Online sample issue GeroPsych 1/2020 by Hogrefe

Volume 33 / Number 1 / 2020

Editors-in-Chief Frieder R. Lang Johannes Pantel Associate Editors Isabelle Albert Julia Haberstroh Eva-Marie Kessler Mike Martin Michael Rapp Peter Schoenknecht

GeroPsych The Journal of Gerontopsychology and Geriatric Psychiatry Special Section Autobiographical Memory and Health

Set of cards covering VIA’s 24 character strengths, 6 virtues – and practical interventions to help your clients! “The cards are insightful, user-friendly, and ready-for-impact – a tangible resource to help clients build character strengths fluency, catalyze exploration, and set clients on a trajectory of strengths application for building well-being, enhancing relationships, and managing stress.” Ryan M. Niemiec, PsyD, Education Director, VIA Institute on Character, Cincinnati, OH

Matthijs Steeneveld / Anouk van den Berg

Character Strengths Intervention Cards 50 Cards With Instruction Booklet 2020, 50 cards + 16-page booklet US $34.80 / € 27.95 ISBN 978-0-88937-566-6 Also available as eBook The VIA character strengths look at what positive character traits help us lead fulfilling and happy lives, rather than looking at what is wrong with us. Research has shown that knowing your strengths and using them more often leads to greater well-being, better performance, and more resilience. With these cards, you can help clients learn about their character strengths. This full-color 50-card set provides cards for each of the 24 VIA character strengths and 6 virtues as well

www.hogrefe.com

as information cards to hand out in groups and individual sessions. On top of that, 16 ready-to-use, evidence-based intervention cards help clients discover and explore their strengths and practice applying them more often. The cards are a valuable addition to the toolboxes of coaches, trainers, and therapists from any background. The card set is an excellent resource to use with the book Character Strengths Interventions: A Field Guide for Practitioners by Ryan M. Niemiec (ISBN 978-0-88937-492-8).

GeroPsych The Journal of Gerontopsychology and Geriatric Psychiatry

Volume 33 / Issue 1 / 2020

Editors-in-Chief Frieder R. Lang Johannes Pantel Associate Editors Isabelle Albert Julia Haberstroh Eva-Marie Kessler Mike Martin Michael Rapp Peter Schoenknecht

Editors-in-Chief

Prof. Dr. Frieder R. Lang Institute of Psychogerontology University of Erlangen-Nuremberg Kobergerstr. 62 90408 Nuremberg Germany Tel. +49 911 5302-96100 geropsych@fau.de

Associate Editors

Isabelle Albert, Esch-sur-Alzette Julia Haberstroh, Siegen Eva-Marie Kessler, Berlin Mike Martin, Zurich Michael Rapp, Potsdam Peter Schoenknecht, Leipzig

Editorial Board

Noel Ballentine, Pennsylvania Sube Banerjee, Brighton Thomas Boll, Esch-sur-Alzette Helen F. K. Chiu, Hong Kong Burcu Demiray, Zurich Barry A. Edelstein, Morgantown Ulrike Ehlert, Zurich Simon Forstmeier, Siegen Helene H. Fung, Hong Kong Denis Gerstorf, Berlin Vjera Holthoff-Detto, Berlin Andrea Horn, Zurich Michael Hüll, Freiburg Gizem Hülür, Zurich Frank Jessen, Cologne Boo Johansson, Gothenburg Stefan Kamin, Nuremberg Anna Kornadt, Esch-sur-Alzette Reto W. Kressig, Basel Ken Laidlaw, Norwich

Publisher

Hogrefe AG, Länggass-Strasse 76, 3012 Bern, Switzerland Tel. +41 (0)31 300 45 00, Fax +41 (0)31 300 45 91 zeitschriften@hogrefe.ch, www.hogrefe.com/j/gro

Media Manager

Josef Nietlispach, Tel. +41 (0)31 300 45 00, inserate@hogrefe.ch

Production Editor

Karoline Grauer, Tel. +41 (0)31 300 45 56, karoline.grauer@hogrefe.ch

Printing Office

Jetoprint GmbH, VS-Villingen, Germany

Cover Artwork

Ó 2018 Arthur Schall for ARTEMIS project (ART Encounters: Museum Intervention Study), Institute of General Practice, Johann Wolfgang Goethe University, Frankfurt am Main, Germany

ISSN

1662-9647 (ISSN-L) 1662-9647 (Print) 1662-971X (Online)

Frequency

4 issues per year

Subscriptions

Annual subscription rates: Libraries / Institutions € 249.00 / US$ 356.00 / CHF 327.00 plus postage and handling (print only; for information online subscriptions, please consult the journals catalogue on www.hgf.io/zftkatalog) Individuals € 106.00 / US$ 148.00 / CHF 143.00 plus postage and handling Individuals e-only € 95.00 / CHF 122.00 Single issue € 56.00 / US$ 74.00 / CHF 71.00 plus postage and handling

Indexing

This journal is regularly abstracted/indexed in EMCARE, ESCI, PSYNDEX, PsycInfo, PsycARTICLES and Scopus.

Electronic Full Data

www.econtent.hogrefe.com/toc/gro/current

GeroPsych (2020), 33(1)

Prof. Dr. Johannes Pantel Institute of General Practice Johann Wolfgang Goethe University Theodor-Stern-Kai 7 60590 Frankfurt am Main Germany Tel. +49 69 6301-6134 pantel@allgemeinmedizin.uni-frankfurt.de

Anja Leist, Esch-sur-Alzette Becca Levy, New Haven Ulman Lindenberger, Berlin anin, Zagreb Jasminka Luc Frank Oswald, Frankfurt am Main Yuval Palgi, Haifa Christina Röcke, Zurich Margund Rohr, Dresden Amit Shrira, Ramat Gan Cornel C. Sieber, Erlangen-Nuremberg Sjacko Sobczak, Heerlen Valentina Tesky, Frankfurt am Main Clemens Tesch-Römer, Berlin Alan Thomas, Newcastle Manuel Trachsel, Zurich Pieter J. Visser, Maastricht Jenny Wagner, Hamburg Susanne Wurm, Greifswald Dannii Yeung, Hong Kong Susanne Zank, Cologne

Ó 2020 Hogrefe

Contents Editorial

The Interpersonal Focus and Specificity of Autobiographical Memories: Theoretical Relevance and Methodological Considerations Burcu Demiray

Full-Length Research Reports

Interpersonal Focus in the Emotional Autobiographical Memories of Older and Younger Adults

Angelina J. Polsinelli, Kelly E. Rentscher, Elizabeth L. Glisky, Suzanne A. Moseley, and Matthias R. Mehl Interindividual Differences in Cognitive Functioning Are Associated with Autobiographical Memory Retrieval Specificity in Older Adults

Sarah Peters and Signy Sheldon Do South African Xhosa-Speaking People with Schizophrenia Really Fare Better? A Longitudinal Mortality Study in Older Patients with Schizophrenia

Dana Niehaus, Esme Jordaan, Riana Laubscher, Taryn Sutherland, Liezl Koen, and Felix Potocnik Perceived Benefits and Costs Contribute to Young and Older Adults’ Selectivity in Social Relationships

Erica L. O’Brien and Thomas M. Hess

Ó 2020 Hogrefe

GeroPsych (2020), 33(1)

Editorial

The Interpersonal Focus and Specificity of Autobiographical Memories Theoretical Relevance and Methodological Considerations Burcu Demiray1,2 1

Department of Psychology, University of Zurich, Switzerland University Research Priority Program “Dynamics of Healthy Aging”, University of Zurich, Switzerland

Cognitive aging is today perhaps one of the most important and popular domains of aging research, and autobiographical memory is a fruitful and relevant area of study: Older adults have a long personal past that is full of meaningful autobiographical memories embedded in the selfmemory system. According to the self-memory system model of autobiographical memory, memories are reconstructed during retrieval in line with the current goals and characteristics of the rememberer (e.g., Conway, Singer & Tagini, 2004). Thus, the socioemotional and cognitive qualities of the individual, which tend to change with age, have an impact on how the individual recalls their memories. This “how” could be studied by examining a diverse range of memory qualities. However, this special section focuses on two qualities that are highly relevant for aging adults: interpersonal focus and specificity. Interpersonal focus is relevant, because individuals emphasize interpersonal goals and positive relationships more with age (e.g., socioemotional selectivity theory; Carstensen, 1992). Specificity is another key memory quality that declines with aging: Cognitive aging influences the episodic component of autobiographical memory, leaving the semantic component relatively intact (e.g., Devitt, Addis, & Schacter, 2017). Polsinelli, Rentscher, Glisky, Moseley, and Mehl (2020) examined the interpersonal focus of negative and positive memories of older adults from the perspective of socioemotional selectivity theory. Consistent with that theory, they found that the positive memories of older adults were more interpersonally focused (and less selffocused) compared to their negative memories and to the positive memories of younger adults. From a cognitive

Ó 2020 Hogrefe

perspective, Peters and Sheldon (2020) showed that the cognitive functioning level of older adults was positively associated with the proportion of specific memories retrieved and the amount of specific details within a single memory. This work focused on healthy older adults along a spectrum of cognitive functioning rather than comparing healthy versus unhealthy groups (e.g., dementia). In doing so, it showed that patterns that exist at extreme ends of cognitive functioning for autobiographical memory specificity can also be observed in healthy older adults with more subtle cognitive functioning differences. In terms of methodology, two issues need to be emphasized for future research: multiple measurement units per participant and the coding of these units. First, both studies included the collection of multiple autobiographical memories per participant. This is common practice in autobiographical memory research that is accompanied by between-subject analyses, in which all memory scores of a participant are summed and averaged to create a personlevel mean. Although this analytical approach is widely used, it has two limitations: First, memories are nested within each participant, so the data have a hierarchic structure with two levels (i.e., person level, memory level). There is more similarity of memories within one person compared with memories of other people (e.g., Wright, 1998). Second, this approach emphasizes interindividual differences and neglects intraindividual variability. In fact, behaviors and psychological processes have been shown to vary within persons across even seconds and minutes (Houben, Van Den Noortgate, & Kuppens, 2015; Sherman, Rauthmann, Brown, Serfass, & Jones, 2015). Thus, the variability of memory scores within a person

GeroPsych (2020), 33(1), 1–2 https://doi.org/10.1024/1662-9647/a000222

contains information that should be taken into account. These two limitations should encourage autobiographical memory researchers to use multilevel modeling, which considers the hierarchical nature of the data and withinperson variations. Peters and Sheldon (2020) present a valuable example of the necessity and advantage of multilevel modeling in handling multiple measurement units per person. Second, both studies relied on the manual coding of memory narratives. Manual coding is a commonly used method in autobiographical memory research, although it is a laborious and demanding task. Polsinelli et al. (2020) used a computerized text analysis tool, the Linguistic Inquiry and Word Count (LIWC), in addition to manual coding in order to bolster their methodology. Social scientists have started investing in the automatization of coding procedures by collaborating with computer scientists to use machine learning to automatically code text (Demiray, Luo, Tejeda-Padron, & Mehl, in press). For example, Yordanova, Demiray, Mehl, and Martin (2019) worked with transcribed real-life conversations and trained a classifier to code for (1) whether or not the participant is talking about an autobiographical memory and (2) what the content and function of the memory is. The classifier was able to code transcripts with 74–99% accuracy. These high accuracy rates are promising for autobiographical memory researchers who work with large transcribed datasets, which will save substantial amounts of time and energy spent on manual coding. In future research, important memory qualities, such as interpersonal focus and specificity could be coded automatically. In sum, these two contributions demonstrate the current status of autobiographical memory research in the context of healthy aging. Both the socioemotional and cognitive perspectives are crucial in understanding and promoting healthy aging (e.g., Demiray et al., in press). Future autobiographical memory research could benefit from methodological and analytical advancements, such as innovative computerized text analyses and more sophisticated statistical analyses.

GeroPsych (2020), 33(1), 1–2

Editorial

References Carstensen, L. L. (1992). Social and emotional patterns in adulthood: Support for socioemotional selectivity theory. Psychology and Aging, 7, 331–338. Conway, M. A., Singer, J. A., & Tagini, A. (2004). The self and autobiographical memory: coherence and correspondence. Social Cognition, 22, 491–529. Demiray, B., Luo, M., Tejeda-Padron, A., & Mehl, M. R. (in press). Sounds of healthy aging: Assessing everyday social and cognitive activity from ecologically sampled ambient audio data. In P. Hill & M. Allemand (Eds.), Personality and healthy aging in adulthood: New directions and techniques. New York: Springer. Devitt, A. L., Addis, D. R., & Schacter, D. L. (2017). Episodic and semantic content of memory and imagination: A multilevel analysis. Memory & Cognition, 45, 1078–1094. Houben, M., Van Den Noortgate, W., & Kuppens, P. (2015). The relation between short-term emotion dynamics and psychological well-being: A meta-analysis. Psychological Bulletin, 141, 901–930. Peters, S., & Sheldon, S. (2020). Interindividual differences in cognitive functioning are associated with autobiographical memory retrieval specificity in older adults. GeroPsych, 33, 15–29. Polsinelli, A. J., Rentscher, K. E., Glisky, E. L., Moseley, S. A., & Mehl, M. R. (2020). Interpersonal focus in the emotional autobiographical memories of older and younger adults. GeroPsych, 33, 3–14. Sherman, R. A., Rauthmann, J. F., Brown, N. A., Serfass, D. G., & Jones, A. B. (2015). The independent effects of personality and situations on real-time expressions of behavior and emotion. Journal of Personality and Social Psychology, 109, 872–888. Wright, D. B. (1998). Modelling clustered data in autobiographical memory research: The multilevel approach. Applied Cognitive Psychology, 12, 339–357. Yordanova, K. Y., Demiray, B., Mehl, M. R., & Martin, M. (2019). Automatic detection of everyday social behaviours and environments from verbatim transcripts of daily conversations. In 2019 IEEE International Conference on Pervasive Computing and Communications (PerCom) (pp. 1–10). IEEE.

Burcu Demiray Department of Psychology University of Zurich Binzmühlestrasse 14 PO Box 24 8050 Zurich Switzerland b.demiray@psychologie.uzh.ch

Ó 2020 Hogrefe

Full-Length Research Report

Interpersonal Focus in the Emotional Autobiographical Memories of Older and Younger Adults Angelina J. Polsinelli1, Kelly E. Rentscher2, Elizabeth L. Glisky3, Suzanne A. Moseley4, and Matthias R. Mehl3 1

Department of Psychiatry & Psychology, Mayo Clinic, Rochester, MN, USA

Cousins Center for Psychoneuroimmunology, Semel Institute for Neuroscience and Human Behavior, University of California, Los Angeles, Los Angeles, CA, USA

Psychology Department, University of Arizona, Tucson, AZ, USA

MN Epilepsy Group, Minneapolis, MN, USA

Abstract: The present study examined the interpersonal focus within autobiographical memories (AMs) of older and younger adults from the perspective of socioemotional selectivity theory (SST). Specifically, we measured interpersonal focus directly through rater codings (relational vs. individual focus) and social word use, and indirectly through personal pronoun use. Forty-five older (Mage = 76.76) and 25 younger (Mage = 18.64) adults recalled positive and negative AMs, which were then coded and processed through computerized text analysis software to obtain word-use counts. Consistent with SST, the positive AMs of older adults were more interpersonally focused compared to negative AMs and younger adults. The results suggest that the positive life experiences of older adults tend to be associated with a high degree of social importance and focus on others. Keywords: interpersonal, LIWC, autobiographical memory, socioemotional selectivity theory

“What matters in life is not what happens to you but what you remember and how you remember it.” Gabriel García Márquez

Introduction Socioemotional selectivity theory (SST) postulates that perception of time changes over the lifespan, resulting in a shift in goal focus (Carstensen, Isaacowitz, & Charles, 1999). Specifically, while younger adults are typically more concerned with knowledge acquisition, older adults tend to focus more on generating and maintaining positive emotional experiences, including greater investment in meaningful and high-quality relationships that contribute to positive affect (Carstensen, Fung, & Charles, 2003). Less understood, however, is how the importance of relationships motivates the way older adults reflect on their personal, emotional life experiences or autobiographical Ó 2020 Hogrefe

memories (AMs). Directly comparing positive and negative AMs facilitates our understanding of processes linking relationship (i.e., interpersonal) focus with emotional experiences and affect in later life. The present study examines how older adults choose to narrate their emotional AMs. Specifically, we examine the social content and pronoun use of the positive and negative AMs of older and younger adults to address the link between interpersonal focus and positive life experiences in later life that would be predicted by SST.

Socioemotional Selectivity Theory SST posits that the relative importance of an individual’s set of goals (e.g., knowledge-acquisition, emotional) changes as a function of their perception of time until the end of life (Carstensen et al., 1999, 2003). When future time is perceived as expansive and open-ended, as in early adulthood, goals tend to be future-oriented. Consequently, SST GeroPsych (2020), 33(1), 3–14 https://doi.org/10.1024/1662-9647/a000220

predicts that younger adults prioritize knowledge and resource acquisition to prepare for the future. But when future time is perceived as constrained, as in later life, focus shifts to more present-day-oriented goals, which tend to be more emotional. Accordingly, SST predicts that older adults prioritize emotional gratification and regulation (Carstensen et al., 2003; Penningroth & Scott, 2012). Lang and Carstensen (2002) demonstrated a robust negative association between age and future time perspective (FTP) and found that replacing FTP with chronological age in models predicting goal focus produced very similar patterns. Specifically, emotion-regulatory and generativity goals were prioritized by those whose FTP was time limited (older adults), and autonomy and social acceptance were prioritized by those whose FTP was open-ended (younger adults). Carstensen and colleagues also acknowledge that time perspective can be manipulated, resulting in FTP that is different from what would be predicted based on age alone (e.g., in younger adults with terminal illness). A rich literature documenting the influence of goals on information processing (Neisser, 1979) suggests these age-related changes have important implications for cognition and behavior relevant to emotion regulation. One mechanism of emotion regulation is through positive relationships and social interactions (Urry & Gross, 2010). Compared to younger adults, older adults report that their choice of social partners is more heavily influenced by the likelihood of a positive interaction (Fredrickson & Carstensen, 1990). Accordingly, older adults tend to invest time and energy in a selective group of meaningful and emotionally beneficial social partners (English & Carstensen, 2014; Lang, 2000). This importance of relationships in older adults’ lives may also be reflected in how they select and choose to narrate their personal life experiences.

SST and Autobiographical Memory AM is a highly reconstructive and self-relevant process, shaped by an individual’s goals and motivations at the time of retrieval (Conway, 2005). These emotional and relational goals may influence the way in which older adults recall and narrate personal life events. Indeed, compared to younger adults, the self-reported ratings of the AMs of older adults are more positive (Tomaszczyk & Fernandes 2012) and narrative content analysis demonstrates a greater emotional focus (e.g., Schryer, Ross, St. Jacques, Levine, & Fernandes, 2012). Historically, rater coding has been the main methodology of narrative content analyses, ranging from broad or global themes to small details, ideas, and/or constructs (Pasupathi, Mansour, & Brubaker, 2007). A more recent

GeroPsych (2020), 33(1), 3–14

A. J. Polsinelli et al., Emotional Autobiographical Memory

and complementary approach involves examining narrative content at the level of frequency of individual and categories of words (e.g., pronouns), using computerized text analysis programs. Although coding and text analysis may appear redundant, they are independent and complimentary tools for analyzing narratives (Weston, Cox, Condon, & Jackson, 2016). Using computerized text analysis to examine word use allows researchers to indirectly track an individual’s attentional focus during narration (Tausczik & Pennebaker, 2010). The words that people spontaneously choose demonstrate what is most important to them at a particular time, and in this way it creates a window into their current set of values, goals, and emotions. The use of substantive words, like verbs and nouns, is likely under conscious control and can reveal important themes (e.g., family, happiness). In contrast, function words, specifically personal pronouns, are less consciously monitored and can reveal whether one’s focus is on the self (first-person singular pronouns; I-talk) or others (third-person pronouns; she, he, they; Mehl, Raison, Pace, Arevalo, & Cole, 2017; Pennebaker, 2011; Pennebaker, Mehl, & Niederhoffer, 2003).

SST and Word Use Computerized text analysis of the verbal communications and narratives of older and younger adults (e.g., oral interviews, expressive writing) has revealed age-related differences in relational or interpersonal focus. The social media posts of older adults frequently concern family and relationships, something less common among younger adults (Kern et al, 2014; Schwartz et al., 2013), and there appears to be a shift in pronoun use across the lifespan. While I-talk decreases with age, first-person plural (we-talk) and third-person pronouns increase with age (Pennebaker & Stone, 2003; Schwartz et al., 2013; cf. Neysari et al., 2016) suggesting a shift from focusing on the self to focusing on others. Studies using computerized text analysis to examine interpersonal focus in the emotional AMs of older and younger adults are relatively scarce and offer mixed results. Robertson and Hopko (2013) found that, although older adults tend to use more family-related words in positive AMs compared to younger adults, pronoun use did not differ. There are several caveats to these findings including failure to (1) separate I-talk from we-talk in analyses, (2) control for overall pronoun use, (3) compare positive to negative AMs within age groups, and (4) control for the age of recalled AMs (i.e., participants were asked to recall only one positive and one negative life event from any time in their lives).

Ó 2020 Hogrefe

A. J. Polsinelli et al., Emotional Autobiographical Memory

Present Study The present study used rater codings and computerized text analysis to assess interpersonal focus in the emotional AMs of older and younger adults. Specifically, we used rater codings to examine the overall relational orientation or theme of the AMs and computerized text analysis (Pennebaker, Francis, & Booth, 2015) to examine social word and pronoun use. We aimed to replicate previous work showing older adults to be more interpersonally focused in general and extend the research by showing the specificity of the interpersonal focus to positive life experiences (i.e., AMs), as would be predicted by SST. In manipulating the emotional valence of retrieval cues, we were able to examine content that older adults spontaneously associate with positive and negative affective experiences. Additionally, using computerized text analysis allowed us to access memory retrieval processes that are not necessarily available when relying on self-report or rater codings. For example, self-report ratings of AMs engage conscious deliberation and self-evaluation (i.e., metacognition) that includes an appraisal of the relevance of the memory for the self (Schwarz, 1996). In turn, spontaneous word use, particularly pronoun use, revealed through computerized text analysis, is typically not under conscious cognitive control (Pennebaker, 2011; Pennebaker et al., 2003) and may therefore tap more automatic cognitive and emotional processes that largely “bypass” meta-cognitive processes during spontaneous retrieval.

Hypotheses In accordance with SST, we hypothesized that the positive AMs of older adults would be more interpersonally focused compared to their negative AMs and compared to the positive AMs of younger adults. We do not expect to see this pattern (interpersonal focus: positive > negative) in younger adults. Specifically, relational orientation (coded by raters), social word use (text analysis), and other-focused pronouns (we-talk, third-person pronouns; text analysis) should be higher in the positive AMs of older adults (compared to their negative AMs and the positive AMs of younger adults), reflecting interpersonal focus. In contrast, we expected that older adults would use less I-talk in positive AMs (compared to their negative AMs and the positive AMs of younger adults), consistent with a shift away from self-focus.

Method Participants Participants were 70 older (n = 45) and younger (n = 25) adults. Originally, 55 older adult participants (range Ó 2020 Hogrefe

65–90) were selected from a larger study of emotional AM and were included if they gave permission to use their data in subsequent studies. Forty-four younger adults (range 18–26) were initially recruited from introductory psychology courses for the same larger study of emotional AM. They received course credit for their involvement. Exclusion criteria for both groups included (1) neurological disorder or injury, (2) history of substance abuse, (3) significant psychiatric history, (4) current psychotropic medication use, and (5), consistent with standard clinical administration, scores of 10 or higher on the Geriatric Depression Scale (GDS; Yesavage et al., 1982) or 14 or higher on the Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996). Fifteen participants were excluded because of (1) significant depressive symptomatology (13 younger adults), (2) inability to complete the AM task (one older adult), or (3) recent history of alcohol use disorder (one older adult). In addition, technical problems with audio equipment prevented the recording of six younger and eight older adults AMs. Thus, the data for the present analyses included 45 older adults and 25 younger adults. Unfortunately, because of practical limitations on recruitment, we were unable to recruit additional younger adults, resulting in uneven samples. The demographic information for each group is shown in Table 1. Older adults had higher levels of education, t(68) = 9.41, p < .001, d = 2.35, and, consistent with SST, reported higher positive affect (measured with the Positive and Negative Affect Scale [PANAS; Watson, Clark, & Tellegen, 1988]), t(68) = 3.55, p = .003, d = .76, and lower negative affect (PANAS), t(68) = 2.70, p = .014, d = .74. There were a greater number of women than men in both groups, but no statistically significant difference between groups, w2(1) = 2.90, p = .088.

