Facial emotion recognition (FER), one of the essential components of social
cognition (Adolphs, 2001), represents the ability to recognize facial emotional
expressions. It enables individuals to sense their social environment and modify their
behavior accordingly (McCade, Savage, Guastella, Lewis, & Naismith, 2013); it also
contributes to more efficient social interactions (Sze, Goodkind, Gyurak, & Levenson,
2012). Thus, this ability is undoubtedly crucial for social behavior (Hargrave,
Maddock, & Stone, 2002); furthermore, engaging in satisfying social interactions and
avoiding social isolation are important to our health and well-being throughout life
(Cacioppo, Berntson, Bechara, Tranel, & Hawkley, 2011). Consequently, deficits in
this ability may contribute to difficulties in social communication, damage
self-esteem, and even diminish the quality of life (Ciarrochi, Chan, & Caputi, 2000).
FER has drawn considerable attention in clinical and functional imaging studies
recently. Studies have demonstrated that dissociable neural substrates are associated
with the facial recognition of basic emotions (Hennenlotter & Schroeder, 2006;
Schroeder et al., 2004). The occipital and posterior temporal cortices are responsible
for the perceptual analysis of facial expressive features (Haxby, Hoffman, & Gobbini,
2000; 2002), and the extraction of emotional meaning from faces is linked to the
orbitofrontal, ventral prefrontal cortex-related, and somatosensory regions (Adolphs,
2002). However, these emotional circuits, including the hippocampus, amygdala, and
frontal regions, were reported to show age-related neurological changes (Greenwood,
2000). In addition, certain types of neurodegenerative diseases, such as
frontotemporal dementia (FTD) and Alzheimer’s disease (AD), can also damage these
brain regions (Keane, Calder, Hodges, & Young, 2002; Pietschnig et al., 2016). Thus,
deficits in decoding specific emotions have been reported in normal aging as well as
in patients with neurodegenerative diseases (Keane et al., 2002; Torres et al., 2015).
To help families realize the patients’ difficulties and improve their life quality,
choosing an appropriate clinical assessment for early detection of deficits in FER is
undoubtedly crucial.
Furthermore, studies have indicated that several characteristics of emotional
stimuli could affect the accuracy and memory of FER, including cultural, gender-, and
age-based factors (Bäckman, 1991; Hess, Blairy, & Kleck, 1997; Malpass & Kravitz,
1969; Wells, Gillespie, & Rotshtein, 2016). Indeed, the own-race bias refers to the
tendency of recognizing and memorizing one’s own race or ethnicity relatively more
accurately than another race or ethnicity (Malpass, & Kravitz, 1969). Gender has also
been reported to have different effects depending on the type of expressions (Wells et
al., 2016); for example, female faces are reported to be easier to recognize with regard
to expressions of happiness (Hess et al., 1997), while male faces are better recognized
in expressions of disgust, sadness (Hess et al., 1997), and anger (Becker, Kenrick,
Neuberg, Blackwell, & Smith, 2007). To our knowledge, very few studies have
examined the effects of photo age on FER. This is why studies conducted so far in the
domain of age group differences in processing emotional expressions have mostly
used younger faces (some included middle-aged faces) but did not systematically vary
the age of the presented faces. However, the study by Lamont, Stewart-Williams, and
Podd (2005) using neutral faces as stimuli found that observers of different ages
recognize faces of their own age more accurately and rapidly as opposed to those of
other ages (referred to as the own-age bias; Bäckman, 1991). Such findings suggest
that the age of a face constitutes an important factor that influences how we attend to,
encode, and remember faces. Evidence of the own-age bias challenges any
interpretation of observed age group differences in FER, as older observers may have
been at a disadvantage relative to younger observers when the stimuli consisted only
of faces of young individuals.
