Experiment 1 is the first in a series of three experiments that were designed to
investigate emotion recognition in older adults (OAs) using tasks that were controlled for possible methodological confounds. The subsequent experiments are reported in Chapters 5 (non-verbal vocalisations) and 6 (single words) but this chapter focuses on static faces. Each experiment included an emotion recognition task as well as a non- emotion task with both tasks closely matched on tasks demands. All three chapters need to be considered together to give a broad understanding of emotion recognition ability in OAs (see Chapter 6). Importantly, compared to YA participants the OA participants were not disadvantaged in many aspects that may influence emotion recognition ability such as extroversion, empathy, alexithymia, and intelligence. Thus findings can be more confidently attributed to emotion recognition ability rather than previous research that has not accounted for these abilities.
To allow for comparisons with previous research it is important for the current study to include an emotion recognition task using static facial expressions. The
literature in the field provides evidence that OAs, compared to YAs, have difficulties in recognising emotions from faces (e.g., Demenescu, Mathiak & Mathiak, 2014;
Grainger, Henry, Phillips, Vanman, & Allen, 2015; Sasson et al., 2010). Furthermore, a meta-analysis suggests that OAs are less accurate than YAs at recognising the negative emotions of sadness, fear, and anger (Ruffman, Henry, Livingstone, & Phillips, 2008). However, it should be noted that not all studies report emotion recognition deficits for all three emotions. For instance, as shown in Table 1.1 (Appendix 1.1) of the 18 studies that measured emotion recognition of fear, anger, and sadness from static faces only six studies reported recognition deficits in OAs, compared to YAs, across all three emotion types. Moreover, in the same 18 studies OAs were less able than YAs to recognise fear or anger in 11 studies and deficits for sadness were observed in 15 studies. Whilst the
evidence suggests that OAs are less able than YAs to identify at least one negative emotion the specific emotion deficit can differ between studies. The fact that some studies failed to find emotion recognition difficulties in OAs for fear and anger, and to a lesser extent sadness, demonstrates that some inconsistencies exist in the field regarding which emotion types OAs are less able to recognise than YAs.
The observed between-study variations for age-related recognition deficits in OAs for fearful, angry, and sad faces also extend to other emotion types including happiness and disgust. For example, OAs are less able than YAs to recognise happiness from static faces in four out of 18 studies included in Table 1.1, whereas other findings suggest there are no differences in recognition ability for happiness between OAs and YAs. Regarding disgust recognition from static faces four studies indicate that OAs are more able than YAs, four other studies state that OAs are less able than YAs, and ten studies report no differences between the age groups (see Table 1.1). Therefore, OAs’ recognition ability for disgusted and happy faces remains unclear.
The variation in findings in the field may reflect methodological differences between studies, such as the number of response options, and these limit cross-study comparisons. Furthermore, few studies attempt to account for many of the potential differences in sample characteristic between OAs and YAs. It is possible that these methodological and sample differences can account for some of the between-study inconsistencies regarding which emotions are difficult for OAs to recognise compared to YAs.
Methodological differences can include variations in task demands such as: time constraints (self-paced versus restricted time); working memory demands (whether or not the face is visible when the response is given); and processing demands (more forced-choice options requires increased processing demands as additional options need to be eliminated before arriving at a decision). The established decline in several
cognitive domains with natural aging (e.g., Hartshorne & Germine, 2015) may mean that OAs are disadvantaged more than YAs with an increase in the task demands. Thus, OAs may have lower emotion recognition ability than YAs when tasks are more
demanding.
Given the possible effect of task demands on emotion recognition accuracy it is important to try to isolate the ability to meet task demands from emotion processing. A non-emotion task can be used to measure processing ability in the absence of emotion content. However, as documented in previous chapters comparisons between emotion and non-emotion tasks are best achieved when the tasks share similar task demands, procedures, and stimuli. Essentially the main difference between the current emotion and non-emotion tasks was the type of judgement that participants were required to make, for instance whether between emotion types or by sex and age group. To the researcher’s knowledge the current experiment is the first to use an emotion and non- emotion task that have been carefully designed to reduce the methodological differences between the tasks as far as possible.
In addition to the differences in methodology, between-studies dissimilarities in sample characteristics including cognitive abilities, emotional skills, or personality traits may account for the variability in findings in the field, as is discussed below. If sample differences are not understood then it is difficult to explain any age-related emotion recognition deficits in OAs as a function of difficulties in emotion processing, as the impairments may arise from other confounding group differences. Unfortunately, there is inconsistency across studies in the field regarding which cognitive, emotion, or personality aspects are measured alongside emotion recognition ability (see Table 3.1 in Appendix 3.1). These omissions present two main challenges: First, it is difficult to understand which characteristics may affect emotion recognition ability in OAs. Second, it remains unclear whether findings in some previous studies might be better
explained by age-related differences in a variable other than emotion recognition ability, per se.