Procedure All participants gave informed consent. Procedures were approved by the University of Arizona’s Institutional Review Board. Participants completed the GDS (older adults) or the BDI-II (younger adults) to rule out significant depressive symptomatology, a brief health questionnaire to rule out exclusion criteria, and the PANAS. Participants then completed the AM task. Autobiographical Memory Task (AM) In the complete protocol, participants recalled three positive, three negative, and three neutral AMs. However, neutral AMs were excluded from analyses as the focus was on the comparison between positive and negative AMs (see Appendix A for neutral AM data). AMs were required to have occurred between 1 month and 3 years ago. The recent time frame was chosen to reduce the inherent difficulty of GeroPsych (2020), 33(1), 3–14

comparing the remote AMs of younger and older adults because of substantial differences in the age and importance of the memory (Rubin, Rahhal, & Poon, 1998). A research assistant provided participants with an explanation and example of an AM, adapted from Lemogne and colleagues (2009) and instructions for AM recall were as follows: I would like you to recall (one of your most negative/one of your most positive/a neutral) autobiographical memory that occurred in the past 3 years but is at least 1 month old. A (negative/positive/neutral) autobiographical memory is a memory for an event you perceived as (unpleasant/ pleasant/not particularly pleasant or unpleasant) at the time you were experiencing it. Give as many details as possible about this memory including where and when this event occurred. Participants were then asked to recall three AMs of the specified valence category before moving on to the next category, and their responses were audio recorded. Category order was counterbalanced. Participants recalled each AM without interruption until they reached a natural ending point, at which time there were asked a standardized prompt (i.e., “Is there anything else you can tell me about the memory that you haven’t already?”). Participants were not given word cues, a time limit, or specific instructions regarding the types of details to include (other than requesting a specific date of the event to verify it occurred within the past 3 years). After recalling each AM, participants used Likert scales to rate the AM on a number of qualitative factors. Of the factors selected for the larger emotional AM study, four were deemed relevant to the present study: (1) an “emotionality” scale (emotional valence) was used to check our valence manipulation, (2) an “amount of selfawareness” scale (degree to which participants felt they were focused on themselves during the AM) was used as a self-report measure of self-focus, and (3) “frequency of recalling/retelling the AM” (i.e., rehearsal) and “importance of the AM to the self-concept” scales were used as proxies for assessing potential reminiscence bump effects (greater recall of AMs occurring between ages 10–30; Rubin et al., 1998). We specifically requested recent AMs to prevent older adults from recalling AMs from the reminiscence bump period. However, younger adults were in the midst of the reminiscence bump period, and although Janssen, Chessa, and Murre (2005) demonstrated that the reminiscence bump effect is strongest in later life, it has also been observed in earlier life.

Measures Coding of Relational Orientation Two trained raters, blind to the study hypotheses, coded each AM in its entirety as either relationally focused (on GeroPsych (2020), 33(1), 3–14

A. J. Polsinelli et al., Emotional Autobiographical Memory

Table 1. Means and standard deviations of demographic variables Older adults

Younger adults

Age

76.76 (5.94)

18.64 (.81)

Education

16.73 (2.29)

12.32 (.63)

Sex (% female)

71.11

64.00

GDS/BDI-II

3.56 (2.64)

5.12 (3.48)

PANAS-positive

37.24 (6.97)

31.20 (6.56)

PANAS-negative

11.60 (1.88)

13.24 (3.21)

Note: BDI-II = Beck Depression Inventory – II; GDS = Geriatric Depression Inventory; PANAS = Positive and Negative Affect Scale.

others or interactions) or individually focused (on self), oriented with criteria adapted from Wang and Ross (2005). (See Appendix B for the full coding procedures and criteria.) The two sets of ratings showed adequate convergence (ICC[2,2] = .77), but in the case of disagreement, a third blinded rater independently coded the AMs to resolve the discrepancy. Orientation coding scores reflected the proportion of AMs in a valence category (positive vs. negative) that was relationally oriented, with higher scores indicating greater relational orientation. Word Use Verbatim transcripts of each audio-recorded AM were subjected to linguistic inquiry and word count (LIWC; Pennebaker et al., 2015) software, which generates wordcategory scores that are proportions of total word count (in this case, within each individual AM). The social words category – comprised of nouns (e.g., friend, neighbor), verbs (e.g., spoke, fought), and personal pronouns (e.g., you, they) that reflect an interpersonal focus or exchange – was validated in prior research (Tausczik & Pennebaker, 2010). LIWC also extracted scores for first-person singular (I-talk), first-person plural (we-talk) and third-person (he, she, they) pronouns. To control for individual variability in overall pronoun use, we calculated a set of ratio scores for each pronoun variable (I-talk, we-talk, and third-person pronouns) that represented each variable as a proportion of the total pronoun use (e.g., we-talk/(I-talk + we-talk + third-person pronouns)). Analyses For LIWC variables, we calculated an average from the three AM scores within each valence category, and these averaged proportion scores served as dependent variables for the study. Nine of the older participants were unable to retrieve three AMs for a particular valence category, resulting in 15 missing data points. Thus, for five participants, one of their valence category scores reflects only one AM and for four participants it reflects an average of two AMs. All analyses were performed using 2 (Age group: older, younger) 2 (valence: positive, negative) mixed model ANOVAs. Ó 2020 Hogrefe

A. J. Polsinelli et al., Emotional Autobiographical Memory

did not significantly differ in overall AM importance, F(1,68) = .43, p = .512, η2 < .01, and the Age Valence interaction was not significant, F(1,68) = 1.41, p = .239, η2 = .02.

Results Preliminary Analyses Mean scores for the measures below in older and younger adults’ positive and negative AMs appear in Table 2. AM Age The age of AMs was not significantly different between positive and negative AMs, F(1,68) = 1.63, p = .207, η2 = .02, or older and younger adults, F(1,68) = 2.02, p = .159, η2 = .03. The Age Valence interaction was also not significant, F(1,68) = .03, p = .871, η2 < .01. Self-Rated Emotionality Consistent with the manipulation, all participants rated their positive AMs more positively and their negative AMs more negatively, F(1,67) = 954.73, p < .001, η2 = .93. Younger and older adults did not differ in self-reported AM emotionality, F(1,68) = .44, p = .510, η2 = .01, and the Age valence interaction was not significant, F(1,68) = 1.42, p = .237, η2 = .02. Self-Rated Self-Awareness Younger and older adults did not significantly differ in overall self-awareness, F(1,68) = 2.64, p = .109, η2 = .04. Selfawareness was higher in positive AMs than in negative AMs, F(1,68) = 8.15, p = .006, η2 = .11, and this was qualified by a significant Age Valence interaction, F(1,68) = 9.99, p = .002, η2 = .13. Older adults reported similar self-awareness regardless of AM valence, t(44) = .26, p = .797, d = .05, and reported an amount of self-awareness in positive AMs that was similar to younger adults, t(68) = .79, p = .430, d = .20. Younger adults were more self-aware during their positive compared to negative AMs, t(24) = 3.74, p = .001, d = 1.07. Self-Rated Importance (to Self-Concept) Importance was higher in positive AMs than negative AMs, F(1,68) = 27.50, p < .001, η2 = .29. Younger and older adults

Self-Rated Frequency of Retelling/Recalling Positive and negative AMs were recalled at similar frequencies, F(1,68) = .00, p = .994, η2 < .01. Younger and older adults did not differ in the number of times they had recalled their AMs, F(1,67) = .89, p = .348, η2 = .01, and the Age Valence interaction was not significant, F(1,68) = .18, p = .673, η2 < .01. Word Count Participants used a similar number of words when recalling negative and positive AMs, F(1,68) = 2.81, p = .098, η2 = .04; there was no significant main effect of age, F(1,68) = 3.07, p = .084, η2 = .04, and no significant Age Valence interaction, F(1,68) = .16, p = .692, η2 < .01. Affective Words Affective words were used similarly within AM valence categories, F(1,68) = .59, p = .446, η2 < .01, and among older and younger adults, F(1,68) = .14, p = .707, η2 < .01. There was no significant Age Valence interaction, F(1,68) = 2.54, p = .115, η2 = .04. Gender Differences We explored potential gender differences in the preliminary and primary outcome measures in both samples. There were no statistically significant gender differences in either sample (all ps > .05) with two exceptions for our preliminary outcomes: Older women (M = 8.29, SD = .86) rated the emotionality of their positive memories higher than older men (M = 7.54, SD = .73), t(43) = 11.56, p = .027, d = .89: and younger men (M = 4.56, SD = 1.60) used more affect words to describe their positive AMs than younger women (M = 3.21, SD = 1.07), t(23) = 2.54, p = .018, d = .79.

Table 2. Means and standard deviations for emotionality, age, and length of positive and negative AMs Older adults

Younger adults

Positive AMs

Negative AMs

Positive AMs

Negative AMs

η2

Age of memory (months)

11.36 (7.64)

13.03 (10.31)

9.27 (4.93)

10.56 (5.30)

.871

< .01

Self-rated emotionality

8.16 (0.88)

2.02 (0.81)

7.85 (1.00)

2.17 (0.96)

.237

.02

Self-rated self-awareness

6.78 (1.97)

6.88 (1.93)

7.15 (1.52)

5.33 (1.85)

.002

.13

Self-rated importance

6.23 (1.68)

4.53 (2.42)

6.19 (1.54)

5.12 (1.91)

.239

.02

Self-rated frequency of recall

4.20 (2.02)

4.09 (1.93)

4.47 (1.66)

4.58 (1.79)

.673

.00

Total word count

268 (170)

294 (209)

193 (116)

236 (154)

.692

< .01

.04 (.02)

.04 (.01)

.115

.04

Proportion of affective words

Note: AMs = Autobiographical memories. Participants rated emotionality of their AMs on a Likert scale ranging from 1 (very negative) to 9 (very positive). Participants rated self-awareness of their AMs on a Likert scale ranging from 1 (not at all self-aware) to 9 (very self-aware). Participants rated importance of the AM to their self-concept on a Likert scale ranging from 1 (not at all important) to 9 (very important). Participants rated frequency of recall on a Likert Scale ranging from 1 (never recalled) to 9 (recalled frequently). P-values and η2 associated with Age Valence interaction term.

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 3–14

A. J. Polsinelli et al., Emotional Autobiographical Memory

Primary Analyses Table 3 provides examples of positive and negative AMs for older and younger adults as well as the most frequent AM themes across ages and valences. The means and standard deviations for primary outcome variables are presented in Table 4. To account for multiple comparisons (i.e., five mixed-model ANOVAs), we used Bonferroni’s correction, which lowered our significance value to p < .01 for each interaction term. Rater Coding for Relational Orientation There was no significant main effect of age, F(1,68) < .01, p = .980, η2 < .001, on relational orientation. The significant main effect of valence, F(1,68) = 6.25, p = .015, η2 = .08, was qualified by an Age Valence interaction, F(1,68) = 12.48, p = .001, η2 = .15. Older adults showed greater relational orientation when recalling positive AMs compared to (1) negative AMs, t(44) = 4.99, p < .001, d = .98, and (2) the positive AMs of younger adults,

t(68) = 2.45, p = .017, d = .61. Younger adults did not show valence differences in relational orientation, t(24) = 0.66, p = .52, d = .17. Social Word Use There was no significant main effect of age on social word use, F(1,68) = .81, p = .370, η2 = .01. The main effect of valence, F(1,68) = 4.83, p = ..031, η2 = .07, was qualified by an Age Valence interaction, F(1,68) = 6.53, p = .012, η2 = .09. Specifically, older adults used more social words when recalling positive AMs compared to (1) negative AMs, t(44) = 3.79, p < .001, d = .67, and (2) the positive AMs of younger adults, t(68) = 2.44, p = .017, d = .67. There was no significant valence difference for the social word use of younger adults, t(24) = .25, p = .807, d = .33. I-talk Consistent with prior language research on the effects of age (Pennebaker & Stone, 2003; Kern et al., 2014) and affect (Rude, Gortner, & Pennebaker, 2004), younger

Table 3. Themes and examples of emotional autobiographical memories for older and younger adults Most frequent themes Older adults

Positive

Family gathering (e.g., birthdays) Friendship Travel

Negative Injury Death (pets, family members) Interpersonal conflict (generally with nonfamily members)

Younger adults

Positive

Personal accomplishment (e.g., graduation, getting driver’s license) Romantic relationship Friendship

Negative Interpersonal conflict (especially with significant other and family members) Car accident Injury

GeroPsych (2020), 33(1), 3–14

Example [positive, interpersonal-focus] Ok, in June of 2010 my granddaughter XXXX graduated from high school, and we were there to observe it at her high school, XXXX, in North Carolina, and it was wonderful. We sat in the second row off to the right just a little bit. It was so hot. She received high honors and walked up there on stage to get her diploma. She was smiling so much. Oh, we were just very pleased to share in her joy with her and her family. My daughter and son-in-law and our grandson were all there. [negative, self-focus] Ok, this would’ve been about in June of last year and we were in South Dakota taking a walk. We walk around the lake and over to a dock and then we walk back home, but we come back a different way when we come back home. We got there the gate was locked, there was a gate across it. It’s a gate that has solid bars on it so it’s kind of like a ladder you can climb up over it. And I got up on the top of it and I had my hands full of some garbage I’d picked up along the road. And I wasn’t hanging on very carefully and I fell off the top of it and landed on my back and it was excruciating. And my husband kept saying “Can I help you? Can I pull you up?” And I says “No, leave me alone.” And finally managed to get up on my feet and he said “I’ll go home and get the car” and I says “No, I’ll walk home somehow.” And then we walked very slowly the rest of the way home which was about half a mile. And then I just took some painkiller and kind of laid down and rested for the rest of the day. I thought, “Boy I really hurt myself this time, that was really stupid.” There was no reason to fall if I had been paying attention and watching myself. [positive, self-focus] Finishing my first semester of college, so December 2010. The day I took my last final I just felt relieved and after taking my last final I went home, and I just slept I was so tired. And just glad and so happy that it was done. I got up and panicked for a second before I realized I was done. Done. I was supposed to go out with friends, but I was so tired. I just stayed home and slept. I think I made mac and cheese for dinner or something just as bad. I experienced the first semester. I just looked back on the whole semester and that was it. It was right after my math final, and I was just relieved. [negative, interpersonal-focus] My sister is notorious for taking peoples’ things without asking, and she always takes my clothes and ruins them. Which is frustrating and I once she I remember she took something I told her she could never borrow because I spent a lot of money on it. And it was gone from my closet and I freaked out and I took all of her clothes out of her closet and put it in my parents’ room, and then she came home. I was yelling at her and we got in a huge fight. I was really upset. It was a year ago. Maybe March. One of my sister’s friends was with her, and they were in her room and started talking about it, and I could hear them and then I went in her room and got really mad again. I started yelling at them and they yelled back at me. It was a mess. We didn’t talk for a few days after because we were both so mad.

Ó 2020 Hogrefe

A. J. Polsinelli et al., Emotional Autobiographical Memory

Table 4. Means and standard deviations for main outcome variables in positive and negative AMs Older adults

Relational orientation

Younger adults η2

Positive AMs

Negative AMs

Positive AMs

Negative AMs

.81 (.24)a,c

.50 (.34)a

.63 (.31)c

.68 (.28)

.001

.15

a,c

.11 (.03)c

.12 (.04)

.012

.09

.004

.12

Social word use

.13 (.03)

I-talk

.44 (.14)a,d

.61 (.14)a

.58 (.14)d

.63 (.14)

Third-person pronouns

.35 (.13)b,d

.28 (.13)b

.19 (.08)b,d

.25 (.09)b

We-talk

.21 (.14)

.11 (.08)

.23 (.11)

.12 (.11)

.11 (.03)

< .001

.17

.945

< .01

Note. AMs = Autobiographical memories. p-values and η2 associated with Age Valence interaction term. Significant paired-samples t-tests: ap < .001; bp .01. Significant independent samples t-tests: cp < .05; dp < .001.

adults used more I-talk than did older adults, F(1,68) = 8.51, p = .005, η2 = .11, and participants used more I-talk when recalling negative compared to positive AMs, F(1,68) = 28.03, p < .001, η2 = .29. These main effects were qualified by an Age Valence interaction, F(1,68) = 9.10, p = .004, η2 = .12. Specifically, older adults used less I-talk in positive AMs compared to (1) negative AMs, t(44) = 6.45, p < .001, d = 1.00, and (2) the positive AMs of younger adults, t(68) = 4.19, p < .001, d = 1.21. Younger adults used similar amounts of I-talk across valence categories, t(24) = 1.69, p = .103, d = .35. Third-Person Pronouns There was no significant main effect of valence on use of third-person pronouns, F(1,68) = .24, p = .623, η2 < .01. The main effect of age, F(1,68) = 18.29, p < .001, η2 = .21, was qualified by an Age Valence interaction, F(1,68) = 14.15, p < .001, η2 = .17. Specifically, older adults used more third-person pronouns when recalling positive AMs compared to (1) negative AMs, t(24) = 3.19, p = .003, d = .56, and (2) the positive AMs of younger adults, t(68) = 4.85, p < .001, d = 1.39. Younger adults used more third-person pronouns when recalling negative compared to positive AMs, t(24) = 2.76, p = .01, d = .59. We-Talk Participants used more we-talk when recalling positive compared to negative AMs, F(1,68) = 44.83, p < .001, η2 = .40. There was no significant main effect of age on we-talk, F(1,68) = .50, p = .480, η2 < .01. The Age Valence interaction was also not significant, F(1,68) < .01, p = .945, η2 < .01.

Discussion We examined interpersonal focus directly in the content (rater coding and social word use) and indirectly in the pronoun use of the emotional AMs of older and younger adults. Consistent with our hypotheses, based on SST, content analyses suggested that older adults were more relationally oriented (rater codings) and used more social words (LIWC) in their positive AMs compared to (1) their negative AMs and (2) the positive AMs of younger adults. Analyses of Ó 2020 Hogrefe

pronoun use similarly confirmed the hypotheses that older adults were less focused on the self (i.e., less I-talk) and more focused on others (i.e., higher third-person pronoun use) when recalling positive AMs compared to (1) the negative AMs and (2) the positive AMs of younger adults. Across measures, as hypothesized, younger adults did not show valence differences, with one exception: Younger adults used more third-person pronouns in their negative AMs, suggesting a focus on others during negative compared to positive AM recall. We found only partial support for hypotheses involving we-talk.

Emotion and Interpersonal Focus in AMs The results of this study contribute to the SST literature and speak to the importance of emotional valence (positive vs. negative) in examining interpersonal focus. Older adults selected social/relational interactions to narrate when cued for a positive life event. They appeared less attentive to their own presence and/or perspective and more attentive to the presence, actions, and perhaps perspectives of others in positive AMs. Interestingly, older adults reported a similar degree of self-awareness during the AM event regardless of valence, suggesting self-report might be a less sensitive method for capturing these attentional differences in older adults. In younger adults, valence either did not influence interpersonal focus or produced the opposite pattern – a focus on others during negative AMs. Indeed, an overwhelmingly frequent theme of the negative AMs of younger adults was interpersonal conflict, especially with significant others and family members. Additionally, and in contrast to older adults, younger adults reported being less self-aware during negative AM events compared to positive AM events (which were overwhelmingly about personal achievements such as graduation). This focus on others was captured only by third-person pronouns, suggesting that pronouns may be a uniquely sensitive measure of attentional focus in younger adults.

Age Differences in the Use of We-Talk Contrary to our hypotheses and the majority of prior research (cf., Neysari et al., 2016), older adults did not GeroPsych (2020), 33(1), 3–14

use more we-talk than younger adults overall. Further, higher we-talk in positive than negative AMs was not specific to older adults. Compared to I-talk and third-person pronouns, we-talk is a unique pronoun category that may represent aspects of both self and others. “We” includes the self but also implies the presence of others and/or the extension of self to include others. In this way, it may represent a middle ground between two extreme ends of the pronoun-use continuum: I-talk (complete self-focus) and third-person pronouns (complete other-focus). Though there were no age differences in overall rates of we-talk, our findings hint that it may be a worthy scientific pursuit to investigate potential differences in how older and younger adults use we-talk. For example, younger adults may use “we” to indicate the presence of others – but from a focus that is still solidly on the self (as interpersonal focus was generally equal across valence categories but third-person pronoun use was higher in negative than positive AMs). In contrast, older adults may use “we” in a more interpersonal way that reflects a true focus on others with a linguistic “acknowledgement” that the self is also present (evidenced by greater interpersonal focus and third-person pronoun use in positive than negative AMs).

Possible Differences in Self and Other Perspective As already discussed, in the context of positive AMs, older adults, but not younger adults, appear to shift away from focusing on the self toward focusing on others. But this pattern of pronoun use may also represent a decrease in using one’s own perspective and an increase in taking another’s perspective in narrating a positive personal event. For example, one older adult stated, “He must have been so happy when they visited,” reflecting insight into another’s emotional state. In this way, pronouns have the potential to track the narrator’s point of view in the retrieval process and provide information about how point of view is related to emotional experiences at the time of recall. Thus, taking another person’s perspective may be associated with positive experience in later life. Indeed, prior work suggests that older adults who are more empathetic experience more positive and meaningful social interactions (Gruhn, Rebucal, Diehl, Lumley, & Labouvie-Vief, 2008).

Methodological Considerations Researchers studying personal narratives have embraced computerized text analysis (e.g., Cox & McAdams, 2019; Himmelstein, Barb, Finlayson, & Young, 2018; Hirsh & Peterson, 2009), which complements more traditional self-report and narrative coding approaches. For example, directly comparing computerized text analysis and GeroPsych (2020), 33(1), 3–14

A. J. Polsinelli et al., Emotional Autobiographical Memory

narrative coding reveals that they capture distinct information and independently predict outcomes (Weston et al., 2016). Although we acknowledge a common and valid criticism of text analysis, namely, that it cannot account for the context in which words are used, the resistance of this methodology to self-report and rater biases may offset this limitation (Mehl, 2006; Tausczik & Pennebaker, 2010). In the context of self-report, self-evaluation and word use arguably tap different, albeit complementary, cognitive and memory processes. Whereas self-evaluation is a reflective, metacognitive process, word use may more closely align with dynamic processes of re-experiencing as the event unfolds through narration in the present moment. Another important methodological consideration is that we attempted to reduce biases in AM content by refraining from providing any specific cue words or event types during the recall task. We believe this may be closer to emulating a more naturalistic memory retrieval process – while acknowledging that this was not entirely akin to spontaneous or involuntary retrieval (Schlagman & Kvavilashvili, 2008). As AM is a dynamic reconstruction of a past event, it will be important in future studies to tease apart whether interpersonal focus in AMs of older adults reflects (1) actual prior experiences (e.g., positive interpersonal interactions are more frequent in later life), (2) cognitive biases at encoding (e.g., enhanced encoding for positive interpersonal interactions) or retrieval (e.g., active forgetting of negative interpersonal interactions or increased accessibility of positive interpersonal interactions), and/or (3) the emotional, psychological, and social functions that AM may serve (Bluck, Alea, Habermas, & Rubin, 2005; Wolf & Demiray, 2019). For example, younger adults rate selfcontinuity as a more important function of their AMs than middle-aged adults (Demiray & Freund, 2014), and older adults tend to reminisce for social functions (Webster & Gould, 2007). These questions will have important implications for how social structure and AM function in later life, particularly as they relate to emotional experience and regulation.

Limitations Sample sizes were uneven, there was a notable gender bias favoring women, and participants were not demographically representative of the American population. Replication is needed in larger and more diverse samples. The data were not collected at multiple time points, which would have assisted in demonstrating the reliability and stability of these content and language variables over time. Although SST posits that age is inextricably associated with time perspective (Carstensen et al., 1999), we did not include a direct measure of FTP. In future research, it will be Ó 2020 Hogrefe

A. J. Polsinelli et al., Emotional Autobiographical Memory

important to obtain this measure to verify its association with age and primary outcomes. There are two additional considerations relevant to results from the direct comparison of the positive AMs of older and younger adults: potential age differences in the recency effect – especially as it relates to the reminiscence bump – and AM specificity. Regarding the recency effect and reminiscence bump, younger adults were in the midst of the reminiscence bump period. We did not find any age differences in factors potentially related to the reminiscence bump (i.e., self-rated frequency of AM recalling/ retelling and importance of the AM for the self-concept). However, these data cannot rule out the possibility of reminiscence bump effects and broader differences in the recency effect which might contribute to age differences in interpersonal focus in positive AMs. With respect to the second consideration, younger and older adults differ in the degree of AM specificity (e.g., greater semantic or generalized knowledge and less episodic detail in older compared to younger adults’ AMs; Levine, Svobada, Hay, Winocur, & Moscovitch, 2002). We did not measure AM specificity or frequency of AM appraisal/evaluation (which would arguably result in more pronoun use, likely more I-talk) and therefore cannot rule out the possibility that these factors contributed to the age differences in interpersonal focus in positive AMs.

