Words matter: age-related positivity in episodic memory for abstract but not concrete words

ABSTRACT

Research continues to assess potential boundary conditions for the age-related positivity effect in emotional information processing. Beyond the valence and arousal characteristics of a stimulus, other features may play a role in the manifestation of positivity effects. Differences between abstract and concrete words (i.e., the level of imageability, presence of affective content) may lead to differences in downstream processing outcomes. The present study examined whether additional features of word stimuli, beyond valence and arousal, could influence the emergence of age-related positivity in episodic memory. Fifty-two younger adults and 51 older adults completed a categorization task where they separated a series of positive, negative, and neutral words into “abstract” or “concrete” categories. A surprise recognition task followed after a short delay. Results revealed a three-way Age × Valence × Word Type interaction. No age differences in overall recognition was observed for concrete words, regardless of emotion; however, for abstract words, an Age × Valence interaction emerged whereby older adults recognized more positive than negative words, while valence differences were less pronounced among younger adults. Concrete words were remembered better than abstract words by both age groups, regardless of valence. Conversely, preferential processing appeared to occur for positive abstract words, especially for older adults. We contend that these results provide further evidence of the top-down and controlled nature of age-related positivity effects in episodic memory.

KEYWORDS:

• Aging
• positivity effect
• episodic memory
• emotion
• word type

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. We examined the distribution of nouns, adjectives, verbs, and multiple-meaning words across valence and word type factors during the initial word sort. Chi-square analyses revealed that there were more abstract adjectives (n = 42) than concrete adjectives (n = 7) and more concrete nouns (n = 69) than abstract nouns (n = 28), Χ² = 48.63, df = 3, p < .001. Additionally, there were slightly more negative adjectives (n = 33) than positive adjectives (n = 16), Χ² = 9.19, df = 3, p = .027. However, these valence differences did not appear when isolating words based on time of presentation and word type (i.e., shown during initial encoding or retrieval), ps > .07. Nouns, adjectives, verbs, and multiple-meaning words were distributed similarly across valence, notwithstanding the fact that more abstract words were adjectives, and more concrete words were nouns.

2. It was reasonable to suspect that overall differences in performance during the initial encoding task (i.e., how accurately abstract and concrete words were categorized as such) may have differed across word type and valence. We conducted a 2 (Word Type: Abstract, Concrete) × 2 (Valence: Positive, Negative) × 2 (Age: Younger, Older) mixed ANOVA on percentage scores as a measure of accuracy. All main effects were significant (all ps < .05), most notably the main effect of age, with younger adults outperforming older adults. However, these effects were qualified by a significant interaction between valence and word type, F(1,101) = 268.15, p < .001, η_p² = .73. It was clear that negative abstract words were driving the interaction, as they were categorized most accurately by younger (M = .842, SD = .26) and older adults (M = .937, SD = .09). Importantly, this pattern held across all word forms (i.e., nouns, adjectives, verbs). and was opposite the word recognition pattern during the retrieval task. Thus, it appears that corrected recognition scores were relatively unaffected by performance on the initial word sorting task.

3. This omnibus ANOVA was conducted with neutral words added as a third condition within the Valence factor. When included, all main effects and interactions remained significant. The pattern of the neutral words matches the concreteness effect described earlier; however, we do not discuss these results as our focus is on assessing positive versus negative preferences. Furthermore, we did not control how neutral words were selected by their respective concrete or abstract status, resulting in a large mismatch between the number of abstract (n = 20) and concrete (n = 80) neutral words. This was because neutral words were meant to act as a set of filler items, not as a substantive level of a key independent variable.