Similarity to the Self Affects Memory for Impressions of Others in Younger and Older Adults

Abstract

Objectives.

Similarity to the self has been shown to affect memory for impressions in younger adults, suggesting a self-reference effect in person memory. Because older adults show comparable self-reference effects, but prioritize memory for positive over negative information relative to young adults, we examined age differences in self-similarity effects on memory for positive and negative impressions.

Method.

Younger and older adults formed positive and negative impressions of others differing in the degree of similarity to the self (high, medium, low).

Results.

For positive impressions, both groups showed enhanced memory for self-similar others relative to dissimilar others, whereas for negative impressions, memory was poorer for those similar to the self. When collapsed across similarity to the self, younger adults remembered negative impressions better than older adults, but interestingly, older adults exhibited a trend for better memory for the positive impressions.

Discussion.

Results suggest that self-reference effects in impression memory are preserved with age and that older adults exhibit positivity effects in person memory consistent with previous findings.

Processing information relative to the self improves memory in older and younger adults (Glisky & Marquine, 2009; Gutchess, Kensinger, Yoon, & Schacter, 2007; Leshikar & Duarte, 2012, 2013; Mueller, Wonderlich, & Dugan, 1986; Rogers, Kuiper, & Kirker, 1977) and sometimes leads to age-equivalent memory performance (Experiment 1, Cassidy & Gutchess, 2012; Rosa & Gutchess, 2011). We recently demonstrated that self-reference effects extend to memories about other individuals in younger adults (Leshikar & Gutchess, 2013). Specifically, participants in that study formed either positive or negative impressions of others exhibiting characteristics (e.g., kindness) that varied in self-descriptiveness. For positive impressions, memory was better for those showing characteristics similar to oneself. In contrast, memory for negative impressions was poorest for those showing characteristics similar to the self.

Previous work has shown age differences in memory for specific behaviors performed by other individuals (Hess & Tate, 1991). No investigation, however, has examined age differences in memory for positive or negative impressions of others, apart from memory for the specific behaviors. Testing memory for impressions separately from memory for specific behaviors is important considering that people can remember their impressions even when they cannot remember the behaviors upon which their impressions were based (Todorov & Uleman, 2003). Because memory for specific details is impaired with age (e.g., Spencer & Raz, 1995; Schacter, Kaszniak, Kihlstrom, & Valdiserri, 1991), we examined whether impression memory—remembering that this impression was associated with that person—shows age-related decline, or whether impression memory might be spared, consistent with work showing preserved memory for specific details concerning materials rich in social detail in older adults (Rahhal, May, & Hasher, 2002).

Valence is inherently important in impressions, which typically reflect a general positive or negative assessment of someone (Srull & Wyer, 1989). Valenced information is remembered differently with age, with younger adults remembering proportionally more negative than positive information whereas older adults tend to remember proportionally more positive than negative information (Charles, Mather, & Carstensen, 2003, 2005). One investigation showed that older adults had significantly better memory for happy faces than angry or neutral faces, whereas younger adults’ memory did not differ across expressions (Ebner & Johnson, 2009). Thus, older adults may show relatively preserved memory for positive impressions compared with younger adults, while exhibiting poorer memory for negative impressions.

We examine memory for positive and negative impressions of others in younger and older adults as a function of self-similarity (high, medium, low). We predict that similarity to the self would affect impression memory, with both groups showing better memory for others associated with positive impressions who are similar to the self, whereas self-similarity should lead to poor memory for others associated with negative impressions, consistent with our prior work in younger adults (Leshikar & Gutchess, 2013), and literature showing similar self-reference effects on memory across age (Dulas, Newsome, & Duarte, 2011; Glisky & Marquine, 2009; Hamami, Serbun, & Gutchess, 2011; Leshikar & Duarte, 2013; Mueller et al., 1986; Rosa & Gutchess, 2011). We further predict that valence, collapsed across similarity to the self, will differentially affect impression memory across age, with older adults showing preserved memory for the positive impressions and poorer memory for negative impressions relative to younger adults, consistent with previously reported positivity effects (Charles et al., 2003; Mather & Carstensen, 2005; Scheibe & Carstensen, 2010).