Conclusion The self-selected emotional AMs of older adults suggest that greater interpersonal focus is associated with positive life experiences, more so than what is found in young adulthood. In contrast, focusing on others (as measured with pronoun use) is associated with negative life experiences in younger adults. We believe these findings reflect the importance of interpersonal interactions in generating and maintaining positive experiences, specifically for older adults. These results can best be understood within the context of SST and suggest important implications for the emotional benefits of interpersonal engagement for older adults.

References Beck, A. T., Steer, R. A., & Brown, G. K. (1996). Manual for the Beck Depression Inventory-II. San Antonio, TX: Psychological Corporation. Bluck, S., Alea, N., Habermas, T., & Rubin, D. C. (2005). A tale of three functions: The self–reported uses of autobiographical memory. Social Cognition, 23, 91–117. Carstensen, L. L., Fung, H. H., & Charles, S. T. (2003). Socioemotional selectivity theory and the regulation of emotion in the second half of life. Motivation and Emotion, 27, 103–123.

Ó 2020 Hogrefe

Carstensen, L. L., Isaacowitz, D. M., & Charles, S. T. (1999). Taking time seriously: A theory of socioemotional selectivity. American Psychologist, 54, 165–181. Conway, M. A. (2005). Memory and the self. Journal of Memory and Language, 53, 594–628. Cox, K. S., & McAdams, D. P. (2019). Mean level differences in the vividness, meaning, and coherence of life story high and low points: How valenced life stories do and do not differ. Identity, 19, 128–143. Demiray, B., & Freund, A. M. (2014). Michael Jackson, Bin Laden and I: Functions of positive and negative, public, and private flashbulb memories. Memory,1–14. English, T., & Carstensen, L. L. (2014). Selective narrowing of social networks across adulthood is associated with improved emotional experience in daily life. International Journal of Behavioral Development, 38, 195–202. Fredrickson, B. L., & Carstensen, L. L. (1990). Choosing social partners: How old age and anticipated endings make people more selective. Psychology and Aging, 5, 335–347. Gruhn, D., Rebucal, K., Diehl, M., Lumley, M., & Labouvie-Vief, G. (2008). Empathy across the adult lifespan: Longitudinal and experience-sampling findings. Emotion, 8, 753–765. Himmelstein, P., Barb, S., Finlayson, M. A., & Young, K. D. (2018). Linguistic analysis of the autobiographical memories of individuals with major depressive disorder. PLoS One, 13(11), e0207814. Hirsh, J. B., & Peterson, J. B. (2009). Personality and language use in self-narratives. Journal of Research in Personality, 43, 524–527. Janssen, S., Chessa, A., & Murre, J. (2005). The reminiscence bump in autobiographical memory: Effects of age, gender, education, and culture. Memory, 13, 658–668. Kern, M. L., Eichstaedt, J. C., Schwartz, H. A., Park, G., Ungar, L. H., Stillwell, D. J., . . . Seligman, M. E. P. (2014). From “Sooo excited!!!” to “So proud”: Using language to study development. Developmental Psychology, 50, 178–188. Lang, F. R. (2000). Endings and continuity of social relationships: Maximizing intrinsic benefits within personal networks when feeling near to death. Journal of Social and Personal Relationships, 17, 155–182. Lang, F. R., & Carstensen, L. L. (2002). Time counts: Future time perspective, goals, and social relationships. Psychology and Aging, 17, 125–139. Lemogne, C., Bergouignan, L., Piolino, P., Jouvent, R., Allilaire, J.-F., & Fossati, P. (2009). Cognitive avoidance of intrusive memories and autobiographical memory: Specificity, autonoetic consciousness, and self-perspective. Memory, 17, 1–7. Levine, B., Svoboda, E., Hay, J. F., Winocur, G., & Moscovitch, M. (2002). Aging and autobiographical memory: Dissociating episodic from semantic retrieval. Psychology and Aging, 17, 677–689. Mehl, M. R. (2006). Quantitative text analysis. In M. Eid & E. Diener (Eds.), Handbook of psychological measurement: A multimethod perspective. Washington, DC: American Psychological Association. Mehl, M. R., Raison, C. L., Pace, T. W., Arevalo, J. M., & Cole, S. W. (2017). Natural language indicators of differential gene regulation in the human immune system. Proceedings of the National Academy of Sciences, 114, 12554–12559. Neisser, U. (1979). The control of information pickup in selective looking. Perception and its development: A tribute to Eleanor J. Gibson, 201–219. Neysari, M., Bodenmann, G., Mehl, M. R., Bernecker, K., Nussbeck, F. W., Backes, S., . . . Horn, A. B. (2016). Monitoring pronouns in conflicts: Temporal dynamics of verbal communication in couples across the lifespan. GeroPsych, 29, 201–213. Pasupathi, M., Mansour, E., & Brubaker, J. R. (2007). Developing a life story: Constructing relations between self and experience in autobiographical narratives. Human Development, 50, 85–110.

GeroPsych (2020), 33(1), 3–14

Pennebaker, J. W. (2011). The secret life of pronouns. New York, NY: Bloomsbury. Pennebaker, J. W., Francis, M. E., & Booth, R. J. (2015). Linguistic inquiry and word count: LIWC 2015. Austin, TX: LIWC (www.liwc. net). Pennebaker, J. W., Mehl, M. R., & Niederhoffer, K. G. (2003). Psychological aspects of natural language use: Our words, our selves. Annual Review of Psychology, 54, 547–577. Pennebaker, J. W., & Stone, L. D. (2003). Words of wisdom: Language use over the life span. Journal of Personality and Social Psychology, 85, 291–301. Penningroth, S. L., & Scott, W. D. (2012). Age-related differences in goals: Testing predictions form selection, optimization, and compensation theory and socioemotional selectivity theory. The International Journal of Aging & Human Development, 74, 87–111. Robertson, S. M. C., & Hopko, D. R. (2013). Emotional expression during autobiographical narratives as a function of aging: Support for the socioemotional selectivity theory. Journal of Adult Development, 20, 76–86. Rubin, D. C., Rahhal, T. A., & Poon, L. W. (1998). Things learned in early adulthood are remembered best. Memory & Cognition, 26, 3–19. Rude, S., Gortner, E. M., & Pennebaker, J. (2004). Language use of depressed and depression-vulnerable college students. Cognition & Emotion, 18, 1121–1133. Schlagman, S., & Kvavilashvili, L. (2008). Involuntary autobiographical memories in and outside the laboratory: How different are they from voluntary autobiographical memories? Memory & Cognition, 36, 920–932. Schryer, E., Ross, M., St. Jacques, P., Levine, B., & Fernandes, M. (2012). Emotional expressivity in older and younger adults’ descriptions of personal memories. Experimental Aging Research, 38, 345–369. Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Dziurzynski, L., Ramones, S. M., Agrawal, M., . . . Ungar, L. H. (2013). Personality, gender, and age in the language of social media: The open-vocabulary approach. PLoS One, 8(9), e73791. Schwarz, N. (1996). Cognition and communication: Judgmental biases research methods and the logic of conversation. Hillsdale, NJ: Erlbaum. Tausczik, Y. R., & Pennebaker, J. W. (2010). The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology, 29(1), 24–54. Tomaszczyk, J. C., & Fernandes, M. A. (2012). A positivity effect in autobiographical memory, but not phonemic fluency, in older adults. Aging, Neuropsychology, and Cognition, 19, 699–722. Urry, H. L., & Gross, J. J. (2010). Emotion regulation in older age. Current Directions in Psychological Science, 19, 352–357.

GeroPsych (2020), 33(1), 3–14

A. J. Polsinelli et al., Emotional Autobiographical Memory

Wang, Q., & Ross, M. (2005). What we remember and what we tell: The effects of culture and self-priming on memory representation and narratives. Memory, 13, 594–606. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scale. Journal of Social Psychology, 54, 1063–1070. Webster, J. D., & Gould, O. (2007). Reminiscence and vivid personal memories across adulthood. International Journal of Aging and Human Development, 64, 149–170. Weston, S. J., Cox, K. S., Condon, D. M., & Jackson, J. J. (2016). A comparison of human narrative coding of redemption and automated linguistic analysis for understanding life stories: Redemption and automated linguistic analysis. Journal of Personality, 84, 594–606. Wolf, T., & Demiray, B. (2019). The mood-enhancement function of autobiographical memories: Comparisons with other functions in terms of emotional valence. Consciousness and Cognition, 70, 88–100. Yesavage, J. A., Brink, T. L., Rose, T. L., Lum, O., Huang, V., Adey, M., & Leirer, V. O. (1982). Development and validation of a geriatric depression screening scale: A preliminary report. Journal of Clinical Psychology, 42, 918–946. History Received May 30, 2019 Accepted September 19, 2019 Acknowledgments We are indebted to our team of research assistants – Ira R. Adler, Tomas Martinez, Nathan Sherman, Patrick Murray, Irene Cheff, Michelle Costales, Collin Nelson, Amanda Oliver, Alexandra Logue, Caitlin Berard, Meli’sa Crawford, Natalie Goodreau, and Samantha Contreras – for their dedicated and diligent work. We are also sincerely grateful to our participants, who volunteered their time and energy and who made this research possible. Conflict of Interest The authors declare no conflict of interest. Angelina J. Polsinelli, Ph.D. Department of Psychiatry & Psychology Mayo Clinic 200 1st St. SW Rochester, MN, 55904 USA polsinelli.angelina@mayo.edu

Ó 2020 Hogrefe

A. J. Polsinelli et al., Emotional Autobiographical Memory

Appendix A Data from Neutral Autobiographical Memories Of note, neutral AMs were not part of the study hypotheses, and as a result, the raters did not code neutral AMs for relational vs. individual focus. For this reason, the results involving neutral AMs are reported as supplementary material solely for completeness; these data are purely exploratory, and analyses have not been corrected for multiple comparisons. See Table 1 for means and standard deviations.

Social Word Use Younger and older adults used similar amounts of social words in their neutral AMs, t(68) = 1.46, p = .148, d = ..33. Among older adults, neutral AMs contained similar amounts of social words as negative AMs, t(44) = 1.18, p = .244, d = .33, but fewer social words compared to positive AMs, t(44) = 6.07, p < .001, d = 1.00. Among younger adults, neutral AMs contained fewer social words compared to both positive, t(24) = 2.72, p = .012, d = .67, and negative, t(24) = 3.10, p = .005, d = .85, AMs.

I-Talk Younger and older adults used similar amounts of I-talk in their neutral AMs, t(68) = 1.43, p = .158, d = .35. Among

older adults, neutral AMs contained more I-talk compared to positive, t(44) = 7.66, p < .001, d = 1.24, but similar amounts of I-talk compared to negative AMs, t(44) = .97, p = .339, d = .19. Among younger adults, neutral AMs contained more I-talk compared to both positive, t(24) = 3.40, p = .002, d = .83, and negative, t(24) = 2.53, p = .018, d = .58, AMs.

Third-Person Pronouns Older adults used more third-person pronouns than younger adults in their neutral AMs, t(68) = 2.84, p = .006, d = .72. Among older adults, neutral AMs contained fewer third-person pronouns than both negative, t(44) = 4.40, p < .001, d = .91, and positive, t(44) = 8.83, p < .001, d = 1.50, AMs. The pattern was the same in younger adults (compared to: positive, t(24) = 4.96, p < .001, d = 1.20; negative, t(24) = 6.94, p < .001, d = 1.86).

We-Talk Younger and older adults used similar amounts of we-talk in their neutral AMs, t(68) = .24, p = .808, d = .08. Among older adults, neutral AMs contained similar amounts of we-talk as positive AMs, t(44) = 1.16, p = .251, d = .22, but more we-talk than negative AMs, t(44) = 3.97, p < .001, d = .68. The pattern was the same in younger adults (compared to positive, t(24) = 1.19, p = .247, d = .33; negative, t(24) = 3.67, p = .005, d = .58).

Table A1. Means and standard deviations for main outcome variables in positive, neutral, and negative AMs Older adults Positive AMs .13 (.03)

.44 (.14)

Third-person pronouns

.35 (.13)

We-talk

.21 (.14)

Social word use I-talk

Neutral AMs .10 (.03)

.11 (.03)

.64 (.18) .17 (.11) .18 (.13)

Younger adults Negative AMs

Neutral AMs

Negative AMs

.11 (.03)

.09 (.03)

.12 (.04)

.70 (.15)

.63 (.14)

.25 (.09)

.12 (.11)

.58 (.14)

.28 (.13)

.19 (.08)

.11 (.08)

.23 (.11)

.61 (.14) d

Positive AMs

.10 (.07) .19 (.13) b

Notes: AMs = autobiographical memories. Significant paired-samples t-tests comparing to neutral AMs: p < .05; p < .01; p < .001. Significant independent samples t-tests: dp < .01.

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 3–14

Appendix B Relational vs. Individual Coding Procedures and Criteria Each trained, undergraduate rater, blind to study hypotheses, independently rated each AM for overall interpersonal focus. Rating criteria were based on Wang and Ross (2005): Each memory was coded as either “individual” or “social.” Individual memories focused on purely personal experiences (e.g., success, frustration, fears, nightmares). Social memories centered on activities of a social group such as the family, neighborhood, or school. Memory content was categorized based on the central focus rather than contextual background of the memory (Wang & Ross, 2005, p. 598). We provided the following complete instructions to the raters: You will code each individual autobiographical memory (AM) for either “relational” or “individual” focus. A relationally focused AM is one in which the central focus is about an interpersonal interaction or relations with others. For example, an AM with a relational focus could be about time spent with friends or family, a neighborhood event, or a conflict with another person. An individually focused AM is one in which the central focus is about the self and how the self may be related to objects, the environment, or

GeroPsych (2020), 33(1), 3–14

A. J. Polsinelli et al., Emotional Autobiographical Memory

events. For example, an AM with an individual focus could be about a personal achievement or failure, learning a new skill, or running errands. There are many different parts of memories, including the general background or theme in which the event occurred. We’ll call this the “context.” For example, if the memory took place at a party, we might call that the “context.” Another part of the memory is what the person chooses to spend their time telling you about. We’ll call this the “central focus.” For example, speaking about the different people the individual met in the context of the party. Although the context and the central focus are likely to match in most cases, they don’t always. For example, an individual could tell you that they went to a party (context) and focus on telling you all about one individual who was particularly interesting (central focus). In this case, the context and the central focus match. Or, they could tell you they were at a party (context) and focus on telling you all about the delicious food they had there (central focus), barely mentioning the people. In this case the context and the central focus don’t match. The former would be an example of a relationally focused AM because the central focus, namely, what they chose to spend their time telling you about, was about the person they met. The latter would be an example of an individually focused AM because the central focus was on their experience of the food rather than on the people around them.

Ó 2020 Hogrefe

Full-Length Research Report

Interindividual Differences in Cognitive Functioning Are Associated with Autobiographical Memory Retrieval Specificity in Older Adults Sarah Peters and Signy Sheldon Department of Psychology, McGill University, Montreal, QC, Canada

Abstract: We examined whether interindividual differences in cognitive functioning among older adults are related to episodic memory engagement during autobiographical memory retrieval. Older adults (n = 49, 24 males; mean age = 69.93; mean education = 15.45) with different levels of cognitive functioning, estimated using the Montreal Cognitive Assessment (MoCA), retrieved multiple memories (generation task) and the details of a single memory (elaboration task) to cues representing thematic or event-specific autobiographical knowledge. We found that the MoCA score positively predicted the proportion of specific memories for generation and episodic details for elaboration, but only to cues that represented event-specific information. The results demonstrate that individuals with healthy, but not unhealthy, cognitive status can leverage contextual support from retrieval cues to improve autobiographical specificity. Keywords: Autobiographical memory, episodic memory, fluency, recollection, interindividual differences

Introduction As individuals age, autobiographical memory – the ability to recollect one’s own personal history – changes over time (Levine, Svoboda, Hay, Winocur, & Moscovitch, 2002; Nyberg, Bäckman, Erngrund, Olofsson, & Nilsson, 1996; Nyberg et al., 2003; Piolino, Desgranges, Benali, & Eustache, 2002). Research indicates that cognitive aging disproportionately targets the episodic component of autobiographical memory, leaving semantic processes intact. Given that episodic processes support the specificity of autobiographical retrieval, i.e., the recollection of contextualized events and their associated details, one might expect retrieval specificity deficits in older adults. However, within this population, there are substantial differences in global cognitive functioning that can influence how memory tasks are performed (Dennis, Bowman, & Peterson, 2014; McIntyre & Craik, 1987; Spencer & Raz, 1995). Indeed, studies have labelled some older adults as “low cognitive performers” and some as “high cognitive performers” based on whether they perform encoding and retrieval tasks at a level comparable to their younger adult Ó 2020 Hogrefe

counterparts (Cabeza et al., 2018; Glisky, 2007; Lighthall, Huettel, & Cabeza, 2014; Van Petten, 2004; Van Petten et al., 2004). While previous research has contrasted autobiographical retrieval specificity between healthy older adults and those with pathological cognitive deficits (e.g., amnestic mild cognitive impairment [aMCI] or Alzheimer’s disease; Addis & Tippett, 2004; Barnabe, Whitehead, Pilon, Arsenault-Lapierre, & Chertkow, 2012; Donix et al., 2009; Leyhe, Müller, Milian, Eschweiler, & Saur, 2009; Matuszewski et al., 2009; Murphy, Troyer, Levine, & Moscovitch, 2008), it remains unclear whether differences in cognitive functioning among healthy older adults is associated with autobiographical retrieval specificity. The current study assesses whether interindividual differences in cognitive functioning within an older adult sample relates to the specificity of autobiographical memory retrieval, a metric of episodic memory engagement. Ultimately, we aimed to understand whether differences in autobiographical specificity can distinguish between healthy and unhealthy aging trajectories in preclinical older adults. Autobiographical memory retrieval requires accessing event-related information from different levels within a GeroPsych (2020), 33(1), 15–29 https://doi.org/10.1024/1662-9647/a000219

knowledge base or hierarchical structure. This knowledge base is organized such that specific, contextualized event information is embedded within broader conceptual or semantic elements of an experience (Conway, 2005; Conway & Pleydell-Pearce, 2000). Age differences in episodic memory influence the ability to access and associate together the specific, perceptually based elements of a past event when constructing an online representation of a recollected experience (“elaboration” retrieval; Levine et al., 2002; Peters, Fan, & Sheldon, 2019; Piolino et al., 2010; Sheldon, McAndrews, & Moscovitch, 2011; Sheldon et al., 2015). A number of studies demonstrated that, compared to younger adults, older adults provide fewer specific episodic details (“I was wearing a blue scarf. It was raining that day”) when describing past events, instead incorporating more personal or general semantic aspects of the experience (“I love travelling,” “Paris is the capital of France”; Addis, Roberts, & Schacter, 2011; Levine et al., 2002; Sheldon et al., 2011; although see Aizpurua & Koutstaal, 2015). Interestingly, age differences in episodic memory ability may also influence the ability to access and associate together multiple specific (i.e., episodic) event representations within the autobiographical knowledge structure (“generation” retrieval). Experiments using autobiographical fluency measures found that, when presented with a memory cue, healthy older adults tend to retrieve significantly fewer specific episodic events (“Visiting the Eiffel Tower Last Summer”) and more repeated/extended events (“Travelling in my 20s,” “Travelling to France”) than younger adults (Peters et al., 2019; Piolino et al., 2002). Differences in autobiographical specificity among older adults at the level of events have been interpreted in different ways. One line of work linked this deficit to age-related episodic memory deficits (e.g., Peters et al., 2019), a proposal supported by neuroimaging research (Sheldon, McAndrews, Pruessner, & Moscovitch, 2016; Sheldon & Moscovitch, 2012). Another body of work suggested different mechanisms might underlie the ability to access specific autobiographical information at the level of event versus detail (Kyung, Yanes-Lukin, & Roberts, 2016; Piolino et al., 2010; Roberts, Yanes-Lukin, & Kyung, 2018). Thus, there remains open questions as to whether cognitive functioning in older adults relates to the specificity of autobiographical memory when generating episodic events versus elaborating on the details of a single recollected experience. When studying the association between cognitive functioning and autobiographical specificity, it is important to consider the cue used to trigger retrieval as memory cues can direct access to different levels of the autobiographical knowledge structure. For instance, a memory cue can direct access via a higher level, when general thematic information is activated by the cue or, from a lower level, when GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

event-specific knowledge is activated (Sheldon & Chu, 2017). In terms of cognitive status, empirical evidence suggests that older adults with superior cognitive ability can flexibly leverage information contained within an environmental cue (e.g., a context) to more effectively approach a given task (Craik, Klix, & Hagendorf, 1986; Craik & McDowd, 1987; Craik & Schloerscheidt, 2011). By comparison, older adults with lower cognitive functioning may not be able to engage in such flexible behavior and are, therefore, unable to benefit from the support provided by external cues. These effects have been demonstrated both when older adults encode new information (Dando, 2013; Verhaeghen, Marcoen, & Goossens, 1992) and when they retrieve laboratory-based stimuli (Craik & Byrd, 1982). However, it is not yet known whether healthy older adults can effectively leverage external support to compensate for episodic memory decline in the context of naturalistic autobiographical memory retrieval. Based on the above-described model of autobiographical memory organization, we propose that, when a retrieval cue directs access to event specific knowledge (i.e., contextual or perceptual elements of an experience), the cue can be leveraged to activate episodic information about the recollected event, improving the specificity of autobiographical memory. In healthy older adults, we propose that high but not low cognitive performers are able to benefit from the external support provided by this type of retrieval cue. In the present study, we restricted our focus to two types of memory cues that direct access to different levels in the structure of autobiographical memory knowledge. Openended cues are those that trigger retrieval via higher-order semantic information and are thought to represent a wide variety of conceptually related experiences. For example, the cue “celebrating” can trigger the reactivation of a diverse array of past events (e.g., a birthday party, dancing in the kitchen, a drink with a friend) that are, nonetheless, all related to the concept of “celebrating.” Conversely, when a retrieval cue is closed-ended, it triggers retrieval via the more episodically specific information shared among the associated events (e.g., an action or environmental context). For example, the retrieval cue “restaurant” reactivates past experiences situated within this specific environmental context. Given that autobiographical retrieval is a complex task, it is important to examine how (or if) cue type influences the relationship between autobiographical specificity and cognitive functioning across different forms of retrieval. We propose that, during event generation, autobiographical memory retrieval cued by open-ended cues will benefit from semantic processing to guide access to specific episodic information via the abstracted, conceptual links between events. By comparison, when the generation of events is guided by episodically specific information Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

(i.e., closed-ended cues), we expect autobiographical specificity in older adults with low cognitive functioning will suffer because this task disproportionately relies on episodic memory processes. By comparison, during elaboration retrieval, open-ended cues can leverage higher-order conceptual information to access and bind together the specific, episodic details of a recollected event – akin to “top-down” processing. This may not be the case for closed-ended cues, which necessarily require episodic processes to access and bind together episodic details in a more “bottom-up” fashion, as guided by the specific contextual or perceptual information represented in the cue. This framework makes specific predictions about where interindividual differences in cognitive functioning are most strongly related to the specificity of autobiographical memory, namely, during elaboration retrieval to closed-ended cues, where episodic demands are highest. Clarifying the nature of the relationship between interindividual differences in cognitive functioning and the specificity of autobiographical memory has the two-fold benefit of providing a more comprehensive understanding of age differences in autobiographical memory retrieval and in identifying a behavioral marker of poor or unhealthy cognitive status.