The own-age effect (in most studies called own-age bias or own-age advantage, while we used the term “own-age effect” because we did not want to emphasize it as
good or bad) is explained by two main theories: experience (or expertise) accounts
(Rhodes & Anastasi, 2012) and social-cognitive accounts (Sporer, 2001). The former
means that more experience and contact with own-age groups increases the
individual’s familiarity with the expressive style of own-age faces, and thus, decoding
of own-age faces is more efficient (Rhodes & Anastasi, 2012). The latter means that
there is a greater motivation to process and attend to the characteristics of own-age
faces (Sporer, 2001); thus, individuals who identify with an ethnic or social group will
exert more effort when decoding the emotional expressions of the own-group
(Thibault, Bourgeois, & Hess, 2006). The own-age effect was initially proposed in
facial recognition memory studies, indicated that facial recognition memory is
superior for own-age relative to other-age faces (Bäckman, 1991; Lamont et al., 2005;
Wright & Stroud, 2002). Further studies have also observed the own-age phenomenon
in tasks that involve recognizing facial emotional expressions across different fields.
For example, participants tended to look longer at own-age faces, and this was
thought to predict more accurate FER in own-age faces (Ebner, He, & Johnson, 2011).
Functional magnetic resonance imaging studies also reported different activities for
own-age and other-age faces regarding neutral and happy expressions (Ebner et al.,
2013). In addition, studies that used electroencephalography reported partly own-age
and own-race effects on the event-related potentials for neutral expressions (Melinder,
Gredebäck, Westerlund, & Nelson, 2010). Based on these empirical evidence and
theories, it may be assumed that own-age photos can enhance the accuracy of FER for
observers. That is, the own-age effect might appear in FER.
Indeed, this hypothesis has been proposed and investigated in several studies
over the last decade, and some have confirmed this effect in older observers. For
example, Riediger, Voelkle, Ebner, and Lindenberger (2011) used posed expression
with multi-dimensional response format and found that middle-aged and older
observers performed well in their target ratings of happiness and anger by the age of
the own-age photos than did young observers. Another study by Riediger, Studtmann,
Westphal, Rauers, & Weber (2014) which only used spontaneous and posed smile as
the test material also supported that older participants could better identify older
rather than younger faces.
However, contrary to the results of the above studies, most research that was
carried out by modifying the age of the photographed or videoed individuals indicated
that there was no own-age effect or that it was observed only for younger observers.
For example, Borod et al. (2004) presented younger, middle-aged, and older female
observers as stimuli, and the results showed that the expressions of older posers were
rated significantly less accurately than those of younger posers for all groups. Further
studies by Ebner and Johnson (2009), Murphy, Lehrfeld, and Isaacowitz (2010), and
Hühnel, Fölster, Werheid, and Hess (2014) also reported similar patterns. In addition,
Malatesta, Izard, Culver, and Nicolich (1987) found that this effect exists only in
younger observers. Older observers were better at rating older faces than they were at
rating younger faces, while the difference was not significant. A study by Richter,
Dietzel, and Kunzmann (2010) also supported this finding in younger observers.
Nevertheless, the results of these studies were inconsistent, and it should be
noted that some methodologic limitations existed in all these studies. First, the gender
of the stimuli and observers in some studies was exclusively female (Borod et al.,
2004; Hühnel et al., 2014; Murphy et al., 2010), even though it is known that gender
can influence the accuracy of the results based on the type of emotion (Wells et al.,
2016). Second, the numbers of photos and observers in some studies were too small
(Borod et al., 2004; Ebner & Johnson, 2009; Hühnel et al., 2014). Third, the target
emotions in these studies were inconsistent; besides, some examined the own-age
effect by averaging the accuracy of emotions (Malatesta et al., 1987). These factors
not only make it difficult to conclude the type of emotion which was reported
consistently enough to show the own-age effect, but also make it hard to analyze the
different effects of distinct emotions based on the finding that different types of
expressed emotions have different effects on accuracy (Wells et al., 2016). Therefore,
it is necessary to assess enough types of emotions and to examine their effects
separately rather than as averages. In conclusion, gender imbalance, small stimuli and
observer sample sizes, selecting incomparable types of emotions, and ignoring the
effect of different emotions were existing methodologic problems in prior studies, and
these might have resulted in inconsistent results. The present study sought to address
these methodological limitations of earlier investigations.