Methods

Twenty-nine younger (age: 19.9, SD: 2.1, range 18–26, 19 females) and 29 older adults (age: 74.6, SD: 5.2, range 66–87, 18 females) participated in this experiment. An additional three younger and four older adults were excluded for failing to use each of the levels of self-similarity ratings (high, medium, low). Participants gave written informed consent in compliance with the Brandeis Institutional Review Board.

There were 216 face-sentence pairs, with faces drawn from a face database (Minear & Park, 2004). Normed sentences (Uleman, 1988) described a behavior (e.g., “This person returned the wallet with all the money in it.”) that implied a trait (e.g., honesty). Sentences were selected so that half would elicit positive and half negative impressions.

There were three study-test blocks. At study, participants formed impressions for 144 trials (see Figure 1). For each trial, participants were instructed to form a positive or a negative impression of that person based on the sentence and report that impression via a button press (1 = positive; 2 = negative).

Figure 1.

Trial schematic for the study (impression formation) and test phases of the experiment.

The recognition memory test consisted of 216 trials (144 studied and 72 unstudied items). For each trial, participants made two judgments: First, participants were shown a face and decided whether they had generated a positive or negative impression for that face at study, or decided whether the face was unstudied. Correct impression trials were those where participants accurately remembered their impressions generated at study. Second, participants decided which sentence had been paired with the face during study. Participants selected from the correct sentence, a sentence of the same valence (e.g., positive, negative) paired with a different face at study (e.g., re-paired lure), and the option to call the face new.

Immediately following the last study-test block, participants completed a self-paced, post-test rating for 144 trait words. Each word was a trait implied by a behavioral sentence seen at study. Participants rated the self-descriptiveness of the trait on a three-point scale (1 = describes me a lot, 2 = describes me a little, 3 = does not describe me). This rating allowed us to back-sort all trials into high-, medium-, and low-self-similarity.

Results

No group differences emerged in the generation of positive (younger, 56%, SD: 8%; older, 58% SD: 7%) or negative impressions (younger, 44%; older, 42%), ts = 0.68, ps = .50. Proportions of correct trials (e.g., trials given a “positive” impression by the participant at study subsequently endorsed as “positive” at test) and incorrect impression memory trials (e.g., trials given a “positive” impression at study endorsed as “negative” at test) as well as the proportion of studied items endorsed as new (misses) are shown in Table 1, along with responses to the novel (unstudied) trials. Proportions of trials binned into high-, medium-, and low-self-similarity are shown in Table 1. A chi-square analysis showed that the distribution of bin sizes differed across categories, χ²(5) = 17.7, p < .01, which was driven by fewer low-self positive and high-self negative trial types in both age groups. Importantly, however, bin sizes did not differ across age for any bin category, ts < 1.4, ps > .16.

Table 1.

Proportion of Studied Items Associated With Correct Impression Responses, Incorrect Impression Responses, and Item Misses, as Well as the Proportions of New Items Endorsed as Positive (FA-Positive), Negative (FA-Negative), or New (CR), Is Shown for Younger and Older Adults

		Impression memory						Trial proportions
		Correct impression		Incorrect impression		New (miss)
		Younger	Older	Younger	Older	Younger	Older	Younger	Older
Studied items as a function of:
Self-relevance	Initial impression
High	Positive	.69 (.13)	.79 (.12)	.11 (.09)	.08 (.08)	.20 (.10)	.13 (.08)	.46 (.15)	.48 (.16)
	Negative	.42 (.22)	.35 (.28)	.36 (.21)	.43 (.28)	.22 (.22)	.22 (.26)	.14 (.06)	.13 (.11)
Medium	Positive	.66 (.15)	.72 (.12)	.14 (.09)	.11 (.08)	.20 (.13)	.17 (.11)	.40 (.14)	.36 (.17)
	Negative	.60 (.17)	.41 (.25)	.23 (.15)	.39 (.23)	.17 (.10)	.21 (.15)	.42 (.13)	.35 (.18)
Low	Positive	.52 (.20)	.60 (.25)	.27 (.18)	.21 (.23)	.21 (.17)	.18 (.16)	.14 (.06)	.16 (.08)
	Negative	.66 (.16)	.44 (.21)	.17 (.10)	.35 (.22)	.18 (.11)	.21 (.12)	.44 (.13)	.52 (.20)
Unstudied items		FA-positive		FA-negative		New (CR)
		Younger	Older	Younger	Older	Younger	Older
		.19 (.11)	.31 (.03)	.11 (.08)	.12 (.02)	.70 (.14)	.57 (.04)