Current Study The present study investigates how interindividual differences in cognitive functioning among older adults relates to the specificity of autobiographical memory during two types of autobiographical memory retrieval. To this end, we recruited a sample of healthy older adults with subclinical differences in cognitive functioning, as estimated by the Montreal Cognitive Assessment (MoCA; Nasreddine et al., 2005). We used linear mixed-effects modeling to examine the relationship between cognitive functioning and autobiographical specificity during a generation retrieval task, where participants retrieved multiple related events; and an elaboration retrieval task, where participants described a single recollected event in detail. Importantly, we asked whether this relationship was modulated by the nature of the retrieval cue by presenting participants with both closed-ended and open-ended retrieval cues. First, we predicted that, if interindividual differences in global cognitive functioning capture episodic memory ability, these differences will be related to measures of autobiographical retrieval specificity. Second, we predicted that cognitive functioning would be associated with retrieval specificity both when accessing related autobiographical event representations (generation) and when integrating the details of a single event representation (elaboration). Third, we predicted that, across retrieval tasks, high cognitive performance would be associated with improved specificity of autobiographical memory when cued with event-specific Ó 2020 Hogrefe

information (close-ended cues) but not when cued with high-order semantic memory information (open-ended cues).

Materials and Method Participants Forty-nine older adults were recruited from the Montreal area via flyers distributed in the community and via advertisements in local newspapers. Our planned sample size (n = 50) was based on prior related research investigating interindividual (Baudouin, Vanneste, Isingrini, & Pouthas, 2006; Craik, Eftekhari, Bialystok, & Anderson, 2018) and group differences (D’Angelo et al., 2016) in cognitive functioning. Although this sample included individuals with a broad range of global cognitive ability, none of the recruited participants had a formal diagnosis of dementia, aMCI, or any other neurological disorders associated with overt cognitive deficits. All participants had normal or corrected-tonormal vision and hearing, spoke English fluently, were free from major medical complaints, and lived independently (i.e., without external community support or assistance). We excluded participants who had a history of concussion or other head trauma as well as those reporting a significant past or current psychiatric history. Crystallized intelligence and vocabulary were estimated using the Shipley Vocabulary Test (Schear & Harrison, 1988). All participants were above the suggested cut-off for healthy aging (a score of 33 out of 40; Mason & Ganzler, 1964). Participants gave informed consent and received monetary compensation for being part of the study. Table 1 displays the average demographic characteristics as well as estimates of intelligence and cognitive functioning for all participants. For a detailed breakdown of demographic and neuropsychological characteristics as a function of global cognitive status, please see Table E1 in the Electronic Supplemental Material (ESM 1).

Stimuli Eight retrieval cues were collected from previously published reports (Peters et al., 2019; Levine et al., 2002; Sheldon & Chu, 2017; see Table 2). Of these eight cues, half were closed-ended in that they direct access to eventspecific knowledge (i.e., contextual or perceptual elements of an experience). In other words, they trigger retrieval of perceptually related events and, as such, are predicted to result in the retrieval of highly similar event representations. The remaining cues were open-ended in that they direct access to higher-order semantic information (i.e., activity themes) shared among the associated events. GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

Table 1. Demographic characteristics and estimates of cognitive functioning for the tested cohort (n = 49; 24 males) Mean

Median

Range

Age (years)

69.93

4.98

59–86 10–30

Education (years)

15.45

3.93

Shipley (/40)

34.63

3.63

33–40

MoCA Total (/30)

25.82

2.86

16–30 0–5

MoCA Memory (/5)

3.35

1.52

MoCA Executive (/5)

3.82

1.09

1–5

4.89

16.5

6–28

MoCA Fluency

14.6

Table 2. Retrieval cues used in the current study Open-ended cues

Closed-ended cues

Time when you were Celebrating

Times when you were at Home

Times when you were at Galleries

Times when you were at the Mall Times when you were at the Office Times when you were Eating

Times when you were Travelling Times when you made Accomplishments

Such cues are predicted to trigger the retrieval of a wide variety of conceptually related experiences. Cue categorization was based on ratings provided by 50 online older adult participants recruited from Amazon’s Mechanical Turk (MTurk). MTurk participants were shown each cue in sequence and asked to judge the number of unique events that could be represented by that cue (Likert scale: 1 [few events; similar to one another] to 4 [many events; different from one another]). After removing data from three raters who provided incomplete or invalid responses, we were left with a final sample of 47 MTurk participants (22 males, age: M = 63.1 years, SD = 6.08; education: M = 15.3 years, SD = 2.75), and from these data we calculated the average rating for each cue. Using a median-split, cues that received high ratings (i.e., associated with many distinct events) were classified as “open-ended” and those that received low ratings (i.e., associated events are similar/ overlapping) were classified as “closed-ended.” As expected, the average MTurk rating was significantly greater for the open-ended (M = 2.97, SD = 0.47) than for the closed-ended cues (M = 2.04, SD = 0.64; t(46) = 9.54, p < .001, Cohen’s d = 1.39).

Experimental Procedure In an initial screening session, approximately 1 h in length, participants completed a demographics questionnaire, a language screening questionnaire, the Shipley Vocabulary test, and the MoCA cognitive screener. The latter is

GeroPsych (2020), 33(1), 15–29

comprised of 30 items that provide an estimate of executive functioning, memory, language, abstraction, attention, and orientation. The sum of total responses on these items provides a metric of global cognitive ability. The MoCA memory subscale consists of a standard wordlist learning task with 1-point awarded to each correctly recalled word after a short ( 5 minute) delay (maximum of 5 points can be awarded). The MoCA executive function subscale consists of a set-switching task (Trails B) and two visuoconstructive tasks (cube-copy and clock-drawing tests), and performance on these tasks is summed (to a maximum of 5 points) to generate an estimate of executive functioning performance. The MoCA “F” phonemic fluency task is folded into the language subscale but is considered separately in the present study. In this task, participants have 60 s to retrieve as many words as possible beginning with the letter “F.” Language, abstraction, attention, and orientation subscales were not considered independently in the present study. For details we refer readers to Nasreddine and colleagues (2005). Participants completed the experimental task in a single session, 1–5 days after the initial screening session. The experiment was presented via Eprime software (Version 2.0; Psychology Software Tools, Pittsburgh, PA). Detailed instructions were presented to the participants both visually and orally, and all participants completed two practice trials, identical in structure and timing to the experimental trials before beginning the experiment. Participants received detailed feedback on their performance and proceeded to the experimental trials only if it was clear they had understood and could comply with task instructions. Generation Task Over a series of eight trials, participants were presented with open-ended and closed-ended cues on the computer screen and were given 90 s to generate as many specific past personal events as possible to each cue (Dritschel, Williams, Baddeley, & Nimmo-Smith, 1992). Participants were told that a specific event was one that took place in a particular location, lasting minutes or hours but not longer than a single day (e.g., “Going to the park for my 10th birthday”). These instructions remained on the screen throughout the generation task. Participants provided a short verbal statement to describe each retrieved event, and all responses were both audio-recorded and written down by the experimenter. If the participant stopped producing responses during the 90-s fluency period, they were given a general prompt (“Can you think of another memory?”). If they began generating nonspecific responses, they were given a specificity prompt (“Think of a more specific memory”). For scoring purposes, all generation task responses were later transcribed verbatim from the audio recordings.

Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

Elaboration Task After the 90-s fluency period had elapsed, participants were shown a list of their previously generated events (written down by the experimenter) and asked to select one event per cue to elaborate upon in detail. Importantly, participants were instructed to select a specific event that they could remember clearly. If participants did not generate any specific event to a given cue, they were instructed to use one of their general responses to generate a specific event to subsequently elaborate upon. For instance, if they chose “Going to the beach in my teenage years,” they were instructed to “Bring to mind a specific instance or example of a time that you went to the beach as a teenager.” Once an event was retrieved, it was shared with the experimenter for verification before proceeding. In short, all events included in the elaboration task were specific in time and place. Once a specific event had been selected, participants were given 3 min to describe the memory in as much detail as possible. Instructions remained on the screen throughout the elaboration task. If participants began describing a different event or general/factual information, a single prompt was given (“Try to describe only the details of the chosen event, and be as specific and detailed as possible”). Memory descriptions were audio recorded and later transcribed verbatim for scoring. For each detailed event description, participants provided ratings of vividness (0 = not at all vivid to 100 = extremely vivid), familiarity (0 = not at all familiar to 100 = extremely familiar), and estimates of when the event occurred (1 = past week, 2 = past year, 3 = 1–5 years, 4 = 5– 10 years, 5 = > 10 years old, 6 = I don’t know).

Scoring Generation Task The transcribed responses were categorized as either specific or nonspecific events using the Autobiographical Memory Test scoring (AMI) procedure (Williams & Broadbent, 1986). According to this procedure, specific responses capture events that occurred in a particular place and within a defined time period (minutes or hours but less than one day, e.g., “bowling with my niece last June”). Nonspecific responses are those that describe an event lasting longer than 1 day (extended event. e.g., “my trip to Paris”), multiple events occurring in the same location (repeated event, e.g., “going to the gym every Saturday”), general semantic information (e.g., “I am generally a happy person”), or repetitions of responses generated previously to the same cue. Coders were trained on an independent set of data (responses to the practice trials) and scored events as outlined in the AMI procedure. Two independent coders blind to the experimental design scored all generation task data. Given the categorical nature of these data, interrater reliability was assessed by Ó 2020 Hogrefe

calculating Cohen’s κ, which indicated greater than substantial agreement between our raters (κ = 0.86). Following coding and assessment of reliability, extended, categoric, semantic, and repetition responses were collapsed into one “nonspecific” response category. Raw data for the generation task, indicating the average number of responses generated per cue in each AMI coding category, are reported in Table E2 (ESM 1). Elaboration Task The transcribed descriptions were scored according to the Autobiographical Interview (AI) scoring procedure (Levine et al., 2002), which involves segmenting the descriptions into distinct units of information (often a grammatical clause) that independently convey information. Each unit is then coded according to the nature of information it conveys (e.g., occurrence, person, perceptual detail, fact, statement, thought, emotion). Units were then collapsed into two broad categories, internal (episodic) and external (nonepisodic) details. Internal details are those describing specific information pertaining to the main event being recalled (e.g., who was there, perceptual, contextual, and emotional elements) and measure episodic memory processing. External details describe semantic knowledge or general facts (including personal semantics or facts/knowledge about the self), tangential event information (i.e., specific information relating to a different event), or metacognitive statements, and they capture both semantic processing and task adherence. Three blind coders were trained on an independent set of memory descriptions (provided by the developers of the AI) using the procedure laid out in the AI scoring manual. Because of the time-consuming nature of the task, two of the coders scored distinct sets of elaboration descriptions, the third coder rescored a random selection of descriptions (n = 20), which were then used for reliability estimates. All detail categories were collapsed into “internal” and “external” details, and reliability was assessed by calculating Cohen’s κ for these categories across the selected descriptions, which indicated near perfect agreement between raters (κ > 0.90 for both internal and external details). Raw data for the elaboration retrieval task, indicating average number of details generated per cue in each AI coding category, are reported in full in Table E3 (ESM 1).

Analyses To control for individual differences in verbal fluency (output), which is influenced by factors like mental processing speed and executive functioning, we assessed autobiographical specificity by calculating the proportion of specific-to-total responses and internal-to-total details, which were subsequently used as out primary outcome variables. GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

These data were analyzed using linear mixed-effects modelling. Compared to the standard ANOVA, which looks at group-level effects, mixed-effects models offer the opportunity to treat global cognitive functioning (MoCA scores) as a continuous variable, which aligns well with our experimental questions. All models were structured as follows:

Y i ¼ μZ i þ βXi þ εi ; where Yi denotes a vector containing the values of the predictand or dependent variable (autobiographical specificity measures) for the ith participant, μ a vector of q random-effects estimates, Zi a matrix of q randomeffect predictors for the ith participant, β a vector of p fixed-effects beta weight estimates for each predictor included in the model, Xi a matrix of p predictors or independent variables for the ith participant, and ei the model fit error, capturing the discrepancy between the prediction made by the model for each observation from the ith participant and the measured value. Two models were constructed for the generation retrieval task. In the first, the predictor variables were MoCA total score and cue type along with their interaction. In the second, the predictor variables were MoCA Memory score, MoCA Executive Functioning score, MoCA “F” fluency task score, and cue type, along with the two-way interaction between each MoCA scale and cue type. In both models, the predictand was the ratio of specific-to-total memories generated. The MoCA “F” fluency task score was included as a predictor variable to estimate the contribution of strategic search, inhibition, and cognitive control processes. We felt it was important to include estimates of these nonepisodic processes as they are theoretically related to our outcome measure (Alvarez & Emory, 2006), which is more generally akin to an autobiographical fluency task and to task adherence (Ford, Rubin, & Giovanello, 2014). Similarly, two models were constructed for the elaboration retrieval task. In the first, the predictor variables were MoCA total score and cue type along with their interaction. In the second, the predictor variables were MoCA Memory score, MoCA Executive Functioning score, and cue type, along with the two-way interaction between each MoCA scale and cue type. In both models, the predictand was the ratio of internal-to-total details described. Three additional models were constructed to examine the association between global cognitive functioning and the subjective experience of memory recall, as provided by elaboration retrieval task ratings. The fixed effects of the predictors MoCA total score and cue type along with their interaction were modelled. The predictands were vividness, familiarity, and date ratings for each of the three models, respectively. For all mixed model analyses reported, participant and cue type were modelled as random effects predictors to account for GeroPsych (2020), 33(1), 15–29

idiosyncratic variance due to individual differences in task performance and cue variability, respectively. For all models, the regression coefficients and the p-values used to establish statistical significance were based on Satterthwaite approximations for denominator degrees of freedom, established using the “lme” test performed in jamovi (version 0.9.5.12; The jamovi project, 2019). Where appropriate, we confirmed our findings at the group level. We split our older adult sample into two clinically distinct samples using the established cutoff score for the MoCA of 26. This cutoff has been shown to have excellent specificity and sensitivity for differentiating between individuals with marginal or impaired cognitive health (e.g., mild cognitive impairment [MCI]) and healthy controls (Nasreddine et al., 2005) and has practical significance since this same cutoff is often applied in clinical settings to screen for individuals with cognitive impairment. Applying this cutoff to our sample of older adults yielded two groups, one characterized by high cognitive performance (n = 26, 13 males; age: M = 69.20, SD = 4.79; education: M = 14.80, SD = 2.81; total MoCA: M = 28.00, SD = 1.22) and a second by low cognitive performance (n = 23, 11 males; age: M = 70.80, SD = 5.23; education: M = 16.10, SD = 4.95; total MoCA: M = 23.40, SD = 2.23). See Table E1 (ESM 1) for a complete reporting of the demographic characteristics and estimates of cognitive functioning for these two groups. We ran separate mixed-design ANOVAs on the ratio of specific-to-total responses and ratio of internal-to-total details, for generation and elaboration retrieval, respectively, with cue type (open-ended vs. closed-ended) as a within-subjects factor and group (high vs. low cognitive performers) as a between-subjects factor. Posthoc comparisons were made, where indicated, using Tukey’s HSD. We report effect sizes and their corresponding confidence intervals for all findings. For ANOVA main effects and interaction effects, we report eta square and 90% confidence intervals, and for posthoc comparisons we report Cohen’s d and 95% confidence intervals (see Steiger, 2004, for a detailed discussion of this topic). ANOVAs and the corresponding effect sizes were calculated in jamovi (version 0.9.5.12; The jamovi project, 2019). Confidence intervals for eta squared and Cohen’s d were calculated using the ci.pvaf and ci.smd functions, respectively, in the MBESS package (version 4.6.0; Kelley, 2019) of R Studio (version 1.1.453; R Core Team, 2018).

Results Generation Task A linear mixed model was performed to test the effect of global cognitive functioning and cue type on autobiographical Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

specificity during generation retrieval. The ratio of specificto-total generated events was included as the dependent variable with cue type (open-ended or closed-ended), MoCA score, and their interaction as fixed factors in the model. We modelled the random effects of subject and cue. The fixed effect omnibus tests revealed a main effect of MoCA scores (F(1,46.93) = 7.59, p = .008), where older adults with high MoCA scores generated a greater ratio of specific-to-total events than did older adults with low MoCA scores (ß = .02, SE = .007, t(46.93) = 2.75, p = .008; Figure 1). We did find neither a main effect of cue type (F(1,6) = 0.26 p = .63) nor an interaction between these factors (F(1,333.21) = 0.12, p = .73). Variability from the random factors of subject and cue was SD = 0.07 and SD = 0.14, respectively. To confirm these results, we split our sample into two groups (high vs. low cognitive performers) using the clinically derived cutoff of 26 from the MOCA screener. A mixed-design ANOVA was run on the average ratio of specific-to-total responses generated to each cue with type with group (high vs. low cognitive performers) as a between-subjects factor and cue type (open-ended vs. closed-ended) as a within-subjects factor. Results from this analysis revealed a main effect of cue type (higher level of specificity for openended vs. closed-ended; F(1,47) = 4.24, p = .04, η2 = 0.030, 90% CI [.00; .22]) and group (higher level of specificity for high vs. low cognitive performers; F(1,47) = 8.24, p = .006, η2 = 0.095, 90% CI [.03; .29]) but did not find

Figure 1. A visualization of the relationship between global cognitive functioning and the specificity of autobiographical memory during the generation retrieval task. The average ratio of specific-to-total events generated by older adult participants to open-ended and closedended retrieval cues during 90-s retrieval period is associated with performance on the Montreal Cognitive Assessment (MoCA). MoCA scores are mean-centered and error bands represent standard error of the mean. Ó 2020 Hogrefe

a two-way interaction between these terms (F(1,47) = 0.01, p = .918, η2 = 0.000, 90% CI [.00; .02]). A second linear mixed-effects model was performed, as outlined above, but with the ratio of specific-to-total generated events as the dependent variable and with cue type, MoCA Memory, MoCA Executive, and MoCA “F” fluency task scores as fixed factors in the model. The fixed effects omnibus tests revealed a significant main effect of MoCA Memory scores (F(1,43.73) = 7.39, p = .009), viz. older adults with high MoCA Memory scores generated a greater ratio of specific-to-total responses than those with low MoCA Memory scores (ß = .04. SE = .01, t(79.18) = 2.72, p = .009). We did not find significant fixed effects of MoCA Executive score (F(1,84.01) = 3.16, p = .08), cue type (F(1,6) = .26, p = .63), MoCA “F” fluency score (F(1,72.70) = 3.49, p = .07), nor did the interactions between cue type and MoCA Memory score (F(1,343.93) = 0.46, p = .50), cue type and MoCA Executive score (F(1,343.69) = .02, p = .92) or cue type and MoCA “F” fluency score (F(1,337.20) = 0.12, p = .73) result in significant fixed effects. Variability from the random factors of subject and cue was SD = 0.04 and SD = 0.14, respectively.

Elaboration Task As mentioned above, a linear mixed model was performed to test the effect of global cognitive functioning and cue type on autobiographical specificity during elaboration retrieval. The ratio of internal-to-total details was included as the dependent variable with cue type, MoCA score, and their interaction term included as fixed effects. We also modelled the random intercepts for both cue and subject. The fixed effect omnibus tests failed to find a main effect of MoCA scores (F(1,47.33) = 2.18, p = .15) or cue type (F(1,6) = .75, p = .42), but we did find a significant interaction between MoCA score and cue type (F(1,332) = 4.02, p = .04). Unpacking this interaction, we found that older adults with high MoCA scores described a greater ratio of internal-to-total details than did older adults with low MoCA scores – but only for closed-ended cues (ß = -.007, SE = .004, t(33.62) = -2.01, p = .04). There was no relationship between MoCA score and autobiographical specificity for open-ended cues (Figure 2). Variability from the random factors of subject and cue was SD = 0.11 and SD = 0.013, respectively. We confirmed these results at a group level by comparing the episodic richness of event elaborations between high and low cognitive performers. To do so, we ran a mixeddesign ANOVA on the average ratio of internal-to-total details generated to each cue with group (high vs. low cognitive performers) as a between-subjects factor and cue type (open-ended vs. closed-ended) as a within-subjects factor. We did not find a main effect of cue type (F(1,47) = 0.04, p = .849, η2 = 0.000, 90% CI [.00; .04]) or group (F(1,47) = 1.99, p = .166, η2 = 0.032, 90% CI [.00; .16]), but GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

retrieval to determine whether interindividual differences in cognitive functioning relate to the subjective experience of memory recall. We constructed separate models with each of the subjective ratings as the dependent variable (vividness, familiarity, date) and with MoCA score and cue type as predictors. None of these models revealed any association between subjective ratings, main effects of MoCA score, or cue type nor any interaction effects (see Tables 3 and 4).

Discussion

Figure 2. A visualization of the interaction between cue type and global cognitive functioning on the specificity of autobiographical memory during the elaboration retrieval task. The average ratio of internal-to-total details generated by older adult participants during a 3-min retrieval period is associated with performance on the MoCA for closed-ended but not open-ended retrieval cues. MoCA scores are mean-centered and error bands represent standard error of the mean.

we did find a significant interaction between these terms (F(1,47) = 4.23, p = .046, η2 = 0.088, 90% CI [.00; .22]). Tukey’s posthoc comparisons revealed that high cognitive performers generated more episodically rich event elaborations than low cognitive performers for closedended cues (t(47) = 3.26, p = .012, Cohen’s d = 1.02, 95% CI [-0.54; 0.58]) – but not for open-ended cues, where performance was comparable (t(47) = 0.17, p = .998, Cohen’s d = 0.072, 95% CI [0.34; 1.52]). We ran a second linear mixed-effects model with MoCA Memory and Executive subscores as predictors and the ratio of internal-to-total details as the dependent variable (Figure 3). This analysis revealed no main effect of cue type (F(1,6.03) = .76, p = .42), MoCA Memory score (F(1,46.12) = .09, p = .76), or MoCA Executive score (MoCA Executive, (F(1,46.01) = 1.73, p = .19). We also failed to find an interaction between cue type and MoCA Executive scores (F(1,331.25) = .006, p = .94). However, we did find a significant interaction effect between cue type and MoCA Memory score (F(1,331.39) = 5.11 p = .02), viz. older adults with high MoCA Memory scores generated a greater ratio of internal-to-total details than those with low MoCA memory scores for events described to closed-ended but not to open-ended retrieval cues (ß = -.02, SE = .007, t(331.39) = -2 .26, p = .02). Variability from the random factors of subject and cue was SD = 0.11 and SD = 0.013, respectively. Finally, we examined the relationship between the MoCA scores and the subjective ratings of memory elaboration GeroPsych (2020), 33(1), 15–29

The ability to retrieve specific information about past personal experiences or the specificity of autobiographical memory depends on episodic memory processes (Eichenbaum, Yonelinas, & Ranganath, 2007; Naveh-Benjamin, Hussain, Guez, & Bar-On, 2003; Nyberg et al., 1996; Olsen, Moses, Riggs, & Ryan, 2012; Tulving, 2002). Healthy cognitive status is associated with impairments in episodic memory ability which influence the specificity of autobiographical memory (e.g., Peters et al., 2019). However, age differences in episodic memory are far from homogeneous, and the association between subclinical episodic memory impairments and autobiographical specificity is not yet known. To address this, we asked whether interindividual differences in cognitive functioning in healthy older adults are associated with the ability to engage episodic processes during autobiographical memory retrieval. To this end, we established cognitive functioning in a sample of older adult using the MoCA cognitive assessment tool (Nasreddine et al., 2005) and then measured the specificity of autobiographical memory across two retrieval tasks. The first was an autobiographical fluency task, where participants retrieved multiple specific autobiographical events (generation task), and the second was a memory description task in which participants constructed a detailed account of a single recollected episode (elaboration task). We used established scoring systems to quantify autobiographical retrieval specificity. During the generation task, we calculated the proportion of specific (contextualized) events generated by participants, and during the elaboration task we calculated the proportion of internal (contextual-perceptual) details included in memory descriptions. Prior research showed that episodic memory processes are required to retrieve both specific events and their embedded details from the autobiographical memory knowledge structure (Bryan & Luszcz, 2000; Devitt, Addis, & Schacter, 2017; Holland & Rabbitt, 1990; Levine et al., 2002; Madore, Gaesser, & Schacter, 2014; Piolino et al., 2002, 2006, 2010; Sheldon et al., 2011), making autobiographical specificity a useful metric of episodic memory engagement during retrieval. Within each retrieval task, we examined Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

(A)

(B)

Figure 3. A visualization of the relationship between cognitive domain scores on the MoCA and the specificity of autobiographical memory during the elaboration retrieval task when triggered by open-ended and closed-ended retrieval cues. (A) The average ratio of internal-to-total details generated by older adults during a 3-min retrieval period is associated with performance on the MoCA Memory subscale for closed-ended but not open-ended retrieval. (B) The average ratio of internal-to-total details generated by older adults was not significantly related to MoCA Executive subscale scores nor was there an interaction with cue type. MoCA subscale scores are mean-centered and error bands represent standard error of the mean.

