Theme 4 It is unclear to what extent older age performance on emotion recognition tasks is attributable to cognitive decline.

Given changes in some cognitive abilities in OAs (as outlined in section 1.2) it is important to consider whether studies in the field might be confounded by an age- related demise in some cognitive abilities (Hedden & Gabrieli, 2004; Salthouse, 2009). Specifically, emotion recognition often results from a complex integration of cognitive processes including detecting changing emotions expressions, processing of visual or auditory inputs, and discriminating between emotions (Phan et al. 2002; Sze, Goodkind, Gyurak, & Levenson, 2012). It is, therefore, conceivable that the age-related

impairments in emotion recognition in OAs are a consequence of changes in cognitive processing with natural aging rather than emotion processing (Somerville, Fani, & McClure-Tone, 2011). Thus, it is important to disentangle age-related decline in emotion recognition ability from age-related changes in cognitive processing.

Several studies in the area have explored the relationship between cognitive functioning and emotion recognition. Orgeta and Phillips (2008) report that disparities in recognition accuracy for facial expressions of fear, anger, and sadness between YAs and OAs are associated with slowing of processing speed with older age. Furthermore, accounting for processing speed reduced the size of the age-related differences in emotion recognition ability in OAs from dynamic faces (West et al., 2012). These

findings suggest that processing speed may be a specific predictor of age-related emotion recognition impairments in OAs.

Further evidence has found that fluid intelligence, memory (immediate and delayed recall), and processing speed partially accounted for the variance in recognition ability across emotions from faces (Horning, Cornwall, & Davis, 2012). Specifically fluid intelligence predicted emotion recognition accuracy for fear, anger, disgust, and surprise in OAs. Recognition of fear was also predicted by memory, whilst processing speed predicted recognition for sad and happy faces in OAs. However, once age was accounted for the predictors for fear were lost. The findings imply that cognitive functioning has differential effects on emotion recognition ability in OAs across emotion types and age may be a better predictor of fear recognition than cognitive functioning.

There is further evidence that cognitive decline, specifically executive

functioning, may have differential effects of emotion recognition across emotion types. Executive functioning includes higher order cognitive abilities including set shifting, inhibition, motivation, manipulating information, verbal fluency, and planning (Wechsler, 1981). Circelli, Clark, and Cronin-Golomb (2013) report that executive functioning is related to recognition of sad facial expressions in OAs. Also lower executive functioning (as measured by amalgamated scores of word fluency, arithmetic, card sorting, and memory) was related to lower anger recognition in OAs (Krendl & Ambady, 2010). However, whilst the tasks test cognitive abilities they do not capture all of the cognitive abilities encompassed within the concept of executive function, such as set shifting and inhibition, so may not measure full executive function. Nevertheless, taken together the presented evidence demonstrates variations in findings suggesting a complex relationship between specific cognitive abilities and recognition ability for specific emotion types.

In contrast, some evidence suggests that differences between OAs and YAs in emotion recognition ability are not associated with cognitive abilities. For example, non-verbal emotion recognition ability was not related to education, cognitive status, or memory (Lima, Alves, Scott, & Castro, 2014). However, the OA sample in Lima, Alves, Scott, and Castro (2014) comprised of younger-older adults (mean age of 61 years) as such these OAs may not have undergone cognitive demise to the same extent as older-older adult samples. Hence, in Lima et al. the age-related cognitive decline in OAs may not be sufficiently advanced to demonstrate an association with emotion recognition ability. However, evidence from several other studies corroborate with the findings reported by Lima and colleagues. For instance, age-related impairments in recognising affective prosody in OAs were not accounted for by cognitive ability such as list recognition, story recall, and Stroop tests (Orbelo, Grim, Talbott, & Ross, 2005). Furthermore, Keightley, Winocur, Burianova, Hongwanishkul, and Grady (2006) found that cognitive functioning, measured by several abilities including verbal knowledge and fluency; working memory; and inhibition, did not account for emotion recognition differences from static faces between YAs and OAs. More specifically, fluid

intelligence did not account for the variance in adults’ ability to match vocal expressions to either emotion word labels or facial expressions (Ryan, Murray, & Ruffman, 2010). Impairments for accurately recognising anger and sadness in OAs, compared to YAs, from morphed facial expressions were not attenuated when

processing speed and fluid intelligence was accounted for (Sullivan & Ruffman, 2004a). Nor were age-related differences in emotion recognition ability in OAs from facial expressions related to fluid and crystallised intelligence, or education (Phillips, Maclean, & Allen, 2002) or accounted for by executive function (West et al., 2012). Finally, Lambrecht, Kriefelts and Wildgruber (2012) stated that age-related declines in emotion recognition of facial expressions and prosodic vocalisations still existed once

cognitive ability (working memory and verbal intelligence) was accounted for. Taken together these findings suggest that cognitive demise in older age does not account for differences in emotion recognition ability between YAs and OAs.

Given the disparate findings regarding the association between cognitive performance and emotion recognition ability it remains unclear what impact, if any, cognitive decline with age has on emotion recognition performance in OAs. It is possible that OAs use compensatory mechanisms (such as previous experience) to overcome age-related cognitive change during emotion recognition (Isaacowitz & Stanley, 2011). These adaptive behaviours may serve to mask the effects of an age- related decline in cognitive functioning on emotion recognition ability in OAs.