Apart from the problems we have mentioned above, other methodologic
differences existed might also cause inconsistent results, namely, the types of
emotional expressions presented (dynamic or static and posed or spontaneous),
measured approaches of response (the forced-choice approach and the
multi-dimensional response format), and the stimuli database. First, dynamic
spontaneous stimuli were reported to show more ecological validity; thus, they could
increase accuracy (Bartlett et al., 2006; Murphy et al., 2010), while the results of
examining the own-age effect were still inconsistent after controlling it (Murphy et al.,
2010; Riediger, 2014) due to other methodologic problems. Besides, the dynamic
spontaneous stimuli established so far did not include enough stimuli, and most were
female faces only (Murphy et al., 2010; Richter et al., 2010). Thus, there are no
appropriate stimuli that can be selected yet, even if we do not consider including the
East Asian faces. Second, the multi-dimensional response format, the way that
participants rate the percentage across all emotions within a photograph. And the
responses were considered as accurate if the percentage of the target emotion was
higher than the percentage on the remaining scales (Gunes & Pantic, 2010). It was
developed based on the theory that emotional experiences are often multi-faceted
(Hemenover & Schimmack, 2007), and so it was thought to be more appropriate
(Kreibig, Samson, & Gross, 2013). However, some studies that used the forced-choice
approach still confirmed the own-age effect successfully; thus, it seems that different
types of rating formats did not play an important role in the inconsistency of the
results. In addition, Hühnel et al. (2014) indicated that the hit rates in their study were
relatively low because of using the multi-dimensional response format. Although the
multi-dimensional response format was reported to show more ecological validity, the
forced-choice approach might be more appropriate for developing our task to a
clinical measurement. Finally, most studies used the static posed expressions of the
FACES database (Ebner, Riediger, & Lindenberger, 2010) as materials, while the
remaining studies used stimuli (including photos and videos) developed by their
respective laboratories (Borod et al., 2004; Hühnel et al., 2014; Malatesta et al., 1987;
Richter et al., 2010; Riediger, 2014); thus, the stimuli in those studies are
heterogeneous in nature. To control the influence of race on FER (Young &
Hugenberg, 2012), we chose the stimuli from Taiwanese individuals (Tu, Lin, Suzuki,
& Goh, 2018) and included a large number of static posed photos. In conclusion, we
determined to use the forced-choice rating as our response measurement, emotion
stimuli from Taiwanese individuals with static posed photos as stimuli.
Apart from the changes in the brain in normal aging, abnormal cerebral atrophy
and neuropathological changes occur in patients with AD, resulting in damage to the
circuits related to emotions (McLellan, Johnston, Dalrymple‐Alford, & Porter, 2008;
Spoletini et al., 2008). Thus, AD has been reported to result in deficits in FER, with
gradually increasing impairment, especially in specific emotions, as the disease
progresses (Pietschnig et al., 2016), and changes may be evident even in the early
phases (Virtanen et al., 2017). In addition, the onset of AD mostly begins at an age of
over 65 years. If the own-age effect exists in AD or the preclinical and prodromal of
AD patients, the clinical utility of the assessment protocol which uses younger faces
only would decline and underestimate the ability of older patients. Therefore, in
addition to healthy older adults, it is important to examine whether the own-age effect
exists in those with AD, and moreover, in the preclinical and prodromal AD patients.
However, no study has investigated whether the own-age effect exists in FER in the
preclinical and prodromal of AD, and AD patients. Therefore, most studies that
examined the FER performances in AD and the preclinical or prodromal AD patients
used stimuli either without varied age of photos or did not provide exact information
about the age and number.