Table 3. The fixed effects from three linear mixed-effects models examining the relationship between global cognitive performance, retrieval cue type and the subjective experience of recollection, as reflected in ratings collected during the elaboration retrieval task Dependent variable

Fixed effects predictor

Familiarity

Vividness

Date

MoCA score Cue type

0.38 0.0604

1, 46.6 1, 327.5

0.54 0.81

MoCA score Cue type

0.4737

1, 327.4

0.49

MoCA score Cue type

0.20 0.75

1, 47.53 1, 6.05

0.66 0.42

MoCA score Cue type

0.87

1, 332.43

0.35

MoCA score Cue type

1.37 0.16

1, 47.20 1,6.00

0.25 0.70

MoCA score Cue type

2.40

1, 332.52

0.12

Table 4. The random effects due to subject and cue type from three linear mixed-effects models examining the relationship between global cognitive performance, retrieval cue type, and the subjective experience of recollection, as reflected in ratings collected during the elaboration retrieval task Dependent variable

Random effects predictor

Familiarity

Subject Cue type

20.00 0.00

Vividness

Subject Cue type

1.63 1.52

Date

Subject Cue type

0.59 0.43

how episodic memory engagement is modulated by the nature of the cue used to trigger retrieval. Specifically, we asked how the association between cognitive functioning on autobiographical specificity is modulated when retrieval is directed by higher-order semantic information (openended cues) or more episodically specific information (closed-ended cues). This manipulation emerged from Ó 2020 Hogrefe

prior work suggesting that older adults with high cognitive functioning can leverage support from memory cues to more efficiently complete a retrieval task (Craik et al., 1986; Craik & McDowd, 1987; Craik & Schloerscheidt, 2011). Using linear mixed-effects modeling, we examined the relationship between measures of episodic memory engagement (i.e., specific events, internal details) and estimates of cognitive functioning during both generation and elaboration retrieval. In line with our predictions, interindividual differences in global cognitive functioning was associated with impairments in the specificity of autobiographical memory during both the generation and elaboration retrieval tasks. Importantly, we found that the relationship between cognitive functioning and the specificity of autobiographical memory was driven primarily by interindividual differences in episodic memory functioning (MoCA memory scale) and not by executive functioning (MoCA executive and “F” fluency scales). These data are in line with prior research indicating GeroPsych (2020), 33(1), 15–29

that deficits in the specificity of autobiographical memory are amplified in individuals with impaired cognitive functioning, particularly when accompanied by pronounced deficits in episodic memory, such as in older adults diagnosed with Alzheimer’s disease or aMCI (Donix et al., 2009; El Haj, Antoine, Nandrino, & Kapogiannis, 2015; Murphy et al., 2008; Seidl, Lueken, Thomann, Geider, & Schröder, 2011; Sheldon et al., 2015). Although none of the participants we tested had received a clinical diagnosis (e.g., Alzheimer’s disease or aMCI), our healthy adult sample was characterized by a wide range of cognitive functioning, as established using the MoCA cognitive assessment tool. This particular measure was chosen for its excellent sensitivity and selectivity in distinguishing between healthy and pathological aging trajectories (Nasreddine et al., 2005), and recent findings have linked scores on MoCA memory scale to brain regions implicated in episodic memory processing (Ritter, Hawley, Banks, & Miller, 2017). This suggests that approximately half of our healthy older adult sample presented with subclinical global cognitive and episodic memory deficits, in the absence of dementia, which strongly corresponded to impairments in the specificity of autobiographical memory. Importantly, this relationship was evident across multiple forms of autobiographical memory retrieval, indicating that impairments in cognitive functioning – and episodic memory in particular – manifest throughout this complex cognitive task. This aligns with prior work demonstrating that episodic memory processes are required to associate autobiographical knowledge together during both generation and elaboration forms of retrieval (Peters et al., 2019). However, upon examining the association between cognitive functioning and autobiographical specificity in more detail, we found that it differed between the two tested forms of autobiographical retrieval, and this difference was driven by the nature of the retrieval cue. In the present study, cue type was manipulated to represent different types of autobiographical information, which we postulated would influence how episodic events would be accessed from the autobiographical knowledge structure (Conway, 2005; Conway & Pleydell-Pearce, 2000). Open-ended retrieval cues were designed to represent broader conceptual autobiographical information (e.g., “travelling”), which would lead to the retrieval of generalized or semantic event information that is hypothesized to remain relatively intact in older adults (e.g., Levine et al., 2002). Closedended retrieval cues were designed to represent episodically specific information (e.g., actions, contexts), which would lead to the retrieval of event-specific knowledge. Access to such perceptual-contextual knowledge necessarily requires episodic memory processing and, therefore, can potentially discriminate between healthy and unhealthy aging trajectories (e.g., Levine et al., 2002; Tulving, 2002). GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

While we did not find cue effects during the generation task (addressed below), we did find that cue type influenced the relationship between cognitive functioning and the specificity of autobiographical memory during elaboration retrieval, such that the MoCA score positively predicted the episodic richness of event elaborations for closed-ended, but not for open-ended cues. In other words, when retrieval is triggered by a closed-ended cue, older adults with low cognitive functioning (i.e., low MoCA scores) show a relative impairment in accessing specific episodic details when constructing a detailed event representation. We take this finding as evidence that older adults with high cognitive functioning (i.e., high MoCA scores) were able to leverage the event-specific information represented by the closed-ended cues to improve access to specific autobiographical event details, whereas those with low cognitive functioning could not. This interpretation rests on the classic finding that healthy older adults can benefit from environmental factors, including certain types of retrieval cue, to improve their performance on complex retrieval tasks (Craik et al., 1986; Craik & McDowd, 1987; Craik & Schloerscheidt, 2011). This appears to be especially true for naturalistic autobiographical memory tasks, such as that employed in the present study, which already provide minimal external retrieval support (Craik, 1983). Our results build on literature by demonstrating that, similarly, high cognitive performers can leverage the contextual support represented by closed-ended retrieval cues to maintain or even improve autobiographical specificity. If high and low cognitive performance on the MoCA can be considered an indication of healthy and unhealthy cognitive “status,” respectively (Nasreddine et al., 2005), our findings also suggest that, during elaboration retrieval, closed-ended memory cues can distinguish between healthy and unhealthy aging trajectories. The mechanism by which closed-ended cues improved autobiographical specificity for older adults with high cognitive functioning is open to interpretation, particularly given that dimensions other than “endedness” are reflected in our cue manipulation. For instance, closed-ended cues were more likely to represent specific spatial contexts (“mall”), whereas the open-ended cues were more likely to represent activity themes (“travelling”). Prior research showed that spatial contexts improve autobiographical specificity during elaboration retrieval by providing a supportive scaffold to guide the recollection of episodic details (Robin, 2018; Robin, Wynn, & Moscovitch, 2016; Sheldon & Chu, 2017). Consistent with our own findings, healthy older adults benefit from spatial cues during autobiographical retrieval (Robin & Moscovitch, 2017), whereas populations with global cognitive impairment, such as aMCI and Alzheimer’s disease patients, have difficulty constructing a strong spatial scaffold to support retrieval specificity (Serino, Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

Morganti, Di Stefano, & Riva, 2015; Serino & Riva, 2014). Closed-ended cues could also represent more familiar event information, which high-functioning older adults can use to access more rehearsed, and thus more detailed, memories than low-functioning older adults. Another possibility is that the cues also differed in emotional valence, such that the open-ended cues are more positive (“celebrating”) than the close-ended cues, which tend to be more neutral (“offices”). However, based on findings that healthy but not unhealthy cognitive status is associated with a strong positivity bias (Döhnel et al., 2008; Mather & Carstensen, 2005), we would expect higher MoCA scores to be associated with improved performance to open-ended cues compared to closed-ended cues, which was not the case. Finally, it is plausible that high cognitive performers are able to flexibly recruit additional neurocognitive processes, including those supported by the prefrontal cortex, such as cognitive control, to support elaboration retrieval in closed-ended retrieval scenarios (Cabeza, 2002; Duarte, Ranganath, Trujillo, & Knight, 2006; Lighthall et al., 2014). Because we did not collect comprehensive neuropsychological data in the current sample, it is difficult to confidently speak to compensatory recruitment, although in the present study estimates of executive functioning were not associated with autobiographical specificity, which argues against this interpretation. As noted above, cue type did not modulate the association between cognitive functioning and the specificity of autobiographical memory during the generation retrieval task. Instead we found that, across cue type, high cognitive performers tended to generate a greater proportion of specific events compared to low cognitive performers, and that this relationship was driven primarily by episodic memory ability (MoCA memory subscale). This suggests that, counter to our predictions, both open-ended and closed-ended cues may recruit episodic memory to the same extent when generating specific events. This is likely because episodic processes critically support the formation of associations between disparate autobiographical event representations (Eichenbaum, 2003, 2004), which is critical to generation retrieval irrespective of how it is triggered. This fits with previous research demonstrating that episodic processes are recruited whenever one must associate autobiographical information in mind, regardless of whether at the level of event or detail (see Peters et al., 2019, for a detailed discussion on this topic). However, some have suggested that these forms of autobiographical specificity instead depend on separable processes (Dritschel et al., 1992; Kyung et al., 2016; Martinelli et al., 2013; Rathbone, Holmes, Murphy, & Ellis, 2015; Roberts et al., 2018). For example, generation retrieval is akin to an autobiographical memory fluency task, which some propose is more dependent on executive processes than elaboration retrieval (Martinelli Ó 2020 Hogrefe

et al., 2013; Piolino et al., 2010; Roberts et al., 2018). As such, it could be that low cognitive functioning in older adults primarily reflects an executive processing deficit, which would influence performance on a fluency task, regardless of how retrieval was cued. However, in the present study, autobiographical specificity during generation retrieval was not related to either the MoCA executive subscale or, somewhat surprisingly, to performance on the MoCA “F” fluency task, which suggests that indiscriminate executive deficits are, perhaps, not the best explanation for these findings. To more confidently tease apart executive function contributions to autobiographical specificity, future studies could replicate the reported experiment yet include a more in-depth neuropsychological test battery (see Piolino et al., 2010, for related findings). Interestingly, when examining the average number of specific and nonspecific responses generated to each cue (see ESM 1, Table E2), we found that both high and low performing older adults retrieve more specific events to openended compared to closed-ended cues. This suggests that open-ended cues direct a more effective, “top-down” search strategy as guided by the broader conceptual information represented in the cues This strategy may take advantage of the organizational structure of autobiographical memory and broader conceptual links between events (in that specific information can be accessed via the broader concepts in which it is embedded). By comparison, closedended cues appear to direct a less effective, “bottom-up” search strategy as guided by event-specific information resulting in the retrieval of fewer specific events. This suggests that event knowledge is not organized according to the more “ad-hoc” associations represented by closedended cues (contexts or actions), making it more difficult to access multiple related specific events.

Limitations and Alternate Interpretations As with many studies, there are some methodological limitations worth noting. First, despite basing our sample size on prior work exploring interindividual (e.g., Baudouin et al., 2006; Craik et al., 2018) and group differences (e.g, D’Angelo et al., 2016) in autobiographical memory and other complex forms of cognition, the number of participants tested is relatively small for a study of this nature. While most of our results were associated with large effect sizes, our small sample size raises questions about the statistical power of our findings particularly with respect to cue effects during generation retrieval. To this end, we are encouraged to replicate our findings in future work. Another methodological limitation discussed at some length relates to the characterization of the open-ended and closed-ended cues, which framed how we interpreted GeroPsych (2020), 33(1), 15–29

the association between cognitive functioning and autobiographical specificity, particularly in the context of elaboration retrieval. It certainly could be the case that these cues differ on dimensions other than “endedness” (e.g., spatial contexts vs. event-themes or neutral vs. positive emotion), and it will be important for future research to further refine which dimension represents the “active ingredient.” Nonetheless, we demonstrate that some cues, particularly those representing event-specific information (including spatial contexts), are better at distinguishing between healthy and unhealthy aging trajectories and their association with the specificity of autobiographical memory. Several additional factors cannot be completely ruled out using the methodology employed in the present study. It is possible that older adults with low cognitive functioning simply had more difficulty understanding and complying with task instructions, or that their performance was influenced by more global changes in processing speed, particularly for the generation task. Indeed, explicitly instructing individuals to retrieve many “specific” events within a restricted time period requires one to maintain task goals online, inhibit irrelevant or inappropriate responses and engage in strategic search processes, as well as to rapidly retrieve information, all of which tend to be impaired in older adults (Ford et al., 2014). Therefore, it is possible that results from the generation task suggest that high-functioning older adults are simply better able to understand and adhere to task instructions than low-functioning older adults. If this was the case, we might expect that measures of executive functioning, particularly phonemic “F” fluency, which taps into cognitive processes thought to be important for task adherence as well as speed of retrieval, would be associated with interindividual differences in autobiographical specificity during generation retrieval. This was not the case in the present study, where “F” fluency performance did not relate to the specificity of autobiographical memory during generation (or elaboration). Nonetheless, possible differences in task adherence and processing speed should be considered when interpreting our results. Finally, it is possible that low cognitive performing older adults were simply less engaged with their community or had major lifestyle differences, which could influence both global cognitive functioning and/or the availability of autobiographical events at retrieval. While this is certainly a possibility, all recruited participants responded to advertisements posted in the community, suggesting they are, at least somewhat, active and were confirmed to life independently, which implies minimal functional impairments across our older adult sample. In addition, both high and low cognitive performers disproportionately retrieved remote events (i.e., those experienced over 10 years ago), which argues against the notion that lifestyle differences as a function of recent cognitive changes is GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

influencing the availability of prior events. Nonetheless, longstanding lifestyle differences and subtle differences in functioning impairment were not formally assessed, making it difficult to rule them out as possible contributing factors.

Conclusion The reported results indicate that subtle subclinical memory deficits in older adults, as measured with a simple cognitive screening test, are linked to episodic memory impairments during two forms of autobiographical memory retrieval, when retrieving multiple specific past events or recalling the specific details of a single experience. More generally, our results speak to how interindividual differences in cognitive status can influence the ability to access autobiographical information. Since autobiographical memory processes critically support several nonmnemonic tasks important for effective daily functioning (Bluck, 2003; Pillemer, 2003; Prebble, Addis, & Tippett, 2013), our results suggest that individual differences in normal aging can alter several facets of daily life (e.g., problem-solving, selfconcept or identity). Our results mirror patterns observed when comparing healthy older adults to those with aMCI, a pathological syndrome that targets episodic memory (Donix et al., 2009; Murphy et al., 2008; Sheldon et al., 2015). This parallel raises the possibility that autobiographical memory deficits could represent a useful pre-clinical marker of an unhealthy aging trajectory.

Electronic Supplementary Material The electronic supplementary material is available with the online version of the article at https://doi.org/ 10.1024/1662-9647/a000219 ESM 1. The electronic supplemental materials section contains additional data providing the reader with a more detailed characterization of our tested sample of older adults, including a breakdown according to cognitive status to supplement group-level comparisons made in the manuscript. We also include scoring data for both the generation and elaboration retrieval tasks broken down by sub-category to allow readers to more carefully evaluate our scored data.

References Addis, D., Roberts, R., & Schacter, D. (2011). Age-related neural changes in autobiographical remembering and imagining. Neuropsychologia, 49, 3656–3669.

Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

Addis, D., & Tippett, L. (2004). Memory of myself: Autobiographical memory and identity in Alzheimer’s disease. Memory, 12, 56–74. Aizpurua, A., & Koutstaal, W. (2015). A matter of focus: Detailed memory in the intentional autobiographical recall of older and younger adults. Consciousness and Cognition, 33, 145–155. Alvarez, J. A., & Emory, E. (2006). Executive function and the frontal lobes: A meta-analytic review. Neuropsychology Review, 16, 17–42. Barnabe, A., Whitehead, V., Pilon, R., Arsenault-Lapierre, G., & Chertkow, H. (2012). Autobiographical memory in mild cognitive impairment and Alzheimer’s disease: A comparison between the Levine and Kopelman interview methodologies. Hippocampus, 22, 1809–1825. Baudouin, A., Vanneste, S., Isingrini, M., & Pouthas, V. (2006). Differential involvement of internal clock and working memory in the production and reproduction of duration: A study on older adults. Acta Psychologica, 121, 285–296. Bluck, S. (2003). Autobiographical memory: Exploring its functions in everyday life. Memory, 11, 113–123. Bryan, J., & Luszcz, M. A. (2000). Measurement of executive function: Considerations for detecting adult age differences. Journal of Clinical and Experimental Neuropsychology, 22, 40–55. Cabeza, R. (2002). Hemispheric asymmetry reduction in older adults: The HAROLD model. Psychology and Aging, 17, 85–100. Cabeza, R., Albert, M., Belleville, S., Craik, F. I., Duarte, A., Grady, C. L., . . . Reuter-Lorenz, P. A. (2018). Maintenance, reserve and compensation: The cognitive neuroscience of healthy ageing. Nature Reviews Neuroscience, 19, 701–710. Conway, M. A. (2005). Memory and the self. Journal of Memory and Language, 53, 594–628. Conway, M. A., & Pleydell-Pearce, C. W. (2000). The construction of autobiographical memories in the self-memory system. Psychological Review, 107, 261–288. Craik, F. I. M. (1983). On the transfer of information from temporary to permanent memory. Philosophical Transactions of the Royal Society of London. B, Biological Sciences, 302(1110), 341–359. Craik, F. I. M., & Byrd, M. (1982). Aging and cognitive deficits: The role of attentional resources. In F. I. M. Craik & S. E. Trehub (Eds.), Aging and cognitive processes (pp. 191–211). New York, NY: Plenum Press. Craik, F., Eftekhari, E., Bialystok, E., & Anderson, N. D. (2018). Individual differences in executive functions and retrieval efficacy in older adults. Psychology and Aging, 33, 1105–1114. Craik, F., Klix, F., & Hagendorf, H. (1986). A functional account of age differences in memory. New York, NY: Routledge. Craik, F., & McDowd, J. (1987). Age differences in recall and recognition. Journal of Experimental Psychology: Learning, Memory, and Cognition, 13, 474–479. Craik, F., & Schloerscheidt, A. (2011). Age-related differences in recognition memory: Effects of materials and context change. Psychology and Aging, 26, 671–677. D’Angelo, M. C., Smith, V. M., Kacollja, A., Zhang, F., Binns, M. A., Barense, M. D., & Ryan, J. D. (2016). The effectiveness of unitization in mitigating age-related relational learning impairments depends on existing cognitive status. Aging, Neuropsychology, and Cognition, 23, 667–690. Dando, C. J. (2013). Drawing to remember: External support of older adults’ eyewitness performance. PLoS One, 8(7), e69937. Dennis, N. A., Bowman, C. R., & Peterson, K. M. (2014). Agerelated differences in the neural correlates mediating false recollection. Neurobiology of Aging, 35, 395–407. Devitt, A. L., Addis, D. R., & Schacter, D. L. (2017). Episodic and semantic content of memory and imagination: A multilevel analysis. Memory & Cognition, 45, 1–17. Döhnel, K., Sommer, M., Ibach, B., Rothmayr, C., Meinhardt, J., & Hajak, G. (2008). Neural correlates of emotional working

Ó 2020 Hogrefe

memory in patients with mild cognitive impairment. Neuropsychologia, 46, 37–48. Donix, M., Brons, C., Jurjanz, L., Poettrich, K., Winiecki, P., & Holthoff, V. A. (2009). Overgenerality of autobiographical memory in people with amnestic mild cognitive impairment and early Alzheimer’s disease. Archives of Clinical Neuropsychology, 25, 22–27. Dritschel, B. H., Williams, J. M., Baddeley, A. D., & Nimmo-Smith, I. (1992). Autobiographical fluency: A method for the study of personal memory. Memory & Cognition, 20, 133–140. Duarte, A., Ranganath, C., Trujillo, C., & Knight, R. T. (2006). Intact recollection memory in high-performing older adults: ERP and behavioral evidence. Journal of Cognitive Neuroscience, 18, 33– 47. Eichenbaum, H. (2003). How does the hippocampus contribute to memory? Trends in Cognitive Sciences, 7, 427–429. Eichenbaum, H. (2004). Hippocampus: cognitive processes and neural representations that underlie declarative memory. Neuron, 44, 109–120. Eichenbaum, H., Yonelinas, A. P., & Ranganath, C. (2007). The medial temporal lobe and recognition memory. Annual Review of Neuroscience, 30, 123–152. El Haj, M., Antoine, P., Nandrino, J. L., & Kapogiannis, D. (2015). Autobiographical memory decline in Alzheimer’s disease: A theoretical and clinical overview. Ageing Research Reviews, 23, 183–192. Ford, J. H., Rubin, D. C., & Giovanello, K. S. (2014). Effects of task instruction on the specificity of autobiographical memory in young and older adults. Memory, 22, 722–736. Glisky, E. L. (2007). Changes in cognitive function. In D. R. Riddle (Ed.), Human aging in brain aging: Models, methods and mechanisms (pp. 3–20). Boca Raton, FL: CRC Press, Taylor & Francis Group. Holland, C., & Rabbitt, P. (1990). Autobiographical and text recall in the elderly: An investigation of a processing resource deficit. The Quarterly Journal of Experimental Psychology, 42, 441–470. The jamovi project. (2019). jamovi [Computer Software]. Retrieved from https://www.jamovi.org Kelley, K. (2019). MBESS: The MBESS R Package. R package version 4.6.0. https://CRAN.R-project.org/package=MBESS Kyung, Y., Yanes-Lukin, P., & Roberts, J. E. (2016). Specificity and detail in autobiographical memory: Same or different constructs? Memory, 24, 272–284. Levine, B., Svoboda, E., Hay, J. F., Winocur, G., & Moscovitch, M. (2002). Aging and autobiographical memory: Dissociating episodic from semantic retrieval. Psychology and Aging, 17, 677–689. Leyhe, T., Müller, S., Milian, M., Eschweiler, G. W., & Saur, R. (2009). Impairment of episodic and semantic autobiographical memory in patients with mild cognitive impairment and early Alzheimer’s disease. Neuropsychologia, 47, 2464– 2469. Lighthall, N. R., Huettel, S. A., & Cabeza, R. (2014). Functional compensation in the ventromedial prefrontal cortex improves memory-dependent decisions in older adults. Journal of Neuroscience, 34, 15648–15657. Madore, K. P., Gaesser, B., & Schacter, D. L. (2014). Constructive episodic simulation: Dissociable effects of a specificity induction on remembering, imagining, and describing in young and older adults. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 609–622. Martinelli, P., Sperduti, M., Devauchelle, A.-D., Kalenzaga, S., Gallarda, T., Lion, S., . . . Meder, J. F. (2013). Age-related changes in the functional network underlying specific and general autobiographical memory retrieval: A pivotal role for the anterior cingulate cortex. PloS One, 8(12), e82385.