Note. Impression memory was calculated as the proportion of correct impression out of the proportion of correctly recognized old items (correction plus incorrect impression judgments). Proportions of trials binned into high, medium, and low-self-similarity based on participants subjective post-test rating is also shown as a function of valence for both the younger and older adults. CR = Correct rejection; FA = False alarm.

Impression memory was calculated by taking the proportion of correct impressions out of the correctly recognized old items (See Figure 2; formula: correct impressions/[correct impressions + incorrect impressions]). An Age (younger, older) by Valence (positive, negative) by Self (high, medium, low) repeated measures analysis of variance on impression memory indicated an Age effect, F(1, 56) = 6.35, p = .02, η_p ² = .10, driven by better memory for the younger (M = .73) than the older adults (M = .67), a Valence effect, F(1, 56) = 45.08, p < .001, η_p ² = .44, with better memory for the positive (M = .81) than negative items (M = .59), and a Self effect, F(2, 112) = 3.80, p = .03, η_p ² = .06, driven by better memory for the medium- (M = .73), than high- (M = .68) or low-self trials (M = .69), ts > 2.78, p < .007. (Note that the better memory for the medium- than high- or low-self-similarity trials occurred when collapsing across valence. To calculate high-self memory performance, memory for the positive-high-self trials [which was high] was combined with the performance for the negative-high-self trials [which was low], whereas the contribution of valence was less extreme for the medium-self trials.) There was also a significant Valence by Self interaction, F(2, 112) = 35.54, p < .001, η_p ² = .39. Simple effects analysis showed a significant effect of Self for positive impressions, F(2, 114) = 26.04, p < .001, η_p ² = .31, which resulted from better memory for the high- (M = .88), than medium- (M = .84) or low-self-similarity trials (M = .70), ts > 2.79, p < .007. Simple effects for the negative impression trials also resulted in a Self effect, F(2, 114) = 19.82, p < .001, η_p ² = .26, which was driven by better memory for the low- (M = .68), than either medium- (M = .61), or high-self-similarity trials (M = .48), ts > 2.70, p < .009. There was also an Age by Valence interaction, F(1, 56) = 12.79, p = .001, η_p ² = .19. Simple effects showed a trend for better memory in the older adults (M = .84) for positive impressions than the younger adults (M = .78), F (1, 56) = 3.32, p = .07, η_p ² = .06. Younger adults, however, correctly remembered more negative impressions (M = .68) than older adults (M = .50), F(1, 56) = 12.95, p = .001, η_p ² = .19. We further followed up the age by valence interaction by looking at effects of valence within each age group. Both groups showed a valence effect with positive better remembered than negative trials (younger adults, F(1, 28) = 5.97, p = .02, η_p ² = .18; older adults, F(1, 28) = 45.17, p < .001, η_p ² = .62), although the presence of the interaction and inspection of the patterns indicate that the valence effect was stronger for older than younger adults. Given our primary interest in impression memory, results from the sentence memory task are reported in Supplementary Data, but are not further discussed.

Figure 2.

Impression memory for others as a function of self-similarity (high, medium, low; as displayed left to right) and valence of the initial impressions (positive—solid bars, negative—open bars) are depicted for younger and older adults.

Discussion

We investigated the influence of the self on impression memory in both younger and older adults. We report two findings: First, both age groups showed better memory for self-similar others when the initial impression was positive and worse memory when the initial impression was negative. Second, valence, regardless of similarity to the self, affected memory with older adults showing poorer memory for negative impressions but a trend for enhanced memory for positive impressions compared with the younger adults.

