Metacognitive awareness of the associative deficit for words and names

ABSTRACT

Older adults have considerable impairment in associative recognition despite minimal age differences in item recognition. The magnitude of this associative deficit varies by type of stimuli, strategy utilization, and other mediators and moderators (Old & Naveh-Benjamin, 2008, Psychology and Aging, 23, 104–118). Name pair stimuli have not been used to test the associative deficit hypothesis (ADH), although tests using name–face stimuli support the ADH. Additionally, metacognitive awareness of the ADH has not been investigated. We tested the ADH with word and name pair stimuli, and predicted that age-related associative deficits would be larger for words than names because names, unlike most common nouns, lack certain semantic properties that could be used to bind pairs of names together. Results supported the ADH for words but not names: Younger and older adults recognized equivalently fewer names on the associative test relative to the item test. As predicted, self-efficacy was higher for younger than older adults. Surprisingly, self-efficacy for the associative test was higher than for the item test but post-test estimates of performance success (postdictions) were higher for the item test than for the associative test, suggesting sensitivity by participants to different task demands in the item and associative tests following recognition attempts. Metacognitive accuracy was better for words than names and for the item test than associative test, and equivalent between age groups. Overall, participants overestimated their name recognition abilities. Our findings extend support for the ADH to a conceptually important and ecologically valid domain (names) and provide new data on metacognitive aspects of the ADH.

• Aging
• Associative deficit
• Metacognitive awareness
• Recognition memory
• Self-efficacy

Notes

2. ¹Eight configurations of stimuli were created to counterbalance the effects of test type and stimulus type, and to control for possible order effects related to the stimuli. Participants studied either name pairs followed by word pairs, or word pairs followed by name pairs over four study-test blocks. There were two lists of names and two lists of words used in order to establish generalizability of our results across lists. Analyses of configuration, included as a between subjects variable in the 2 (Age Group: Young/Old) × 2 (Test Type: Item/Pairs) × 2 (Stimulus Type: Words/Names) ANOVAs did not change the magnitude nor pattern of main effects or interaction effects involving age group for recognition, MSE, and postdiction accuracy scores, so it was dropped from further analyses. Analysis of block order, to examine whether differences between test types (item and associative) decreased across blocks, included as a within subjects variable in the 2 (Age Group: Young/Old) × 2 (Test Type: Item/Pairs) × 2 (Stimulus Type: Words/Names) ANOVAs yielded a nonsignificant four-way interaction, and thus block order was dropped from further analyses.

3. ²Participants reported whether they were currently experiencing any health problems and whether they were currently taking any medications. We coded these responses as 1 (“yes”) or 0 (“no”), and conducted two Pearson chi-square tests of association between age group and health problems, and age group and medication use. Both tests were significant, χ²_meduse(82) = 11.70, p = .001, and χ²_hlthprob(82) = 10.07, p = .002. When entered as covariates in the ADH ANOVA, the three-way interaction effect remained significant, changing from p = .009 (η_p² = .083) to p = .025 (η_p² = .063). We also covaried age differences in self-reported hearing from three-way ANOVA test of the ADH, with comparable results: p_change = .024, (η_p² = .063).

4. ³It has been argued that absolute accuracy avoids the problem of simple difference scores cancelling out meaningful differences, e.g., between age groups (see Hertzog et al., 2002). Therefore, we also analyzed absolute accuracy scores. None of the effects involving age were significant, therefore analyses of both relative and absolute accuracy scores support our conclusion that younger and older adults had equivalent levels of metacognitive accuracy.

5. ⁴At the suggestion of one of the reviewers, we examined correlations between the four metacognitive accuracy measures and several individual difference measures, including age, sex, education level, vocabulary, speed, and self-rated health, hearing, and vision, separately within each age group. None of the correlations were significant among younger adults, and only S-R hearing was significant among the older adults, r = .31, p = .047. When S-R hearing was added as a covariate in the three-way analysis of postdiction accuracy, the critical three-way interaction effect remained significant, p = .005, η_p² = .096. Thus, none of the potential individual difference variables changed the metacognitive accuracy effect.

6. ⁵Memory self-efficacy was also examined as a potential mediator of age-related differences on the word associative recognition test based on theory (Bandura, 1993; Berry, 1999), empirical work (Artistico, Cervone, & Pezzuti, 2003), and meta-analysis (Beaudoin & Desrichard, 2011). The total effect of age group on memory was significant (B = –0.024, SE = 0.06), t(80) = –3.83, p < .001, and remained significant when word associative MSE (as mediator) was added to the model (B = –0.21, SE = 0.07), t(79) = –3.22, p = .002. The indirect effect of age group → MSE → recognition memory was nonsignificant (B = –0.0232, Boot SE = 0.03), and its 95% CI (–0.0869, 0.0348) included zero. Total R² was 17%; age explained 16% of this variance (p = .001) and MSE explained 1% (p = .380.) MSE was not a significant mediator of age-related differences in word associative recognition, nor did it contribute unique variance.

7. ⁶A comparable analysis of word associative HR with word associative FAR as a covariate in the ANOVA yielded a similarly significant, but smaller indirect effect of strategy success (B = –0.042, Boot SE = 0.02, with 95% CI (–0.0980, –0.0097).