GeroPsych (2020), 33(1), 15–29

Mason, C. F., & Ganzler, H. (1964). Adult norms for the Shipley Institute of Living Scale and Hooper Visual Organization Test based on age and education. Journal of Gerontology, 19, 419–424. Mather, M., & Carstensen, L. L. (2005). Aging and motivated cognition: The positivity effect in attention and memory. Trends in Cognitive Sciences, 9, 496–502. Matuszewski, V., Piolino, P., Belliard, S., de la Sayette, V., Laisney, M., C., Lalevee., & Desgranges, B. (2009). Patterns of autobiographical memory impairment according to disease severity in semantic dementia. Cortex, 45, 456–472. McIntyre, J. S., & Craik, F. I. (1987). Age differences in memory for item and source information. Canadian Journal of Psychology/ Revue Canadienne de Psychologie, 41, 175–192. Murphy, K. J., Troyer, A. K., Levine, B., & Moscovitch, M. (2008). Episodic, but not semantic, autobiographical memory is reduced in amnestic mild cognitive impairment. Neuropsychologia, 46, 3116–3123. Nasreddine, Z. S., Phillips, N. A., Bédirian, V., Charbonneau, S., Whitehead, V., Collin, I., . . . Chertkow, H. (2005). The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. Journal of the American Geriatrics Society, 53, 695–699. Naveh-Benjamin, M., Hussain, Z., Guez, J., & Bar-On, M. (2003). Adult age differences in episodic memory: Further support for an associative-deficit hypothesis. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29, 826–837. Nyberg, L., Bäckman, L., Erngrund, K., Olofsson, U., & Nilsson, L.G. (1996). Age differences in episodic memory, semantic memory, and priming: Relationships to demographic, intellectual, and biological factors. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 51, P234–P240. Nyberg, L., Maitland, S. B., Rönnlund, M., Bäckman, L., Dixon, R. A., Wahlin, Å., & Nilsson, L.-G. (2003). Selective adult age differences in an age-invariant multifactor model of declarative memory. Psychology and Aging, 18, 149–160. Olsen, R., Moses, S., Riggs, L., & Ryan, J. (2012). The hippocampus supports multiple cognitive processes through relational binding and comparison. Frontiers in Human Neuroscience, 6(146), 1–13. Peters, S. L., Fan, C. L., & Sheldon, S. (2019). Episodic memory contributions to autobiographical memory and open-ended problem-solving specificity in younger and older adults. Memory & Cognition, 47, 1–14. Pillemer, D. (2003). Directive functions of autobiographical memory: The guiding power of the specific episode. Memory, 11, 193–202. Piolino, P., Coste, C., Martinelli, P., Macé, A.-L., Quinette, P., Guillery-Girard, B., & Belleville, S. (2010). Reduced specificity of autobiographical memory and aging: Do the executive and feature binding functions of working memory have a role? Neuropsychologia, 48, 429–440. Piolino, P., Desgranges, B., Benali, K., & Eustache, F. (2002). Episodic and semantic remote autobiographical memory in ageing. Memory, 10, 239–257. Piolino, P., Desgranges, B., Clarys, D., Guillery-Girard, B., Taconnat, L., Isingrini, M., & Eustache, F. (2006). Autobiographical memory, autonoetic consciousness, and self-perspective in aging. Psychology and Aging, 21, 510–525. Prebble, S. C., Addis, D. R., & Tippett, L. J. (2013). Autobiographical memory and sense of self. Psychological Bulletin, 139, 815–840. R Core Team. (2013). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. http://www.R-project.org Rathbone, C. J., Holmes, E. A., Murphy, S. E., & Ellis, J. A. (2015). Autobiographical memory and well-being in aging: The central

GeroPsych (2020), 33(1), 15–29

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

role of semantic self-images. Consciousness and Cognition, 33, 422–431. Ritter, A., Hawley, N., Banks, S. J., & Miller, J. B. (2017). The association between montreal cognitive assessment memory scores and hippocampal volume in a neurodegenerative disease sample. Journal of Alzheimer’s Disease, 58, 695–699. Roberts, J. E., Yanes-Lukin, P., & Kyung, Y. (2018). Distinctions between the specificity of autobiographical memory and detail: Trajectories across cue presentations. Consciousness and Cognition, 65, 342–351. Robin, J. (2018). Spatial scaffold effects in event memory and imagination. Wiley Interdisciplinary Reviews: Cognitive Science, 9, e1462. Robin, J., & Moscovitch, M. (2017). Familiar real-world spatial cues provide memory benefits in older and younger adults. Psychology and Aging, 32, 210–219. Robin, J., Wynn, J., & Moscovitch, M. (2016). The spatial scaffold: The effects of spatial context on memory for events. Journal of Experimental Psychology: Learning, Memory, and Cognition, 42, 308–315. Schear, J. M., & Harrison, W. R. (1988). Estimating WAIS IQ from the Shipley institute of living scale: A replication. Journal of Clinical Psychology, 44, 68–71. Seidl, U., Lueken, U., Thomann, P. A., Geider, J., & Schröder, J. (2011). Autobiographical memory deficits in Alzheimer’s disease. Journal of Alzheimer’s Disease, 27, 567–574. Serino, S., Morganti, F., Di Stefano, F., & Riva, G. (2015). Detecting early egocentric and allocentric impairments deficits in Alzheimer’s disease: An experimental study with virtual reality. Frontiers in Aging Neuroscience, 7, 88. Serino, S., & Riva, G. (2014). What is the role of spatial processing in the decline of episodic memory in Alzheimer’s disease? The “mental frame syncing” hypothesis. Frontiers in Aging Neuroscience, 6, 33. Sheldon, S., & Chu, S. (2017). What versus where: Investigating how autobiographical memory retrieval differs when accessed with thematic versus spatial information. The Quarterly Journal of Experimental Psychology, 70, 1909–1921. Sheldon, S., McAndrews, M., & Moscovitch, M. (2011). Episodic memory processes mediated by the medial temporal lobes contribute to open-ended problem solving. Neuropsychologia, 49, 2439–2447. Sheldon, S., McAndrews, M., Pruessner, J., & Moscovitch, M. (2016). Dissociating patterns of anterior and posterior hippocampal activity and connectivity during distinct forms of category fluency. Neuropsychologia, 90, 148–158. Sheldon, S., & Moscovitch, M. (2012). The nature and timecourse of medial temporal lobe contributions to semantic retrieval: An fMRI study on verbal fluency. Hippocampus, 22, 1451–1466. Sheldon, S., Vandermorris, S., Al-Haj, M., Cohen, S., Winocur, G., & Moscovitch, M. (2015). Ill-defined problem solving in amnestic mild cognitive impairment: Linking episodic memory to effective solution generation. Neuropsychologia, 68, 168–175. Spencer, W. D., & Raz, N. (1995). Differential effects of aging on memory for content and context: A meta-analysis. Psychology and Aging, 10, 527–539. Steiger, J. H. (2004). Beyond the F test: effect size confidence intervals and tests of close fit in the analysis of variance and contrast analysis. Psychological Methods, 9, 164–182. Tulving, E. (2002). Episodic memory: from mind to brain. Annual Review of Psychology, 53, 1–25. Van Petten, C. (2004). Relationship between hippocampal volume and memory ability in healthy individuals across the lifespan: Review and meta-analysis. Neuropsychologia, 42, 1394–1413.

Ó 2020 Hogrefe

S. Peters & S. Sheldon, Aging and Autobiographical Specificity

Van Petten, C., Plante, E., Davidson, P. S., Kuo, T. Y., Bajuscak, L., & Glisky, E. L. (2004). Memory and executive function in older adults: Relationships with temporal and prefrontal gray matter volumes and white matter hyperintensities. Neuropsychologia, 42, 1313–1335. Verhaeghen, P., Marcoen, A., & Goossens, L. (1992). Improving memory performance in the aged through mnemonic training: A meta-analytic study. Psychology and Aging, 7, 242–251. Williams, J. M., & Broadbent, K. (1986). Autobiographical memory in suicide attempters. Journal of Abnormal Psychology, 95, 144–149. History Received June 15, 2019 Accepted October 22, 2019

Conflict of Interest The authors declare no conflict of interest. Funding A NSERC Discovery grant (# RGPIN-04241) awarded to Signy Sheldon supported the work reported in this manuscript.

Dr. Signy Sheldon Department of Psychology McGill University 2001 McGill College Avenue Montreal, Quebec, H3A 1G1 Canada signy.sheldon@mcgill.ca

Acknowledgments The authors would like to thank Carina Fan and Frederik CrepeauHubert for help with data collection and scoring as well as Caterina Agostino and Elizabeth DuTemple for editorial assistance.

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 15–29

Full-Length Research Report

Do South African Xhosa-Speaking People with Schizophrenia Really Fare Better? A Longitudinal Mortality Study in Older Patients with Schizophrenia Dana Niehaus1 , Esme Jordaan2, Riana Laubscher2, Taryn Sutherland3, Liezl Koen1, and Felix Potocnik4 1

Department of Psychiatry and Stikland Hospital, Faculty of Health Sciences, University of Stellenbosch, Bellville, South Africa Biostatistics Unit, Medical Research Council, Bellville, South Africa

2 3

Statistics and Population Studies Department, University of the Western Cape, South Africa

Department of Psychiatry, Faculty of Health Sciences, University of the Witwatersrand, South Africa

Abstract: Objectives: Results from multinational WHO studies suggest that schizophrenia patients in developing countries may have more favorable prognoses and morbidity outcomes than those in developed settings. This study serves to establish whether mortality outcomes in South African Xhosa-speaking schizophrenia patients are more favorable than in the general South African population. Methods: We recruited a group of 981 patients from September 1997 to March 2005 as part of a genetic study in the Western, Southern, and Eastern Cape provinces of South Africa. For this substudy, participants were included when they reached the age of 60 years during the study period (8–15 years). We examined factors associated with the probability of dying and computed survival times using national census data as reference. Results: At the time of follow-up, 73 individuals were 60 years or older (21.9% could not be traced); some 40% of the sample had died at the time of the follow-up assessment (mean age at death = 60.12 years, SD = 4.97). Univariate survival analysis, using duration of disorder, revealed that the number of hospitalizations and psychotic episodes impacted survival time. Compared to the age-specific death rates of the general South African population, the death rate in the Xhosa-speaking schizophrenia sample was higher than expected in the 60–69 years category, but lower than expected in the 70+ years category. Conclusion: This study suggests that increased exposure to inpatient mental healthcare (expressed as number of hospitalizations) at baseline, and number of psychotic episodes, improve survival probability in a group of older South African Xhosa-speaking schizophrenia patients. Keywords: mortality, death rate, schizophrenia, elderly, South African

Introduction Schizophrenia, a serious, chronic mental illness typically requiring long-term follow-up, is associated with an elevated mortality rate relative to the general population. Results from a series of multinational World Health Organization (WHO) studies (Harrison et al., 2001; Jablensky & Sartorius, 2008) suggest that schizophrenia patients in developing countries experience more favorable prognoses, morbidity outcomes, and possibly lower mortality rates than those in more developed settings. The reasons for this phenomenon are presently unknown (Emsley et al., 2002). The suggestion that schizophrenia patients in developing countries enjoy better outcomes is not universally accepted Ó 2020 Hogrefe

(Cohen, Patel, Thara, & Gureje, 2008; Patel, Cohen, Thara, & Gureje, 2006). Most published mortality studies are from developed nations, with fewer from less-developed countries, especially from Africa (Cohen et al., 2008; Lee et al., 2018). Even the large WHO studies include only limited data from Africa and no data from Southern Africa. South Africa (SA) has begun a transformation of its healthcare system – one of the aims being to ensure needs-based care. Research into indicators of disease, including cause-specific mortality rates and factors influencing them, is needed if we are to identify, plan, and evaluate appropriate public-health interventions and services. To our knowledge, to date no mortality studies have compared schizophrenia patients to the general population in SA. GeroPsych (2020), 33(1), 31–41 https://doi.org/10.1024/1662-9647/a000217

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

Globally, patients receiving psychiatric care have a higher all-cause mortality risk than the general population (Lee et al., 2018), and among patients with mental illness, those with schizophrenia have one of the highest mortality rates (Joukamaa, Heliovaara, Knekt, & Aromaa, 2001; Kiseley, Smith, Lawrence, & Maaten, 2005). The life expectancy of schizophrenia patients has been estimated to be 20% lower than that of the general population (Newman & Bland, 1991), and a recent systematic review and metaanalysis suggested a weighted average of 14.5 years of potential life lost (Hjorthoj, Sturup, McGrath, & Nordentoft, 2017). Nielsen, Uggerby, Jensen, and McGrath (2013) reported a lower average age of death in the Danish schizophrenia population than in the general population – a finding in keeping with early observations by Emil Kraepelin (Kraepelin, 1919). Nearly a century later, despite advances in antipsychotic pharmacotherapy, psychosocial interventions, and medical care, the mortality gap between schizophrenia patients and the general population in fact appears to be widening, with the mean standardized mortality ratio (SMR) increasing from 2.2 in the pre-1970s studies to 3.0 in the post-1970s studies (Lee et al., 2018). Upon examining Danish death-register data from 1980 to 2010, Nielsen et al. (2013) found that the respective ages at death of men and women in the general population had increased by 0.28 and 0.31 years per calendar year, while in schizophrenia patients it decreased by 0.04 years and 0.05 years, respectively. The reasons for the apparent failure of schizophrenia patients to benefit from life-prolonging interventions have been a subject of study since Kraepelin (1919) first attributed the increased risk of dying to a combination of environmental and behavioral factors, including negativism, poor treatment adherence, and suicide. Suicide among schizophrenia sufferers accounts for a substantial proportion of the excess mortality (Piotrowski et al., 2017). Suicide alone does not, however, account for the entire excess mortality burden encountered in schizophrenia (Nielsen et al., 2013; Piotrowski et al., 2017). Schizophrenia patients are also at higher risk for metabolic abnormalities and cardiovascular disease (Callaghan, Boire, Lazo, McKenzie, & Cohn, 2009; Fan, Wu, Shen, & Zhan, 2013). Purported contributing factors include increased smoking rates, sedentary lifestyles, poor dietary choices, elevated body mass index, and higher incidence of diabetes (Ringen, Engh, Birkenaes, Dieset, & Andreassen, 2014; Walker, McGee, & Druss, 2015). The metabolic abnormalities may be linked to a complex interplay between genetic, illness-, lifestyle-, and medication-related factors (Jeon & Kim, 2017; Manu, Dima, & Shulman, 2015). Research is inconclusive regarding the association between long-term mortality risk and the use of antipsychotic treatment (Vermeulen et al., GeroPsych (2020), 33(1), 31–41

2017). However, a recent meta-analysis by Vermeulen et al. (2017) showed a reduced risk ratio (0.57; LL 0.46, UL 0.76; p < .001) associated with any exposure to antipsychotic medication. The contribution of cancer to mortality in schizophrenia was the subject of a recent systematic review and meta-analysis by Zhuo, Tao, Jiang, Lin, and Shao (2017). Of the included studies, 15 reported a pooled SMR of 1.40 (95% CI [1.29, 1.52], p < .001) and the other four studies a hazard ratio of 1.51 (95% CI [1.13, 2.03], p < .01), favoring a significantly increased risk of cancer mortality in schizophrenia patients compared with the general population. Schizophrenia sufferers also have higher mortality rates from infectious diseases, including pneumonia (Chou, Tsai, & Chou, 2013; Joukamaa et al., 2001) and HIV (Closson et al., 2019). Lack of access to adequate care may contribute to the increased mortality. Patients with mental illness report greater difficulty in accessing primary healthcare and experience lower quality of care than the general population (Nasrallah et al., 2006). Faasen, Niehaus, Koen, and Jordaan (2014) found a prevalence rate of 44.8% for metabolic syndrome and 13.8% for undiagnosed diabetes mellitus in a group of South African Xhosa-speaking schizophrenia patients on clozapine treatment, in keeping with a report that diabetic patients with severe mental illness received compromised medical care (Goldberg et al., 2007). Schizophrenia patients are more likely than the general population to suffer surgical complications (Cooke et al., 2007; Daumit et al., 2006). Patients with schizophrenia experienced lower quality of postmyocardial infarction care (Druss, Bradford, Rosenheck, Radford, & Krumholz, 2001; Kurdyak, Vigod, Calzavara, & Wodchis, 2012), and women with schizophrenia were less likely to receive routine cervical cancer screening (Martens et al., 2009) or mammographs (Chochinov et al., 2009). In summary, schizophrenia is associated with increased mortality, for a variety of possible reasons, compared to general populations in geographical samples outside of Southern Africa. This study examines whether a cohort of South African Xhosa-speaking schizophrenia patients aged 60+ years undergoing follow-up as part of a larger longitudinal study (Niehaus et al., 2005) also showed increased mortality relative to the general South African population. Survival estimates were computed, and factors associated with the probability of death were examined.

Methods Recruitment and Assessment We enrolled 981 South African Xhosa-speaking patients with DSM-IV-diagnosed schizophrenia from community Ó 2020 Hogrefe

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

clinics and psychiatric hospitals in SA in a prospective genetic study between September 1997 and March 2005. Subjects were followed up 8–15 years later. The recruitment and clinical assessment procedures have been discussed previously (Niehaus et al., 2004; Niehaus et al., 2005). Participants signed informed consent for the study, which had been ethically approved by the Institutional Review Board of the University of Stellenbosch. The present study describes mortality statistics in a subgroup (n = 73) of subjects who were 60+ years on entering the study or during follow-up. Patients and/or family members were interviewed individually, and data was gathered using a locator and/or mortality questionnaire. If the patient was deceased, the mortality questionnaire, containing details of death (date, location, and cause), was completed by family members of the deceased. Every effort was made to trace patients. Where relevant, data were gathered from hospitals, treating physicians, and postmortem results to ascertain cause of death.

Statistical Analysis Subjects were classified into three groups according to their status at follow-up: “alive,” “deceased,” and “lost to followup.” The numeric characteristics were compared among the three groups using an F-test for overall mean differences and a t-test for the pairwise mean comparisons with a Tukey-Kramer adjustment for multiple comparisons. The categorical characteristics were compared among the three groups using Pearson’s chi-square tests. An assessment was made regarding the lost to follow-up (missing) data and the assumption of MAR (missing at random). A missing data indicator was created, and the characteristics were compared between the missing and nonmissing groups using a chi-square or Fisher’s exact statistic. Time (years) to event (death) were calculated using the duration the patients had lived with their illness (“time living with the disorder”) and age attained by the end of the study period (if the patient was deceased, then that age was viewed as age at death). Since data and results were similar for both methods, we discuss only “time living with the disorder.” Estimates of survival times overall and for relevant subgroups were computed using the product-limit method (Kaplan-Meier method). Survivor functions were plotted and compared by the nonparametric log-rank and Wilcoxon tests. Age-specific mortality rates of the patients were compared with national census data (Dorrington, Bradshaw, Laubscher, & Nannan, 2014; Stats SA, 2014). The probability of death for the age groups as per the census data was Ó 2020 Hogrefe

applied to the number of deaths in the cohort to obtain the expected deaths. The observed schizophrenia deaths were then compared to the expected deaths and expressed as a proportion, where > 1 indicate an excess of schizophrenia deaths. Finally, a Cox regression adjusting for the age groups 60–64/65+ were done with the variables; number of psychotic episodes and the number of hospitalizations to assess impact on survivorship. Analyses were performed with SAS software (version 17.0) and the significance level was 5%.

Results At the time of follow-up, 73 individuals (58.9% male: 41.1% female) aged 60+ years were identified from the South African Xhosa-speaking genetics study database.

Sample Characteristics Seventy-three patients were at least 60 years old at some point in the study. Subjects were followed up for a mean period of 12.1 years (SD = 3.67, range 8–15). The projected mean age at follow-up was 66.3 years (SD = 7.42, range 60–98). Table 1 shows the sociodemographic characteristics of the 73 subjects at entry into the study stratified by survival status at the end of follow-up. Illness-related characteristics, as obtained at the initial interview, are shown in Table 2. There were no significant differences among the three groups in terms of these variables; 32% (n = 23) had died by the end of the follow-up period. Table 3 shows the causes of death reported by the families. Sixteen (21.9%) of the patients could not be traced despite exhaustive efforts; 34 patients (46.6%) were found to be alive. Substance use, abuse, or dependency (including nicotine) was similar in all groups (Table 2). The only global item elicited by the DIGS that differed significantly across groups were global bizarre behavior (marginally significant; p < .05). Fewer patients with bizarre behavior had died compared to those without (16% vs. 41%), and more were lost to follow-up (32% vs. 14%).

Survival Times In longitudinal studies with protracted follow-up periods, attrition rates are expected to be high, especially in studies such as the current one, in which patients were generally marginalized, poor, and less formally educated. To determine whether those lost to follow-up were similar to those followed up (deceased and alive), we compared the two groups regarding number of episodes, hospitalizations, and global SAPS and SANS scores. The group lost to GeroPsych (2020), 33(1), 31–41

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

Table 1. Biographical information across three groups at initial assessment (baseline) Alive (% of total)

Deceased (% of total)

Lost to follow-up (% of total)

Total (% of total)

n = 34

n = 23

n = 16

n = 73

Female

13 (38.2%)

10 (43.5%)

7 (43.8%)

30 (41.1%)

Male

21 (61.8%)

13 (56.5%)

9 (56.2%)

43 (58.9%)

Biographical Variable Sex

Marital status Divorced

4 (16.7%)

2 (10.5%)

0 (0.0%)

6 (11.1%)

Single

12 (50.0%)

12 (63.2%)

5 (45.5%)

29 (53.7%)

Separated

3 (12.5%)

1 (5.3%)

1 (9.0%)

5 (9.3%)

Widowed

5 (20.8%)

4 (21.0%)

5 (45.5%)

14 (25.9%)

n = 10

n=4

n=5

n = 19

24 (82.8%)

13 (100.0%)

6 (75.0%)

43 (86.0%)

Missing data Occupational status at first interview (Receives disability grant) Employed (Self, temporary, permanent)

0 (0.0%)

(Unemployed)

5 (17.2%)

0 (0.0%)

2 (25.0%)

7 (14.0%)

(Missing data)

n=5

n = 10

n=8

n = 23

Highest level of education No schooling

2 (7.1%)

1 (7.7%)

1 (12.5%)

4 (8.2%)

Grade 1–7

9 (32.2%)

4 (30.8%)

6 (75.0%)

19 (38.7%)

Grade 8–12

15 (53.6%)

8 (61.5%)

1 (12.5%)

24 (49.0%)

2 (7.1%)

0 (0.0%)

2 (4.1%)

n=6

n = 10

n=8

n = 24

Tertiary Missing data

follow-up did not differ significantly from those who were followed up. Hence, missing data were assumed to be MAR for the variables listed above in the assessment of survival time. Survival times were estimated using both the duration of the illness (“time living with the disorder”) and age attained by the end of the study period (if the patient was deceased, then that age was viewed as age at death). Since data and results were similar for both methods, only “time living with the disorder” is discussed. Duration of illness (difference between age at onset and either age at death or age at follow-up interview if alive) could be calculated for 49 subjects. Product-limit survival estimates were calculated and showed the median survival time with the disease to be 50 years. Of the group still alive at the follow-up interview, the subject who had been living with the disease for the longest period had a duration of illness of 58 years at that interview. Further stratified analysis revealed that the number of hospitalizations (categorized as 0–2 and more than 2), number of psychotic episodes (categorized as 1–3 and more than 3, p-value marginal) and global SANS score (p-value for log rank < .05) had a possible impact on survival time with the disorder, while global SAPS subscale scores, marital status, sex, substance use or abuse, and previous suicide attempts had no significant impact on survival time. The median survival time for 0–2 hospitalizations was 41 years and 50 years for more than two hospitalizations. The median survival time for GeroPsych (2020), 33(1), 31–41

1–3 episodes was 41 years, while it was 50 years for more than three episodes. Product-limit survival curves for the number of hospitalizations (Figure 1) and number of episodes (Figure 2) illustrate that subjects with more than two hospitalizations had a significantly better survival rate than those subjects with two or less hospitalizations (log-rank test p < .05). Those with more than three episodes had a marginally better survival rate compared to those subjects with less than or equal to three episodes (log-rank test p = .05). None of the other covariates considered were significant predictors of survival time. The mortality (death) rate in our sample was compared to the age-specific death rates in the 60+ South African general population (Table 4). These results show that the death rate in the schizophrenia sample at the time of last follow-up was higher than expected in the 60–64 years category – and especially in the 65–69 years category – ut decreased to below the expected death rate in the 70+ years category (Table 5).

Discussion The mean age at death was 60.1 years (n = 23; missing data = 4), and 34 individuals were still alive at a mean age Ó 2020 Hogrefe

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

Table 2. Age at initial interview, age at onset of illness, duration of illness, age at mortality assessment, and number of episodes and hospitalizations across three groups at the initial interview (baseline) Alive

Deceased

Lost to follow-up

Range

47–70

44–83

51–71

Mean

59.9

4.6

8.8

6.3

Range

60–79

60–98

60–86

Mean

63.6

66.6

71.6

4.0

8.8

8.5

Range

10–50

15–52

14–55

Mean

28.8

28.0

33.6

11.2

11.5

14.2 0–50

Age at initial interviewy (years)

Age at follow-upy (years)

Age of onset of illness (years)

Duration of illness (at initial interview) (years) Range

0–50

0–47

Mean

25.3

12.2

12.1

15.2

Number of episodes (at initial interview)à n=3

n=4

n=2

Missing data

5 (16.2%)

5 (26.3%)

5 (35.7%)

2–5

13 (41.9%)

11 (57.9%)

7 (50.0%)

More than 5

13 (41.9%)

3 (15.8%)

2 (14.3%)

1–10

1–11

1–20

Range Mean

5.0

3.7

4.1

3.1

2.8

5.2

Number of hospitalizations (at initial interview)à Missing data Never admitted 1

n=1

n=4

n=1

1 (2.9%)

0 (0.0%)

3 (20.0%)

3 (8.8%)

4 (21.0%)

1 (6.7%)

2–5

17 (50.0%)

12 (63.2%)

9 (60.0%)

More than 5

13 (38.3%)

3 (15.8%)

2 (13.3%) 0–10

Range

0–10

1–11

Mean

4.8

3.5

2.7

3.3

2.8

2.6

0–2

11 (33%)

11 (58%)

9 (60%)

More than 2

22 (67%)

8 (42%)

6 (40%)

0–3

12 (35%)

11 (58%)

11 (79%)

More than 3

19 (63%)

8 (42%)

3 (21%)

Number of hospitalizations (at initial interview)

Number of episodes (at initial interview)

Note: ySignificant intergroup differences detected. See text for details. àNumber of episodes and hospitalizations at mortality interview is not reflected in the data as comparisons cannot be drawn between the groups given the “lost to follow-up” status in the study sample.

of 63.6 years over the mean follow-up period of 12.14 years. In the 78% of the sample that could be followed up, the mortality rate was 40.3%. Although the purpose of the study was to compare the sample mortality rate with the national average, interpretation of the results requires an understanding of international data. Ó 2020 Hogrefe

International Data The International Study of Schizophrenia (IsoS) (Harrison et al., 2001) incorporated multiple country cohort sizes varying between 55 and 148 individuals with follow-up periods ranging from 12–26 years (majority 15 years) and the percentage followed up from 43.7% to 77%. Mortality over

GeroPsych (2020), 33(1), 31–41

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

Table 3. Comorbidity patterns: substance use at baseline Alive (% of total)

Deceased (% of total)

Lost to follow-up (% of total)

Total (% of total)

Alcohol use disorders

8 (23.5%)

4 (17.4%)

0 (0.0%)

12 (16.4%)

Cannabis use or abusey (defined as use of more than once per lifetime)

7 (20.6%)

5 (21.7%)

1 (6.3%)

13 (17.8%)

Tobacco use

19 (55.9%)

12 (52.2%)

7 (43.8%)

38 (52.1%)

Other drugs

0 (0.0%)

Note: ySignificant difference detected. See text for details.