Many studies in the area have, however, included a measure of education; education is a useful measure as it is related to levels of intelligence and social economic status (Rindermann, 2008). Findings from these studies suggest that when OAs either have significantly fewer years in education or lower education levels than YAs, then OAs appear to have age-related recognition deficits for numerous emotion types (e.g., Mill, Allik, Realo, & Valk, 2009; Orgeta & Phillips, 2009). In contrast, when the number of years in education is either matched or higher in OAs, compared to YAs, then OAs often appear to be less able than YAs to recognise one emotion from facial expressions or have a general decline (Circelli, Clark, & Cronin-Golomb, 2013; Grainger et al., 2015b; Suzuki, Hoshino, Shigemasu, & Kawamura, 2007). However, in one study OAs were matched with YAs on years in education and the OAs had
recognition deficits for four out of the six emotions measured (Hunter, Phillips, & MacPherson, 2010a). Thus there are differences in the breadth of emotion recognition deficits between these studies and between-study variations in task demands may provide one possible explanation for these findings. For example, in the study
conducted by Hunter, Phillips, and MacPherson (2010a) the availability of the stimuli was time restricted (5000 ms), whereas the stimulus was available until a response was given in other studies. Thus the findings indicate that higher memory demands may have greater implications on OAs’ emotion recognition accuracy than years in education. Nevertheless, given the suggested pattern regarding education and age- related differences in emotion recognition ability between YAs and OAs, it is prudent for research in the field to match OAs and YAs on education level.
Whilst education maybe related to intelligence, intelligence also provides one measure of cognitive ability but the results in the field regarding the role of intelligence
are somewhat complex. Some evidence suggests that when OAs and YAs were
matched on both verbal and fluid intelligence then OAs were less able to recognise two out of six emotion types (Schaffer, Wisniewski, Dahdah, & Froming, 2009). However, similar age-related deficits are reported when OAs had lower levels of fluid intelligence coupled with comparable verbal intelligence to YAs (e.g., Phillips, MacLean, & Allen, 2002; Sullivan & Ruffman, 2004). Thus it appears that similar levels of intelligence in OAs and YAs may not protect against age-related emotion recognition deficits in OAs. It is possible, however, that lower levels of intelligence in OAs than YAs may
exacerbate any age-related differences. To reduce this possibility OAs should be at least matched with YAs on levels of intelligence.
Few studies in the field using facial expressions have measured sample characteristics of personality, affect, empathy, and alexithymia; thus, it is unclear whether these variables can explain age-related differences in emotion recognition ability in OAs. However, it does appear from one study that OAs still have some emotion recognition deficits compared to YAs when age groups are matched on the big five personality traits and alexithymia (Keightley, Winocur, Burianova,
Hongwanishkul, & Grady, 2006). Therefore, to understand whether any differences in emotion recognition ability between OAs and YAs are direct or indirect effects, the findings need to be discussed within the context of age group differences in
characteristics that may influence emotion recognition ability, such as alexithymia, anxiety, and empathy (Parker, Taylor, & Bagby, 1993; Surcinelli, Codispoti,
Montebarocci, Rossi, & Baldaro, 2006; Sucksmith, Allison, Baron-Cohen, Chakrabarti, & Hoekstra, 2013).
In addition to the differences between OAs and YAs, OA samples may vary between studies on several aspects, such as age. For example, due to natural aging older-older adults, typically over 70 years of age, are likely to have more advanced
cognitive and neurological changes than younger-older adults, between 60-70 years of age (Zec, Markwell, Burkett, & Larsen, 2005). These progressive changes may account for some of the between-study variations in the number of age-related emotion
recognition deficits. It may be expected then that older-older adults will have recognition deficits across more emotion types than younger-older adults.
In summary, the between-study methodological and sample differences may help to explain the variations of findings in the field. Importantly, methodological inconsistencies and a failure to account for possible confounding characteristics between the age groups makes it difficult to understand the extent of age-related differences in emotion recognition ability. The current experiment was designed to address these issues by using an emotion recognition task that is comparable to preceding research in that it uses static facial expressions and a forced-choice task. Essentially however, the experiment was carefully designed to tighten up some of the methodological issues observed in previous research including controlling for
unnecessary task demands. To reduce the possible impact of confounding sample differences the current study specifically recruited OAs who did not have lower levels of education, fluid and verbal intelligence, positive mood, and empathy, or higher levels of negative affect or alexithymia than YAs. In this manner it is unlikely that any age- related emotion recognition deficits in OAs result from these skills and traits; thus findings can be more confidently attributed to emotion recognition ability than some of the previous research in the area. Furthermore, to investigate the influence of advancing older age on task performance the current research not only compared accuracy for emotion recognition between YAs and OAs but comparisons were made between younger-older adults (59-69 years) and older-older adults (70 years and over). Finally, a non-emotion task was included in the experiment to determine whether any age-
related differences in emotion recognition ability could be explained by OAs’ and YAs’ ability to conduct the task. To achieve this the aims of the experiment were:
To use a task specifically designed to be similar to previous research using facial expressions in that the emotion recognition task was forced-choice and
measured the recognition of five basic emotions plus a neutral option. However, the stimuli remained on the screen until the time limit was reached and the photographs of facial expressions had limited contextual cues.
To measure many sample characteristics including education status, intellectual ability, empathy, alexithymia, personality traits, and current affect to help explain the findings regarding emotion recognition ability.
To determine whether differences in emotion recognition ability between YAs and OAs could be explained by age group differences in the ability to meet the task demands, a non-emotion task with similar task demands to the emotion task was included.
To understand whether emotion recognition ability deteriorates further with advancing older age findings were compared between younger-older and older- older adults.
Finally, the emotion task was also designed to be comparable to the emotion recognition tasks used in Experiments 2 and 3.