various factors distort age perception (see Moyse, 2014). Previous studies have specifically focused on the effects of gender and race on perceived age (Dehon & Brédart, 2001; Nkengne et al., 2008).

Interestingly, several studies have reported that humans have a counterintuitive bias regarding age. We associate smiling with youth, that is, it is generally believed that when people see a smiling person, they feel that person is younger. Indeed, previous research has provided evidence that individuals with a smile appear younger than those with other facial expressions (Hass et al., 2016; Voelkle et al., 2012). However, contrary to the commonly held association between smiling and youth, Ganel (2015) showed that a smiling face is estimated to be older than a neutral face. This phenomenon, in which smiling faces are evaluated as being older than neutral faces, is called the aging effect of smiling (AES; Ganel & Goodale, 2021). AES is attributed to wrinkles around the eyes caused by smiling (Ganel, 2015; Ganel & Goodale, 2021). In contrast, when participants were asked to retrospectively estimate the mean age of several faces (i.e., face group), they estimated that the smiling face group was younger than the neutral face group (Ganel & Goodale, 2018). These studies indicate that the effect of emotional expressions on age estimation depends on the method of estimation (i.e., directly or retrospectively).

Recently, our study showed that AES was consistently confirmed regardless of the stimulus or participants’ race or culture (Yoshimura et al., 2021). Specifically, smiling faces were estimated to be older than neutral faces for both Swedes and the Japanese. In contrast, participants in both countries estimated the smiling faces to be younger when estimating the age retrospectively. These results suggest that AES is robust across cultures, although the direction of this effect changes depending on the task.

The AES mechanism, however, remains unclear. What can be assumed is that AES is associated with some type of characteristic information of face perception or emotion processing that accompanies changes in facial expressions. Age is considered a relatively primary piece of information extracted from the face (Bruce & Young, 1986), and previous research suggests that age perception may be based on facial surface or shape information (George & Hole, 2000). However, other studies have suggested that facial expression processing relies on holistic processing (Maurer et al., 2002; Tanaka & Sengco, 1997). Since previous studies have reported that smiling is associated with younger age (Hass et al., 2016; Voelkle et al., 2012), AES may be mediated by smiling. When holistic processing of facial expressions is inhibited, the scope of faces that the observer can process is constricted, and facial features are processed sequentially and independently (Rossion, 2009). Additionally, a previous study suggested that AES is driven by wrinkles around the eyes produced by smiling (Ganel, 2015). Therefore, if inverted faces that inhibit holistic processing are age estimated, the effect of local feature processing can be highlighted in inverted smiling faces. In such cases, inverted smiling faces could be evaluated as older than upright, smiling faces.

To extend AES findings, the present study examined whether AES was mediated by smiling. Here, we used inverted faces in the experiment because they inhibit the holistic processing of facial expressions while maintaining visual information (Rakover, 2013). Specifically, we divided the participants into two groups: one observing upright faces and the other observing inverted faces. They were asked to estimate the ages of smiling, neutral, and surprised faces. The estimated age for each facial expression was then compared between the groups. Even if AES was boosted with an inverted smiling face, we could not rule out the possibility that inversion itself has the effect of making the facial expression stimulus older. Hence, we set a surprised face as the control condition because the surprised expression produces facial morphological changes other than the smiling condition, which has been used in previous studies (Ganel & Goodale, 2018; Yoshimura et al., 2021). Comparing smiling faces with neutral or surprised faces allows us to distinguish whether AES is mediated by smiles rather than by facial morphological changes. If the enhancement of AES was due to the prioritization of local information processing for the smiling expression, as holistic processing was suppressed, there would not be a significant difference between facial orientations in the surprised face.

Method

Results

From the internet protocol addresses collected in the experiment, we checked whether any individuals participated in both the upright and inverted face conditions. In total, 104 Japanese people (52 people per group) were recruited to participate in the online experiment (51 males, 51 females, and two non-respondents; M_age = 29.56, SD = 7.04). Duplicate data were excluded from the analysis because we could not confirm whether they were answered by different individuals. We also excluded data from participants who answered the DQS incorrectly. The final number of valid data used in the analysis was 98 (we excluded data from six participants) (Yoshimura, 2022). We also removed trials where the estimated ages were outside ± 2.5 SD from the participants’ mean in each condition.