Figure 1. Product-limit survival estimates for the categorized variable – number of hospitalizations.

Figure 2. Product-limit survival estimates for the categorized variable – number of episodes.

GeroPsych (2020), 33(1), 31–41

Ó 2020 Hogrefe

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

Table 4. Overall and stratified summary statistics and rank tests for the time living with disorder (n = 49)

Total group

Median

Meany

41.8

1.86

Test of equality over strata p-values

Number deceased

Log-rank

Wilcoxon

0.040

0.048

0.053

0.068

Number of hospitalizations (categorized) 0–2 times

35.6

2.59

> 2 times

45.9

1.76

1–3

35.9

2.38

45.8

1.79

Number of episodes (categorized)

Note: yUnderestimates since largest observation was censored.

Table 5. Mortality (death rate) of Xhosa schizophrenia sample compared to the age-specific death rates in the 60+ South African general population Death rate

Cohort 60+

Observed deaths

Redist unknown

Total observed deaths

Expected deaths

“Observed/Expected deaths”

60–64

0.21862

8.96

1.89y

65–69

0.29751

5.36

3.17y

70–74

0.40033

1.60

0.00

75+

0.60621

6.06

0.84

21.98

1.77

Age group

Total Note: yDeath rate higher than expected.

the follow-up period ranged from 6.3% (SMR 1.8; Chandigarh – urban, India) to 31.2% (SMR 1.86; Agra, India). Non-IsoS studies (Holla & Thirthalli, 2015, provide a detailed description) varied in sample size from 30 to 2071 individuals (six studies had sample sizes of more than 400), with follow-up periods of 2–30 years (11/14 had follow-up periods of less than 12 years). Of these 14 studies, 9 reported on mortality rates (ranging from 2%–25.4%). It is important to note that these international studies did not focus on the “graduation” period (i.e., patients reaching 60 years of age and graduating to psychogeriatric services). By comparison, our study size sample falls in the lower quartile, though it represents a follow-up percentage in the high range and a relatively high mortality percentage during the follow-up period.

The South African Context We used the Rapid Mortality Surveillance System (RMS) 2014 report and the SATS SA census data of 2011. The RMS 2014 report monitors trends in adult deaths, providing “timely empirical estimates” of high-level indicators of mortality. The RMS 2005 and 2014 reports recorded an average life expectancy of 54 and 62 years (Dorrington Ó 2020 Hogrefe

et al., 2014; Stats SA, 2014), respectively, indicating an improved adult mortality probability (45q15) from 46% before age 60 (2009), to a 36% probability (2013). The improvement is seen in both sexes (male 51% to 42%, female 40% to 30%). The total number of deaths in the 60+ years group increased from 143,137 (including 3,635 unnatural deaths) in 2000 to an estimated 191,255 (including 5,684 unnatural deaths) in 2013. Because of methodological differences, discrepancies occur between Stats SA and RMS data (7,000 fewer deaths reported on Stats SA 2011 census data). Overall, despite the age-related mortality changes, the expected future lifespan of a South African 60-year-old remains in a narrow band of between 15 and 20 years.

Hypotheses for Differences Between Schizophrenia and National Census Age-Related Death Rate Several hypotheses exist regarding the increased death rate in the 60–69-year-old Xhosa-speaking schizophrenia group, compared with national census figures. These include loss of social support systems, limited access to healthcare, and increased mortality risks associated with GeroPsych (2020), 33(1), 31–41

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

antipsychotic treatment, suicidality, and comorbid substance use disorders with associated risk behavior.

reluctant to attend clinics because “people would then know that they were ill.”

Impact of Poor Social Connection Loss of a social support system is expected, given that their parents will be beyond expected survival age (> 78 years if patient was born at parental age of 18 years). Also, none of the participants in the follow-up sample were married at the time of baseline assessment, and other social challenges (low employment and income potential with direct impact on accessibility of retirement facilities) are common in South African schizophrenia patients (Brink, Niehaus, Oosthuizen, Muller, & Koen, 2004). The latter mimics a possible variant of the social drift theory (Brink et al., 2004). While the association between marital status and mortality still needs to be clarified for schizophrenia, divorce and the loss of a life partner are associated with an increased mortality risk in the general population (Sbarra, 2015; Shor et al., 2012). The situation of an affected sibling-pair (Niehaus et al., 2004) subsample (n = 99), compared to singleton probands, of the South African Xhosa-speaking schizophrenia genetic group suggested that having an affected sibling has a protective effect. It should also be considered that the proposed loss of caregivers may exacerbate an underlying lack of social connectedness or a social distancing, which are associated with the diagnosis and treatment of schizophrenia and related perceived stigma. Both social connectedness and purposeful goals serve as proposed drivers of longevity (Green, Horan, Lee, McCleery, Reddy, & Wynn, 2018; Wagner, TorresGonzález, Runte Geidel, & King, 2011), and they have also been associated with a decreased mortality risk of 29%– 32% in the general population (Holt-Lunstad, Smith, Baker, Harris, & Stephenson, 2015). Within the South African context, stigma and misperceptions about schizophrenia seem to be pervasive in the general population as well as in medical students and family members of individuals with schizophrenia (Botha, Koen, & Niehaus, 2006; De Witt, Smit, Jordaan, Koen, Niehaus, & Botha, 2019; Hugo, Boshoff, Traut, Zungu-Dirwayi, & Stein, 2003; Mbanga et al., 2002).

Age-Related Medical and Mortality Factors With advanced age, physiological changes make the elderly more sensitive to multiple illnesses, including cardiovascular diseases that may contribute to mortality. The global burden of disease study (Murray & GBD, 2015) reported on country-specific probabilities of death during middle age (defined as 50–75 years of age), and the probability of dying ranged from 10.3% (in women in Andorra) to 76.3% (in men in Lesotho). The causes leading to death in middle age differed among the geographical areas, with stroke and other cardiovascular diseases being major contributors in sub-Saharan Africa. Other important contributors to mortality were HIV/AIDS, tuberculosis, diarrhea, and lower respiratory tract infections. No individual in our sample was reported to have suffered heart disease, but hypertension (n = 1) may represent a proxy marker for heart disease. Strokes (n = 3) and complications from diabetes (n = 3) were commonly noted causes of death in our sample. Surprisingly, cancer deaths were recorded as frequently as strokes. Death from cancer is not unexpected in this population because, although the majority of international studies support a lower overall cancer risk in schizophrenia, the survival rate for comorbid schizophrenia and cancer is lower than that of cancer alone (Chou, Tsai, Su, & Lee, 2011). Only one AIDS-related death was reported. This finding requires further exploration as a possible cohort effect. It is also surprising, given the poor HIV/AIDS risk-behavior knowledge reported in a subgroup of 102 patients from a South African schizophrenia genetic study (Koen, Uys, Niehaus, & Emsley, 2007). Only one case of suicide as primary cause of death was reported. To place this in context, we need to consider that 13.8% of deaths in the Ethiopian 10-year follow-up study (Shibre et al., 2015) were attributed to suicide. It is also less than expected considering previously reported findings in a larger South African Xhosa-speaking schizophrenia genetic study (including all ages at baseline) (Luckhoff, Koen, Jordaan, & Niehaus, 2014; Niehaus et al., 2004). Suicide attempts were common (14.1%) at baseline assessment in the total genetic sample (n = 974) (Niehaus et al., 2004) and correlated very well with the attempt rate in the mortality (follow-up) sample (15% attempted suicide prior to baseline). Since both comparison datasets have a much younger mean age at assessment, it is unclear whether suicide rates plateau in mid to late life in South African Xhosa-speaking schizophrenia patients. In the current mortality (followup) study, the deceased group had the highest rate of previous suicide attempts at baseline (26.1% versus 8.8% in the alive group and 12.5% for the lost-to-follow-up group), but only one suicide death was reported at follow-up. These

Stigma Associated with Schizophrenia Mbanga and colleagues (2002) reported that 67% of relatives of Xhosa-speaking schizophrenia patients believed the development of schizophrenia is associated with witchcraft or possession by evil spirits. The patients were also viewed as dirtier (52%), more unpredictable (45%), and more dangerous (44%) than the average person. Botha et al. (2006) found that most participants with schizophrenia had been subjected to name-calling or verbal abuse. Almost 40% reported physical abuse directly linked to their mental illness, and 16% generally agreed that they were

GeroPsych (2020), 33(1), 31–41

Ó 2020 Hogrefe

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

data also suggest that the majority of patients (> 75%) only attempted suicide once, and that the most common method was poisoning. Associated Socioeconomic Factors Limited access to healthcare following loss of income or retirement (although this is unlikely as the minority of the subjects were active in the workforce), loss of mobility through frailty (a final common pathway, e.g., from stroke), or chronic diseases that were not optimally controlled (i.e., diabetes and/or hypertension) are all factors that may have contributed to the increased mortality. In this follow-up study, 30.4% of deaths were attributed to complications of diabetes mellitus (n = 3), hypertension (n = 1), and strokes (n = 3). These findings regarding limited access to appropriate healthcare are supported by a recently published study by Faasen et al. (2014). Differential care (i.e., general practitioner versus specialist) reported on by other studies (Kiseley et al., 2005) is not a testable hypothesis, as all the patients in this study were being treated in the state sector. Such care can vary significantly from visit to visit (i.e., assessment by nurse, medical officer, registrar, or specialist). Within South Africa, limited access to healthcare seems a likely hypothesis supporting the positive correlation between the patient’s contact with psychiatric health services (number of hospitalizations) and longer survival time. This finding, although overtly surprising and contrary to some international literature is not without precedent in international datasets. Better contact with mental-health services has been associated with the differential mortality gap found between schizophrenia sufferers and the general population (Nielsen, Uggerby, Jensen, & McGrath, 2013; Osby, Correia, Brandt, Ekbom, & Sparen, 2000). Country-Specific Variables Similar to the finding by Emsley et al. (2002), this study also raises the question whether some South African schizophrenia patients show a better response to antipsychotic treatment. Many possible variables with complex interplay may impact country-specific findings. Hypotheses include regional factors (Holla & Thirthalli, 2015), which can encompass ethnicity, sociopolitical status, urbanicity rates, level of social support, family structure, and even expressed emotion levels within a regional group or subgroup. Medication Impact Although numerous observational studies associate an increased mortality risk with the use of conventional antipsychotics, a recent meta-analysis and systematic review by Hulshof, Zuidema, Ostelo, and Luijendijk (2015) of 17 trials involving 2,387 elderly patients does not support this hypothesis (pooled risk difference 0.1%; risk Ó 2020 Hogrefe

ratio 1.07; 95% CI [0.54, 2.13]). This is an important review when we consider that most of our sample was exposed to typical antipsychotics (only 10% were on clozapine) because of the lack of access to atypical antipsychotics (excluding clozapine).

Study Limitations The results should be viewed within the context of the study limitations, which includes a selection bias (see Niehaus et al., 2005, for the recruitment strategy of the larger sample), limited sample with loss to follow-up and survival bias. Sixteen (22%) of the patients were lost to follow-up. The lost-to-follow-up group was compared to the deceased or alive patients regarding their biographical information, age at interview, age at onset of disease, duration of disease, age at mortality assessment, number of episodes and hospitalizations, and substance use, but no significant differences were found (data not shown). One possible problem with the reported results is that of survival bias, i.e., patients living longer have more time to be admitted as well as more possible episodes and outpatient contact. To investigate possible survival bias, we compared two strata – the age groups 60–64 and 65+ years – but we did not find a difference in the survival rate (p > .5). Although the sample size might be too small to adjust for covariates, we did a Cox regression adjusting for the age groups 60–64/65+ and found that there were marginally significant effects of the number of psychotic episodes (p < .05) and the number of hospitalizations (p < .1) on survivorship. We accept that there might be other excluded covariates important to the analysis, but the sample size did not allow any further multiple regression modelling.

Conclusions This study suggests that increased exposure to inpatient mental healthcare (expressed as number of hospitalizations) at baseline and number of psychotic episodes improve survival probability in a group of older South African Xhosa-speaking schizophrenia patients. Although the mortality in this group is higher than that of the general population for 60–69-year-olds, it does improve once an individual reaches 70 years. Although the reasons for the increase in mortality in the 60–69-year-old group remain unconfirmed, the causes of death lie within expected international norms, though there is a low AIDS-related mortality rate. Future studies need to characterize the multiple variables and the complex interplay of specific regional and national factors within large sample sizes, and consider mechanisms that will improve follow-up rates and more GeroPsych (2020), 33(1), 31–41

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

accurate cause-of-death data between the general population and schizophrenia patients.

References Botha, U. A., Koen, L., & Niehaus, D. J. H. (2006). Perceptions of a South African schizophrenia population regarding community attitudes towards their illness. Social Psychiatry & Psychiatric Epidemiology, 41, 619–623. Brink, S., Niehaus, D., Oosthuizen, P. P., Muller, J., & Koen, L. (2004). Occupation and schizophrenia in a Xhosa population: A variant of the social drift theory? South African Medical Journal, 94, 42. Callaghan, R. C., Boire, M. D., Lazo, R. G., McKenzie, K., & Cohn, T. (2009). Schizophrenia and the incidence of cardiovascular morbidity. Schizophrenia Research, 115, 325–332. Chochinov, H. M., Martens, P. J., Prior, H. J., Fransoo, R., Burland, E., & Team, Need. To. Know. (2009). Does a diagnosis of schizophrenia reduce rates of mammography screening? A Manitoba population-based study. Schizophrenia Research, 113, 95–100. Chou, F. H., Tsai, K. Y., & Chou, Y. M. (2013). The incidence and allcause mortality of pneumonia in patients with schizophrenia: A nine-year follow-up study. Journal of Psychiatric Research, 47, 460–466. Chou, F. H., Tsai, K. Y., Su, C. Y., & Lee, C. C. (2011). The incidence and relative risk factors for developing cancer among patients with schizophrenia: A nine-year follow-up study. Schizophrenia Research, 129, 97–103. Closson, K., McLinden, T., Patterson, T. L., Eyawo, O., Kibel, M., Card, K. G., et al. (2019). HIV, schizophrenia, and all-cause mortality: A population-based cohort study of individuals accessing universal medical care from 1998 to 2012 in British Columbia, Canada. Comparative Outcomes and Service Utilization Trends (COAST) Study. Schizophrenia Research, 209, 198–205. Cohen, A., Patel, V., Thara, R., & Gureje, O. (2008). Questioning an axiom: Better prognosis for schizophrenia in the developing world? Schizophrenia Bulletin, 34, 229–244. Cooke, B. K., Magas, L. T., Virgo, K. S., Feinberg, B., Adityanjee, A., & Johnson, F. E. (2007). Appendectomy for appendicitis in patients with schizophrenia. The American Journal of Surgery, 193(1), 41–48. Daumit, G. L., Pronovost, P. J., Anthony, C. B., Guallar, E., Steinwachs, D. M., & Ford, D. E. (2006). Adverse events during medical and surgical hospitalizations for persons with schizophrenia. Archives of General Psychiatry, 63, 267– 272. De Witt, C., Smit, I., Jordaan, E., Koen, L., Niehaus, D. J. H., & Botha, U. (2019). The impact of a psychiatry clinical rotation on the attitude of South African final year medical students towards mental illness. BMC Medical Education, 19, 114. Dorrington, R. E., Bradshaw, D., Laubscher, R., & Nannan, N. (2014). Rapid mortality surveillance report 2013. Cape Town: South African Medical Research Council. Druss, B. G., Bradford, D., Rosenheck, R. A., Radford, M. J., & Krumholz, H. M. (2001). Quality of medical care and excess mortality in older patients with mental disorders. Archives of General Psychiatry, 58, 565–572. Emsley, R. A., Roberts, M. C., Rataemane, S., Pretorius, J., Oosthuizen, P. P., Turner, J., . . . Stein, D. J. (2002). Ethnicity and treatment response in schizophrenia: A comparison of 3 ethnic groups. Journal of Clinical Psychiatry, 63(1), 9–14.

GeroPsych (2020), 33(1), 31–41

Faasen, N., Niehaus, D. J. H., Koen, L., & Jordaan, E. (2014). Undiagnosed metabolic syndrome and other adverse effects among clozapine users of Xhosa descent. South African Journal of Psychiatry, 20(2), 54–57. Fan, Z., Wu, Y., Shen, J., & Zhan, R. (2013). Schizophrenia and the risk of cardiovascular diseases: A meta-analysis of thirteen cohort studies. Journal of Psychiatric Research, 47, 1549– 1556. Goldberg, R. W., Kreyenbuhl, J. A., Medoff, D. R., Dickerson, F. B., Wohlheiter, K., Fang, L. J., . . . Dixon, L. B. (2007). Quality of diabetes care among adults with serious mental illness. Psychiatric Services, 58, 536–543. Green, M. F., Horan, W. P., Lee, J., McCleery, A., Reddy, L. F., & Wynn, J. K. (2018). Social disconnection in schizophrenia and the general community. Schizophrenia Bulletin, 44, 242– 249. Harrison, G., Hopper, K., Craig, T., Laska, E., Siegel, C., Wanderling, J., . . . Wiersma, D. (2001). Recovery from psychotic illness: A 15- and 25-year international follow-up study. British Journal of Psychiatry, 178, 506–517. Hjorthoj, C., Sturup, A. E., McGrath, J. J., & Nordentoft, M. (2017). Years of potential life lost and life expectancy in schizophrenia: A systematic review and meta-analysis. Lancet Psychiatry, 4, 295–301. Holla, B., & Thirthalli, J. (2015). Course and outcome of schizophrenia in Asian countries: Review of research in the past three decades. Asian Journal of Psychiatry, 14, 3–12. Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). Loneliness and social isolation as risk factors for mortality: A meta-analytic review. Perspectives on Psychological Science, 10, 227–237. Hugo, C. J., Boshoff, D. E., Traut, A., Zungu-Dirwayi, N., & Stein, D. J. (2003). Community attitudes toward and knowledge of mental illness in South Africa. Social Psychiatry and Psychiatric Epidemiology, 38, 715–719. Hulshof, T. A., Zuidema, S. U., Ostelo, R. W., & Luijendijk, H. J. (2015). The mortality risk of conventional antipsychotics in elderly patients: A systematic review and meta-analysis of randomized placebo-controlled trials. Journal of the American Medical Directors Association, 16, 817–824. Jablensky, A., & Sartorius, N. (2008). What did the WHO studies really find? Schizophrenia Bulletin, 34, 253–255. Jeon, S. W., & Kim, Y. K. (2017). Unresolved issues for utilization of atypical antipsychotics in schizophrenia: Antipsychotic polypharmacy and metabolic syndrome. International Journal Molecular Sciences, 18(10), 601–612. Joukamaa, M., Heliovaara, M., Knekt, P., & Aromaa, A. (2001). Mental disorders and cause specific mortality. British Journal of Psychiatry, 179, 498–502. Kiseley, S., Smith, M., Lawrence, D., & Maaten, S. (2005). Mortality in individuals who have had psychiatric treatment: Populationbased study in Nova Scotia. British Journal of Psychiatry, 187, 552–558. Koen, L., Uys, S., Niehaus, D. J. H., & Emsley, R. A. (2007). Negative symptoms and HIV/AIDS risk behaviour knowledge in schizophrenia. Psychosomatics, 48, 128–134. Kraepelin, E. (1919). Dementia praecox and paraphrenia. Edinburgh: Livingstone. Kurdyak, P., Vigod, S., Calzavara, A., & Wodchis, W. P. (2012). High mortality and low access to care following incident acute myocardial infarction in individuals with schizophrenia. Schizophrenia Research, 142(1–3), 52–57. Lee, E. E., Liu, J., Tu, X., Palmer, B. W., Eyler, L. T., & Jeste, D. V. (2018). A widening longevity gap between people with schizophrenia and general population: A literature review and call for action. Schizophrenia Research, 196, 9–13.