For the self-similarity effects, both age groups showed improved memory for positive impressions as self-similarity increased, with the best memory for high-self-similar others. This finding is consistent with our previous work in younger adults (Leshikar & Gutchess, 2013), and suggests that self-reference effects in impression memory are evident in older adults, contributing to a growing literature showing stable benefits of self-referencing in older adults as in the young (Dulas et al., 2011; Glisky & Marquine, 2009; Gutchess et al., 2007; Hamami et al., 2011; Mueller et al., 1986; Rosa & Gutchess, 2011). Specifically, recent work has shown that relevance to the self facilitates memory for contextual details in younger (Dulas et al., 2011; Leshikar & Duarte, 2012; Serbun, Shih, & Gutchess, 2011) and older adults (Hamami et al., 2011; Leshikar & Duarte, 2013; Rosa & Gutchess, 2011). Here we extend this prior work by showing facilitated memory for specific details (impressions) in a social context. Interestingly, these effects occurred in a study task that was not explicitly self-referential, consistent with prior work showing that self-reference effects occur spontaneously and effortlessly (Kesebir & Oishi, 2010; Yang, Truong, Fuss, & Bislimovic, 2012). For negative impressions, both age groups showed decreased memory with higher degrees of self-similarity. It may be that when faced with information that is both negative and self-relevant, young and older adults engage in more shallow encoding as a means to protect the self-schema, consistent with self-enhancement effects (D’Argembeau & Van der Linden, 2008; Leary, 2007), which may have led to the poor memory for the high-self-similar negative impression trials. Further work will be necessary to understand this inverse self-reference effect for the negative impressions.

Although these data suggest that impression memory is affected by similarity to the self, there is also an alternative possibility. It is possible that the strength of the impression could have been less strong for the low-similarity positive and high-similarity negative trials. Thus, when seeing a face, participants might have recalled the sentence but have a harder time remembering their original impression because they generated an ambivalent impression. Further work will be necessary to assess how the strength of the impression influences memory.

When collapsed across similarity to the self, valence led to memory differences across age. For positive impressions, older adults, relative to the young, showed a trend for enhanced memory, consistent with previous reports of positivity effects (Mather & Carstensen, 2005). Socioemotional selectivity theory suggests that older adults place increasing importance on positive experiences, such as maintaining positive affect and personally meaningful social relationships (Carstensen, 1992), with some work even showing that older adults perceive stimuli as more positive under some circumstances (Kellough & Knight, 2012). Thus, older adults may be more motivated than the young to remember positive information about others. Indeed, older adults may be more motivated to remember positive impressions because it is only with such individuals that a rich social relationship might develop, reflecting the importance that personal meaningfulness plays in memory in older adults. A recent memory investigation showed disproportionally enhanced memory for older adults when they processed social materials for personal meaningfulness (e.g., “would I want this person to play a role in my life?”) than in a less meaningful context (Cassidy & Gutchess, 2012). For the negative impressions, however, older adults showed poorer memory than the young. Interestingly, older adults showed a stronger memory advantage for positive over negative impressions, which might suggest a negativity-avoidance pattern in the older adults, consistent with prior work (Xing & Isaacowitz, 2006). Because work has shown that older adults are less sensitive to negative cues, such as whether someone is untrustworthy (Castle et al., 2012) it may be that when encountering someone they do not like, older adults may process attributes about them shallowly, which, in a broader context, could potentially contribute to older adults’ greater susceptibility to fraud.

One limitation to the design is that the uneven numbers of high-, medium-, and low-self-similarity trials, an unavoidable consequence of subjective ratings, could have influenced memory performance. For instance, small bin sizes may have been expected to make some trial types more distinctive and thus more memorable, although overall that does not appear to be the case. Importantly, because bin sizes did not differ across age, we believe this possibility did not affect our interpretation of memory effects across groups.

Overall, these data suggest that similarity to the self affects memory for impressions of others similarly in younger and older adults, consistent with reports of stable self-reference effects over the life span. Although older adults showed a trend for better memory for the positive impressions relative to the young, they further showed significantly poorer memory for negative impression trials, which we argue reflects changing information processing goals in older adults who tend to attend more to personally meaningful and positive materials.

Supplementary Material

Supplementary material can be found at: http://psychsocgerontology.oxfordjournals.org/

Funding

This work was supported by National Institute on Aging (R21 AG032382 to A. H. Gutchess) and National Institutes of Health (T32AG000204-21 to E. D. Leshikar).