Nevertheless, the majority of research suggests that cognitive performance is not related to emotion recognition ability (Ruffman, 2011).

Rather than accounting for specific cognitive abilities on emotion recognition accuracy it may be more promising to account for OAs’ ability to meet the processing demands of the task. In this manner, instead of isolating particular cognitive functions, information can be gleaned regarding OAs’ ability to complete the task. One way of measuring the ability to meet the demands of the task is to use non-emotion tasks. Incorporating non-emotion tasks alongside emotion tasks can help to tease apart emotion and cognitive performance. To achieve this a non-emotion task should be carefully designed such that it places the same cognitive demands on the participants as the emotion task. In theory, as long as the emotion and non-emotion tasks are closely matched for task demands, comparable accuracy between YAs and OAs on non- emotion tasks would suggest that OAs are as able to meet the demands of the task as YAs. Therefore, any age-related difficulties in emotion recognition ability in OAs are likely to result from age group differences in emotion processing rather than from more general cognitive abilities.

Sullivan and Ruffman (2004) included several non-emotion tasks (a morphed object, gender discrimination task, and beaker judgement task). Their findings suggest that age-related emotion recognition difficulties in OAs are not due to face processing abilities or task demands, as OAs performed well on the non-emotion and face tasks. However, the emotion and non-emotion tasks used in Sullivan and Ruffman (2004) had methodological variations. For example, the gender identification task only required a judgement from two options whilst the emotion recognition task required judgements to be made from several response options. Good performance on the non-emotion tasks compared to emotion tasks, therefore, might be a consequence of methodological inconsistencies between the tasks making the non-emotion task less demanding and easier than the emotion task. Arguably conclusions from non-emotion tasks are stronger when the emotion and non-emotion tasks are matched as a far as possible on task demands.

In an alternative study Grunwald et al. (1999) assessed emotion recognition ability in OAs and YAs using words and sentences. Participants selected which emotion they believed was represented by a single word (gloomy), a group of three words (putrid, slime, stench), or a sentence (“he felt the urge to hit someone”). Similar non-emotion tasks were also used for comparison. OAs were less accurate across all tasks, with the exception of the word identification tasks, suggesting that OAs found the tasks more difficult than YAs regardless of content. In this instance it is possible that a lesser ability to meet the demands of the tasks in OAs than YAs contributed to the age- related emotion recognition deficits in OAs. However, a lack of methodological detail regarding the non-emotion tasks makes it unclear whether the emotion and non-emotion tasks had similar task demands.

In conclusion, whilst there is a logical argument that emotion recognition impairment with age is related to age-related decline in cognitive ability it appears that

there is little evidence that the two concepts are related. A few studies have found that specific emotion recognition impairments in OAs, compared to YAs, are associated with certain cognitive functions, such as fluid intelligence and processing speed. Yet the majority of the research, using several measures of cognition, have found no or only partial correlations between emotion recognition and cognitive abilities including fluid and verbal intelligence and processing speed. In some cases age is a stronger predictor of emotion recognition than cognitive ability. However, it may be naïve to overlook the potential contribution of cognitive ability on task performance. It would be prudent for researchers to control for any potential age-related differences in cognitive abilities by matching YAs and OAs on several cognitive functions including verbal and fluid intelligence. Furthermore, experiments should include both non-emotion and emotion tasks with similar task demands. In this manner results from emotion recognition tasks can be interpreted more confidently as a consequence of either emotion or cognitive processing.

1.7.4.1 The current study. To reduce the possible confound of cognitive ability on emotion recognition ability for Phase 1 and Phase 2, OAs were at least matched with YAs on fluid and verbal intelligence. The OAs did have higher levels of verbal

intelligence than the YAs but the direction of this age group difference should not reduce OAs’ emotion recognition ability. In this manner any differences in emotion recognition ability between YAs and OAs are not likely to result from lower

intelligence in OAs than YAs. Furthermore, the inclusion in Phase 1 of non-emotion tasks that are as procedurally similar to the emotion task is novel. The similarity in procedure of the emotion and non-emotion tasks (i.e., same number of response items, same format of presentation, and same number of trials) means that the tasks are matched as far as possible for task demands. Whilst other studies have accounted for face processing and procedural ability (e.g., Sullivan & Ruffman, 2004) the current

researcher is not aware of previous research that has rigorously controlled for methodological differences between the emotion and non-emotion task. 1.8 Conclusion

To date a wealth of research in the field has made a fundamental contribution to our understanding of how emotion recognition ability changes into older adulthood. The accumulation of evidence indicates that in general OAs have some difficulties compared to YAs in their emotion recognition ability. Whilst this research is central to our understanding of emotion recognition development there are several factors that need further clarification. Particularly there is a need to provide clarity to the extent of age-related differences in OAs across emotion, presentation, and stimuli types. When trying to understand exactly which emotions are vulnerable to age-related decline a review of the literature reveals a series of inconsistent results that fail to provide a common pattern as to which specific emotions are affected. This lack of clarity might result from differences in methodology, stimuli types, and sample characteristics; hence, there is a need to control for some of these potential confounds. Furthermore, the field would benefit from findings beyond those from facial expressions, basic emotions, and stimuli types with low ecological validity. To clarify the interpretation of emotion recognition differences between YAs and OAs it would be useful to use non-emotion tasks to discern between general processing and emotion processing ability. Finally, research is required to determine whether emotion recognition ability in laboratory tasks is reflective of more global measures of social functioning.