Figure 1 shows the distribution of the mean estimated age for each expression in the upright and inverted face conditions. We conducted a two-way mixed ANOVA with facial expression (smiling, neutral, or surprised) as the within-subjects factors and facial orientation (upright or inverted) as the between-subjects factors. The results revealed a main effect of facial expression (F(2, 189) = 23.11, p < .001, η²_G = .01). However, the main effect of facial orientation (F(1, 96) < 0.000, p = .98, η²_G < .001) and the interaction between facial expression and facial orientation (F(2, 189) = 1.33, p = .27, η²_G = .01) were not significant. Based on the main effect of facial expression, we also conducted a Scheffé multiple comparison test of the facial expressions. The results showed that participants estimated smiling faces to be significantly older than neutral faces (t(96) = 5.52, p < .001, d = 0.56). In addition, the results also showed that they estimated surprised faces to be significantly older than neutral faces (t(96) = 6.40, p < .001, d = 0.65).

Figure 1. Violin and box plots show the mean estimated age distribution for each expression by facial orientation.

The black dots represent the outliers for each facial expression. p < 0.001 = ***.

We performed an equivalence test (Lakens et al., 2018) for facial orientation as a post-hoc and exploratory analysis. We set equivalence bounds to ± 0.5, as the medium effect size (i.e., Cohen's d). The results showed that the mean estimated ages in the upright and inverted conditions were significantly equivalent (t(95.34) = 2.453, p < 0.01).

Discussion

The present study aimed to examine how holistic processing of facial expressions contributes to AES. In the experiment, we asked two groups of participants (given upright or inverted faces) to estimate the age of each facial expression; we then compared the estimated ages. The results showed that smiling faces were estimated to be older than neutral faces, indicating that AES was replicated. However, there was no significant difference in the estimated age between the upright and inverted conditions. More importantly, AES was confirmed even when inverted faces were presented. We predicted that inverted smiling faces would be evaluated as even older than upright faces if AES were mediated by smiling. However, this analysis revealed no significant effect on facial orientation. Furthermore, we conducted an equivalence test for facial orientation as an exploratory and post-hoc analysis and found that the mean estimated ages in both the upright and inverted conditions were significantly equivalent. Thus, these results suggest that AES is insensitive to holistic processing.

Given the results of this experiment, we assumed that holistic processing of emotional expressions is insufficient to modulate AES significantly. Another unexpected result was an increase in the estimated age of surprised faces. This result may reflect the contribution of perceptual processing to AES occurrence. Previous studies have reported that AES is affected by changes in the skin surface and other facial parts (e.g., wrinkles around the eye region) over a lifespan (Ganel, 2015; Ganel & Goodale, 2018, 2021). Considering age estimation results for smiling and neutral faces, AES was less affected by face inversion. Therefore, age estimation for facial expressions is considered insensitive to holistic processing. Additionally, a previous study reported no decrease in the accuracy of age estimation, even when participants estimated the age of inverted, negation, or blurred face stimuli (George & Hole, 2000). These results highlight the possibility that AES is processed based on perceptual analysis of facial features, that is, the structural encoding stage (Calder & Young, 2005). From this point of view, the surprising faces used in this study also have wrinkles on the forehead, and this feature may produce an aging effect on facial expressions. However, it should be noted that age estimation is not only determined by facial features, as a previous study indicated that age estimation also involves holistic processing using the composite paradigm (Hole & George, 2011).

Another possible factor is that the attractiveness of facial parts may affect age estimation. Some previous studies have reported that masked faces were more attractive or vice versa (Hies & Lewis, 2022; Miyazaki & Kawahara, 2016; Patel et al., 2020). The fact that age could be estimated only for the upper half of the face and that age estimation did not depend on holistic processing leads to speculate the potential involvement of attractiveness. Specifically, the change in shape due to facial expressions may have reduced the attractiveness of the parts, thereby altering the apparent age. Therefore, the attractiveness of facial parts is also worth considering in further studies, such as whether it is indeed involved and, if so, what the causal mechanism entails.

The present study's findings indicate that AES is based on more primary and local facial features and extends studies on AES (Ganel, 2015; Ganel & Goodale, 2018, 2021; Yoshimura et al., 2021). However, it remains unclear why smiling faces are rated as younger in memory-based age estimation. Age estimation for memory representations of faces may be processed through different mechanisms than perceptual representations of faces in the process of dissociated facial processing (Weigelt et al., 2014). Furthermore, such a bias in opposite directions in memory and perception for identical stimuli has been observed in spatial processing (Yamada et al., 2011). Future research should further examine these questions.

Data availability