As the preclinical and prodromal stages of AD respectively, subjective cognitive
decline (SCD) and mild cognitive impairment (MCI) have recently received attention.
The literature on the related neuropathological locations remains heterogeneous in
individuals with SCD. However, the recent study found that people with SCD had
higher amounts of neurotic amyloid plaques evident in the medial temporal lobes and
neocortex regions (Studart Neto & Nitrini, 2016). Accordingly, it might be possible
that the underlying neuropathologic changes have partially influenced the FER
performances in individuals with SCD. However, only one study has investigated
FER performance between adults with SCD and healthy adults, and the results
showed no difference (Pietschnig et al., 2016). The study used the Vienna Emotion
Recognition Tasks (36 pictures, including 6 individuals with anger, disgust, fear,
happiness, sadness, and neutral facial expressions) (Derntl, Kryspin-Exner, Fernbach,
Moser, & Habel, 2008; Gur et al., 2002) with an equal number of photos of both
genders as stimuli but younger faces only.
Several studies have indicated emotion-specific deficits in patients with MCI;
different stimuli were used in these studies. For example, Fujie et al. (2008) found
that patients with MCI showed deficits in decoding sadness and anger, while Spoletini
et al. (2008) indicated an impairment only in decoding low-intensity stimuli,
especially in fearful faces. The former study used the Facial Expressions of Emotion:
Stimuli and Tests (FEEST) (60 pictures, including 6 females and 4 males for six basic
and neutral emotions) (Young et al., 2002) as stimuli. The latter used the Penn
Emotion Recognition Test (ER40) (40 pictures, including 4 female faces and 4 male
facial expressions of happiness, sadness, anger, fear, and neutral) (Gur et al., 2002) as
stimuli. Both the FEEST and the ER40 have mentioned that their photos were
controlled for the photo age, while no further information was presented. Moreover,
Weiss et al. (2008) also used ER40 as stimuli and indicated that patients with single
domain (sd)-MCI did not have significantly altered emotion recognition abilities, and
only multiple domains (md)-MCI patients showed impairments in recognizing sad,
fearful, and neutral faces. This observation of deficits only in md-MCI and not
sd-MCI was also supported by Teng, Lu, and Cummings (2007) and Varjassyová et al.
(2013), but the results of these studies did not examine distinct types of emotion;
therefore, we do not know which types of emotions showed deficits. The stimuli used
by Teng et al. (2007) was the Florida Affect Battery (FAB; 20 pictures, including 4
females of happy, sad, anger, fear, neutral) (Bowers, Blonder, & Heilman, 1998), and
the stimuli used by Varjassyová et al. (2013) were only 4 faces (gender was not
mentioned) for six basic and neutral emotions from FEEST; both studies did not
mention the age in their photos.
From the above data, we find that these studies did not put much emphasis on the
effect of photo age. As it cannot be said that the effect of photo ages was controlled in
these studies, we can assume that the inconsistent results might be partly due to not
considering the own-age effect. Besides, as we have mentioned that no research has
investigated whether the own-age effect in FER exists in patients with SCD, aMCI,
and very mild AD. Therefore, it is necessary to examine whether the own-age effect
exists in these patients before investigating their performances.
The first aim of our study was to control the prior methodologic differences and
limitations and then to investigate whether the own-age effect in FER exists in healthy
elderly adults when considering the different effects of distinct emotions. The second
aim extended to patients with SCD, aMCI, and very mild AD; we first questioned
whether the own-age effect in FER exists in patients and then investigated their
performances in FER in case of different types of emotions. Finally, we used the
emotion stimuli from Taiwanese (Tu et al., 2018) individuals. As it is the first face
emotion database from the Taiwanese population, we collected participants to rate the
intensities and accuracies of these photos, and explored the clinical utility of this test
for further study to develop a FER assessment.