Ó 2020 Hogrefe

D. Niehaus et al., Mortality in South African Patients with Schizophrenia

Luckhoff, M., Koen, L., Jordaan, E., & Niehaus, D. (2014). Attempted suicide in a Xhosa schizophrenia and schizoaffective disorder population. Suicide and Life-Threatening Behaviour, 44, 167–174. Manu, P., Dima, L., & Shulman, M. (2015). Weight gain and obesity in schizophrenia: epidemiology, pathobiology, and management. Acta Psychiatrica Scandinavica, 132, 97–108. Martens, P. J., Chochinov, H. M., Prior, H. J., Fransoo, R., & Burland, E., Need To Know Team. (2009). Are cervical cancer screening rates different for women with schizophrenia? A Manitoba population-based study. Schizophrenia Research, 113, 101–106. Mbanga, N. I., Niehaus, D. J., Mzamo, N. C., Wessels, C. J., Allen, A., Emsley, R. A., & Stein, D. (2002). Attitudes towards and beliefs about schizophrenia in Xhosa families with affected probands. Curiatonis, 25(1), 69–73. Murray, C. J. L., GBD 2013 Mortality and Causes of Death Collaborators. (2015). Global, regional, and national age-sex specific all-cause and cause-specific mortality for 240 causes of death, 1990–2013: A systematic analysis for the global burden of disease study 2013. Lancet, 385, 117–171. Nasrallah, H. A., Meyer, J. M., Goff, D. C., McEvoy, J. P., Davis, S. M., Stroup, T. S., & Lieberman, J. A. (2006). Low rates of treatment for hypertension, dyslipidemia and diabetes in schizophrenia: Data from the CATIE schizophrenia trial sample at baseline. Schizophrenia Research, 86, 15–22. Newman, S. C., & Bland, R. C. (1991). Mortality in a cohort of patients with schizophrenia: A record linkage study. Canadian Journal of Psychiatry, 36, 239–245. Niehaus, D. J., Laurent, C., Jordaan, E., Koen, L., Oosthuizen, P., Keyter, N., . . . Emsley, R. A. (2004). Suicide attempts in an African schizophrenia population: An assessment of demographic risk factors. Suicide and Life-Threatening Behaviour, 34, 320–327. Niehaus, D. J. H., Koen, L., Laurent, C., Muller, J., DeLeuze, J. F., Mallet, J., . . . Emsley, R. A. (2005). Positive and negative symptoms in affected sibling pairs with schizophrenia: Implications for genetic studies in an African Xhosa sample. Schizophrenia Research, 15(79), 239–249. Nielsen, R. E., Uggerby, A. S., Jensen, S. O., & McGrath, J. J. (2013). Increasing mortality gap for patients diagnosed with schizophrenia over the last three decades: A Danish nationwide study from 1980 to 2010. Schizophrenia Research, 146(1–3), 22–27. Osby, U., Correia, N., Brandt, L., Ekbom, A., & Sparen, P. (2000). Mortality and causes of death in schizophrenia in Stockholm County, Sweden. Schizophrenia Research, 29(45), 21–28. Patel, V., Cohen, A., Thara, R., & Gureje, O. (2006). Is the outcome of schizophrenia really better in developing countries? Brazilian Journal of Psychiatry, 28, 149–152. Piotrowski, P., Gondek, T. M., Królicka-Dereôgowska, A., Misiak, B., Adamowksi, T., & Kiejna, A. (2017). Causes of mortality in schizophrenia: An updated review of European studies. Psychiatria Danubina, 29, 108–120. Ringen, P. A., Engh, J. A., Birkenaes, A. B., Dieset, I., & Andreassen, O. A. (2014). Increased mortality in schizophrenia because

Ó 2020 Hogrefe

of cardiovascular disease: A non-systematic review of epidemiology, possible causes, and interventions. Frontiers of Psychiatry, 5, 137. Sbarra, D. A. (2015). Divorce and health: Current trends and future directions. Psychosomatic Medicine, 77, 227–236. Shibre, T., Medhin, G., Alem, A., Kebede, D., Teferra, S., Jacobson, L., . . . Fekadu, A. (2015). Long-term clinical course and outcome of schizophrenia in rural Ethiopia: A 10-year follow-up of a population-based cohort. Schizophrenia Research, 161, 414– 420. Shor, E., Roelfs, D. J., Curreli, M., Clemow, L., Burg, M. M., & Schwartz, J. E. (2012). Widowhood and mortality: A metaanalysis and meta-regression. Demography, 49, 575–606. Statistics South Africa (Stats SA). (2014). Mortality and causes of death in South Africa, 2011: Findings from death notification. Pretoria: Stats SA. Retrieved from http://www.statssa.gov.za/ publications/P03093/P030932011.pdf. Vermeulen, J., van Rooijen, G., Doedens, P., Numminen, E., van Tricht, M., & de Haan, L. (2017). Antipsychotic medication and long-term mortality risk in patients with schizophrenia: A systematic review and meta-analysis. Psychological Medicine, 47, 2217–2228. Wagner, L. C., Torres-González, F., Runte Geidel, A., & King, M. B. (2011). Existential questions in schizophrenia: Perceptions of patients and caregivers. Revista Saúde Pública, 45, 401–408. Walker, E. R., McGee, R. E., & Druss, B. G. (2015). Mortality in mental disorders and global disease burden implications: A systematic review and meta-analysis. JAMA Psychiatry, 72, 334–341. Zhuo, C., Tao, R., Jiang, R., Lin, X., & Shao, M. (2017). Cancer mortality in patients with schizophrenia: Systematic review and meta-analysis. British Journal of Psychiatry, 211(1), 7–13. History Received April 30, 2019 Accepted November 13, 2019 Conflict of Interest The authors declare no conflict of interest. ORCID Dana Niehaus https://orcid.org/0000-0001-9696-5605

Prof. Dana J.H. Niehaus, Ph.D, D.Med, MMed, MPhil, MBCHB Department of Psychiatry University of Stellenbosch Fransie van Zijl drive Bellville 7500 Western Cape South Africa djhn@sun.ac.za

GeroPsych (2020), 33(1), 31–41

Full-Length Research Report

Perceived Benefits and Costs Contribute to Young and Older Adults’ Selectivity in Social Relationships Erica L. O’Brien

and Thomas M. Hess

Department of Psychology, North Carolina State University, Raleigh, NC, USA

Abstract: This article explores the influence of perceived benefits and costs on willingness to engage in social interactions in 32 young adults aged 20 to 40 years and 38 older adults aged 65 to 85 years. Results showed (1) increases in perceived benefits and importance of each relationship but decreases in perceived costs associated with increases in network centrality, (2) reduced willingness in older adults to engage with social partners for whom perceived costs outweighed benefits, and (3) perceived costs and benefits subsumed the effects of the affective qualities of social interactions. Findings support an analysis of social behavior based on the selective engagement theory (Hess, 2014), with selection effects in willingness to engage in social interactions related to perceived benefits and costs. Keywords: social networks, benefits and costs, network centrality, willingness, age differences

The size and composition of social networks generally changes as people grow older. Several studies show that network size decreases, with relationships at the periphery of the network (e.g., largely friends and acquaintances) exhibiting the greatest reductions and those at the center remaining relatively stable (e.g., Carstensen, Fung, & Charles, 2003; Lang & Carstensen, 1994; Wrzus, Hänel, Wagner, & Neyer, 2013; cf. Lansford, Sherman, & Antonucci, 1998). Network size and composition also have important consequences for cognitive functioning, psychological well-being, and health-related behavior (e.g., physical and social activity) in older adulthood. Concerning size, findings suggest that a smaller network is generally associated with lower scores on measures of global cognition (Li & Dong, 2018) and lower levels of engagement in social activities (e.g., Huxhold, Fiori, & Windsor, 2013). Networks with more family members – often central figures who have important emotional functions within the network – have also been linked to negative behaviors and outcomes in certain cases. One study found no association between network composition and cognitive functioning in older adults who identified as non-Hispanic White, and lower levels of global cognition in those who identified as African American whose networks consisted mostly of family members (Sharifian, Manly, Brickman, & Zahodne, GeroPsych (2020), 33(1), 42–51 https://doi.org/10.1024/1662-9647/a000218

2019). In another study, depressive symptoms decreased prospectively in older adults overall as a function of having a greater total number of family members in the network but increased when these members made up a greater proportion of the network (Fuller-Iglesias, Webster, & Antonucci, 2015). These associations depended on negativity with family members. In the context of low family negativity, the total number of family members was associated with depression, with greater numbers predictive of fewer symptoms. In the case of high family negativity, depressive symptoms were higher on average but unrelated to the proportion of family members. Finally, having a greater proportion of family members also corresponds to a lower likelihood of engaging in healthy behaviors such as being physical active (Litwin, 2003; Shiovitz-Ezra & Litwin, 2012), having a regular exercise routine, and walking a mile or more at least once a week. The potential impacts of social networks on health in later life raise questions about the factors that drive how individuals manage their personal networks: Why do they maintain some relationships and disregard others? The current study specifically aimed to examine the factors that contribute to an individual’s willingness to engage with members of their networks. Extant work suggests that aging-related shifts in motivation partially account for changes in the make-up of social Ó 2020 Hogrefe

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

networks. Perhaps most notably, socioemotional selectivity theory (SST; Carstensen, Isaacowitz, & Charlels, 1999) proposes that a shift from an expansive perception of future time in young adulthood to a more limited perspective in old age results in an increase in the salience of emotion-based social goals relative to those associated with information seeking. This shift is assumed to lead to increased pursuit of emotionally meaningful social experiences in later life, with a selective narrowing of relationships and a focus on close and familiar others serving to promote emotional well-being. Empirical work in support of these ideas demonstrates that, more than young adults, older adults (1) differentiate between social partners based on familiarity and anticipated affect (Fredrickson & Carstensen, 1990), (2) prefer engagements with familiar partners (Fredrickson & Carstensen, 1990; Fung, Carstensen, & Lang, 2001; Fung, Carstensen, & Lutz, 1999), (3) have more emotionally close social partners in their networks (Fung et al., 2001), and (4) have smaller networks because of reductions in the number peripheral relationships (English & Carstensen, 2014). They further report feeling more satisfied (Lansford et al., 1998) as well as generally more positive and less negative about their networks (English & Carstensen, 2014).1 These findings support the idea that in order to achieve emotional satisfaction in line with their presumed goals, older adults create networks comprised largely of close and important relationships. However, variability does exist in the actual placement of these individuals – whether family (e.g., spouse, child, sibling) or nonfamily (e.g., friend, neighbor, acquaintance) – within the network (English & Carstensen, 2014) and in the degree to which a proportion of them are characterized as negative (e.g., Fuller-Iglesias et al., 2015) and exclusively positive (e.g., Rook, Luong, Sorkin, Newsom, & Krause, 2012). Other factors and motivations, over and above emotion-related ones, likely contribute to satisfaction achieved from interactions with network members. One of these factors relates to the personal resources possessed by individuals for dealing with the demands or costs associated with social interactions. For instance, older adults “poor” in personal resources engage in fewer social activities (Lang, Rieckmann, & Baltes, 2002), and generally perceive greater effort in social relationships than do younger adults (Lang, Wagner, Wrzus, & Neyer, 2013). This suggests that the demands related to social activity –

including those associated with dealing with negative interactions – increase in later life, putting stress on potentially reduced personal resources. Interestingly, older adults also perceive emotionally close relationships as more stressful than do young adults (Lang et al., 2013). The fact that older adults tend to maintain engagement in such relationships suggests a willingness to invest their limited resources in costly social situations if they perceive beneficial results. These findings do not necessarily contradict the idea that older adults’ deliberate pruning of their social networks reflects their strivings to maintain familiar relationships and optimize emotionally beneficial experiences. They do, however, highlight other aspects of social selection processes. Specifically, simply maintaining and engaging in social networks places demands on individuals’ personal resources, suggesting that the differential willingness to invest in familiar versus novel relationships may depend on the benefits and costs associated with each and generally how one wishes to allocate his or her resources. One could think about the impacts of benefits and costs on social behavior through the lens of selective engagement theory (SET; Hess, 2014). This perspective argues that normative declines in the personal resources (e.g., cognitive ability, physiological structures) that support behavior increase the demands – or costs – of engagement. These increased costs make older adults more selective about engagement, reducing their general motivation to expend resources and increasing their sensitivity to the benefits (e.g., personal implications) of a given task. Much of the support for this theory comes from work examining performance on cognitively demanding tasks within the laboratory and everyday life (e.g., Ennis, Hess, & Smith, 2013; Hess, Growney, O’Brien, Neupert, & Sherwood, 2018; Hess, Smith, & Sharifian, 2016; Queen & Hess, 2018; Smith & Hess, 2015). One could consider social interactions in similar terms. Specifically, maintaining and investing in social networks not only comes with specific benefits (e.g., preservation of important relationships), but also puts demands on both physical and cognitive resources, potentially increasing the costs of social interaction in later life. Such effects might have particular consequences in certain situations (e.g., maintaining conversations with unfamiliar others, interacting with negative social partners). Thus, SET may offer a different, albeit complementary, account of aging-related variation in social relationships by suggesting that changes in the motivation

English and Carstensen (2014) examined emotional tone across different relationships within the inner, middle, and outer circles of the network, but not by the type of relationship (e.g., friends, family). The finding by Lansford and colleagues (1998) concerns satisfaction with the number of friends in participants’ networks and does not take into account (a) the actual network size (i.e., the number of relationships reported within and across each level), (2) the placement of friends in the network, or (3) the nonfriends in the network (i.e., family members). The satisfaction finding, however, still supports SST-based arguments.

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 42–51

or willingness to engage resources partly explain the increased selectivity in social networks observed in later life. Indeed, research using both longitudinal (Growney & Hess, 2019) and cross-sectional (Queen & Hess, 2018) data has provided evidence consistent with this motivational perspective. The present study tested these ideas for the first time in the social domain and outside of a traditional experimental setting, specifically by examining (1) how perceived benefits and costs influence selectivity in social relationships and (2) the degree to which age and nature of relationship moderate these influences. We asked young and older adults from the community to identify real-life social partners at various levels of centrality within their social network. We then asked them to evaluate a number of factors associated with each partner, including the benefits and costs involved in interactions with that person and their willingness to engage in them. Based on the SET framework, we generally expected that the centrality of a partner to an individual’s social network would positively relate to benefits and negatively relate to costs (H1). In other words, the perceived benefits and costs of interacting with a person could partially explain that individual’s position within a social network. We also examined affective attributes associated with social partners. Given findings that suggest more positivity in central versus peripheral relationships (English & Carstensen, 2014), we tested the general prediction derived from SST that suggests that centrality should relate positively to positive affect and negatively to negative affect, with these associations being stronger in the older relative to the young group (H2). We further hypothesized, however, that benefits and costs would respectively subsume the positive and negative affective outcomes associated with interaction partners. In other words, affect would function as one of several attributes that contributes to perceived benefits and costs (H3). Of primary interest was the impact of these partnerspecific interaction characteristics on participants’ willingness to expend the effort necessary to engage with those partners. We predicted that perceived benefits would have a positive impact (H4) on willingness, whereas perceived costs would have a negative impact (H5). The nature of the relationship and age of participant was also expected to further moderate the impact of costs. Specifically, and consistent with SET, we predicted a disproportionate reduction in the negative impact of costs in older adults for those relationships with high perceived benefits (H6). That is, normal age-related increases in the costs (e.g., cognitive effort, fatigue) associated with social engagement should make older adults more discriminating and increase their reluctance to engage in interactions high in perceived costs unless they are also perceived as high in benefits.

GeroPsych (2020), 33(1), 42–51

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

Methods Participants Using online and newspaper advertisements, we recruited 32 younger (18 women; Mage = 31.91, range = 20–40) and 38 older (19 women; Mage = 74.53, range = 65–85) adults from the greater Raleigh, North Carolina, community to participate in an online study in exchange for a $20 Amazon.com gift card. Participants had on average about 16 years of education, equivalent to a college degree. Additionally, as expected, young adults reported better physical health (M = 48.33, SD = 5.95) and worse mental health (M = 46.41, SD = 12.42) compared to older adults (physical: M = 45.41, SD = 6.16; mental: M = 55.76, SD = 7.80), as measured by the SF-36 (Ware, 1993). This research project was reviewed and approved by the North Carolina State University IRB.

Materials and Procedure We administered the study entirely online. Participants received a web link to an online survey that contained the informed consent form and all of the tasks and questionnaires affiliated with the study. Once participants had consented, they completed a social network mapping task (Antonucci, 1986). In this task, we asked them to think about people who were important to them at the time they were participating in the study, noting that these people could come from various areas of their lives (e.g., neighbors, friends, family members as well as individuals from social and recreational clubs, religious groups, work or professional organizations, and community programs). We then presented a diagram that showed three concentric circles surrounding a smaller center circle with the word “Me” written in the middle. Participants were instructed to list up to six people in each of the circles, with each circle reflecting the degree of closeness between the participant and network member. They placed people in the innermost circle to whom they felt very close such that they found it hard to imagine life without them. People to whom they did not feel quite as close but who were still very important to them were placed in the middle circle, whereas those to whom they felt the least close but who were important enough to include in their personal networks were placed in the outermost circle. We encouraged participants to simply list people within each circle on a separate piece of paper. To preserve confidentiality, we did not collect the lists participants generated. After creating their diagrams, participants assigned each identified person a number and then considered him or her individually in order to rate: (1) the amount of effort

Ó 2020 Hogrefe

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

involved in interacting with the social partner (i.e., costs); (2) the amount of enjoyment or satisfaction they typically experience from an interaction (i.e., benefits); (3) the positivity and negativity of these interactions; (4) the importance associated with maintaining the relationship (i.e., importance); and, (5) their willingness to initiate an interaction (i.e., willingness). They provided each rating on a 7point scale, which ranged from 1 (not at all) to 7 (a lot). Note that the definition of costs and benefits is specific to SET. For example, costs were defined in terms of effort as opposed to other types of costs (e.g., temporal, financial) that might also influence social engagement. Finally, participants also specified their (1) relationship to the network member (spouse or significant other, relative, friend, acquaintance, or other), (2) frequency of contact with the member, and (3) degree of preference to interact with that person over all other members. Responses for these latter two items ranged on a scale from 1 (rarely or not at all, respectively) to 7 (very often or a lot, respectively). We presented all items on the computer, repeating them for each individual identified by the participant until he or she exhausted their list of network members. Following these two tasks, participants completed the demographic and background assessments. They were then presented with information that fully debriefed them about study goals and procedures and explained how they would receive the promised compensation.

Results General Composition of Network First, we compared the general composition of the social networks generated by younger and older adults to determine whether participants populated the levels of centrality in a manner that made sense based on both our instructions and the literature. We examined the frequency with which members in each level of the personal network (i.e., inner, middle, outer circles) were labeled as either partners (i.e., spouses, significant others), family members, friends, or acquaintances. (Note that data have a nested structured, with network member characteristics nested within participants.) As shown in Table 1, partners and family members were overrepresented by both age groups in the inner and middle circles, whereas friends and acquaintances were found more frequently in the middle and outer circles. In addition, family made up a larger percentage and friends a smaller percentage of members 2 3

Table 1. Relationship of network members: N (% within level) Centrality

Partner

Family

Friend

Other

Total

Younger adults Inner

21 (12.1)

108 (62.1)

45 (25.9)

0 (0.0)

174

Middle

2 (1.2)

52 (31.3)

104 (62.7)

8 (4.8)

166

Outer

1 (0.7)

38 (25.5)

77 (51.7)

33 (22.1)

149

Total

24 (4.9)

198 (40.5)

226 (46.2)

41 (8.4)

489

32 (13.7)

165 (70.8)

36 (15.5)

0 (0.0)

233

0 (0.0)

92 (43.0)

104 (48.6)

18 (8.4)

214

Older adults Inner Middle Outer

0 (0.0)

37 (19.4)

115 (60.2)

39 (20.4)

191

Total

32 (5.0)

294 (46.1)

255 (40.0)

57 (8.9)

638

in the inner and middle circles of older adults’ compared to younger adults’ networks. Across all levels, older adults also listed more family members and fewer friends than did younger adults.

Characteristics of Relationships with Network Partners We next explored the perceived characteristics of individuals at each level of participants’ social networks using multilevel modeling (MLM), treating ratings of each social partner identified by each participant as observations at Level 1 (within-person). We regressed these ratings onto the Level 1 factor of centrality ( 1 = outer circle, 0 = middle circle, 1 = inner circle),2 and the Level 2 factor of age, with the young adults serving as the reference group. Level 1 ratings of romantic partners were excluded from these analyses given that such relationships likely differ in meaningful ways compared to other types.3 Level 1 factors were also grand mean centered and slopes for main effects were allowed to vary freely. Central to the SET perspective, we first focused on the perceived benefits and costs associated with social interchanges. For perceived benefits, we observed a positive main effect of centrality, b = 0.60, t(999) = 5.17, p < .0001, but no effect of age (ps > .25). For costs, in contrast, we found a marginal, negative main effect of centrality, b = 0.28, t(999) = 1.93, p = .054, and a significant Centrality Age interaction, b = .43, t(999) = 2.23, p = .03. The interaction indicated – somewhat unexpectedly – that costs did not vary significantly over levels of centrality in the older group, b = 0.15, p = .23. Thus, we obtained support for H1, with the caveat that age moderated the impact of costs. We next examined age differences in ratings of typical affect associated with social partners and found effects only

Dummy coding centrality resulted in similar findings. Thus, for simplicity, we treated centrality as a continuous variable. Excluding romantic partners did not significantly change the strength or pattern of results. However, we retained the models with romantic partners excluded to remain conservative in our approach.

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 42–51

partially consistent with expectations relevant to H2. Centrality was positively associated with positive affect, b = 0.49, t(999) = 4.68, p < .0001, but was unrelated to negative affect, b = 0.01, p = .92. A main effect of age, however, indicated that older adults had lower ratings of negative affect compared to younger adults (a = 1.76 vs. 2.30, respectively), b = 0.54, t(68) = 3.19, p = .002. Importantly, significant interactions between centrality and age were not observed for either affective rating. This finding seemingly contradicts SST-based expectations that older adults spend more time engaged in emotionally “rewarding” (i.e., positive) experiences (for a review, Carstensen, Gross, & Fung, 1997), which presumably could have a disproportionate impact on the preference for close interaction partners in old age (but see Carstensen, Pasupathi, Mayr, & Nesselroad, 2000; English & Carstensen, 2014).4 We further examined whether the effects of age and centrality on affect could be explained in terms of benefits and costs. To do this, we re-ran the two just-described models but included benefits and costs as additional predictors of positive and negative affect, respectively. We then ran two converse models that predicted benefits and costs using positive and negative affect as covariates, respectively. The model predicting positive affect revealed a positive association with benefits when it was included as a covariate, b = 0.69, t(998) = 23.55, p < .0001, but the previously observed centrality effect was no longer significant (p = .39). The converse analysis indicated that positive affect was also a significant predictor of benefits, b = 0.82, t(998) = 27.41, p < .0001, but that the effect of centrality remained a significant predictor when controlling for positive affect, b = 0.17, t(998) = 2.84, p = .005. Similarly, when costs was included as a covariate in predicting negative affect, a significant positive association was obtained, b = 0.23, t(998) = 6.76, p < .0001, but the previously observed age effect, though significant, was weakened, b = 0.38, t(68) = 2.87, p = .005. Conversely, both negative affect, b = 0.48, t(998) = 7.75, p < .0001, and centrality, b = 0.25, t(998) = 2.14, p = .03, emerged as significant predictors of costs. These results support H3, suggesting that our assessment of benefits and costs represents a more general assessment of the affective attributes associated with interaction partners, accounting for positive and negative emotional responses as well as other factors.

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

In sum, the network characteristics across groups were very similar in terms of the types and qualities of people that made up the three levels of centrality. There were only minor age differences in the distribution of partners based on network position (e.g., the aforementioned difference in the allotment of family and friends to the inner and middle circles). The positive aspects (i.e., benefits, positive affect) of interaction partners also played a similar role in network assignments for both groups. Surprisingly, however, the costs of interactions associated with network partners had less to do with centrality in the older group than in the young group, perhaps relating to the overall lower level of negative affect among partners in the older adults’ networks. Importantly, these analyses also suggest that, whereas affective characteristics of interactions may differentiate individuals at different levels of one’s social networks, a more general assessment of benefits and costs may subsume such characteristics. Taken together, these results indicate that network member characteristics varied in meaningful ways as a function of network placement and, in some cases, age.

Prediction of Interaction Willingness Our main focus was on examining the predictors of an individual’s willingness to exert the necessary effort to interact with each social partner. In our first set of analyses, we once again employed MLM to assess the impact of age group as a Level 2 factor as well as perceived benefits and costs as Level 1 factors. Models also included at Level 1 (1) meancentered importance ratings, under the assumption that they might capture the degree to which interactions with each partner aligns with personal goals, and (2) centrality to account for factors unrelated to motivation (e.g., contact frequency) that might artificially inflate willingness scores.5 Again, all Level 1 main effects were allowed to vary randomly. Finally, we decided to exclude ratings associated with romantic partners (i.e., spouses, significant others) because of the potentially more obligatory nature of this type of relationship compared to other types. Table 2 presents the results for these analyses. Supportive of H4 and H5, importance and benefits positively predicted ratings of willingness, whereas costs

We acknowledge that, although often associated with a positivity bias in studies examining SST (Carstensen et al., 1997; Fredrickson & Carstensen, 1990; Reed & Carstensen, 2012), emotional meaningfulness does not necessarily predict or equate to positivity. For example, in one study, network-related positive affect did not predict positive emotionality (English & Carstensen, 2014). Supplemental analyses also indicated that centrality predicted both importance, b = 0.96, t(999) = 8.55, p < .0001, and contact frequency: b = 1.12, t(999) = 8.39, p < .0001), with the latter being weaker in older adults. We consequently included centrality as an additional control. Similarly, we did not examine preference or contact frequency as outcomes given potential interpretational issues. The latter may reflect opportunity or access as opposed to general willingness and the former may not take into account perceived obligation to interact with difficult members in order to maintain important relationships.

GeroPsych (2020), 33(1), 42–51

Ó 2020 Hogrefe

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

Table 2. MLM results for prediction of willingness from ratings of benefits and costs Effect

p <.0001

Intercept

5.3978

0.0983

54.90

Benefits

0.5599

0.0466

986

12.00

<.0001

Costs

0.1082

0.0392

986

2.76

0.0060

Importance

0.3119

0.0498

986

6.26

<.0001

Age group

0.3776

0.1277

2.96

0.0043

Benefits Costs

0.0337

0.0184

986

1.84

0.0666

Benefits Importance

0.0301

0.0242

986

1.25

0.2131

Benefits Age group

0.3003

0.0589

986

5.10

<.0001

Costs Importance

0.0118

0.0185

986

0.64

0.5237

Costs Age group

0.0327

0.0534

986

0.61

0.5400

Importance Age group

0.0209

0.0658

986

0.32

0.7503

Benefits Costs Importance

0.0075

986

1.00

0.3166

Benefits Costs Age group

0.0742

0.0253

986

2.93

0.0035

Benefits Importance Age group

0.0397

0.0303

986

1.31

0.1898

Costs Importance Age Group

0.0369

0.0271

986

1.36

0.1734

Benefits Costs Importance Age group

0.0053

0.0103

986

0.51

0.6083

Centrality

0.1865

0.0489

986

3.81

0.0001

Note. This model explained 64% and 67% of the Level 2 and Level 1 variance, respectively.

negatively predicted them. In addition, the interactions between age, benefits, and costs suggest differences in the weightings assigned to these ratings across age groups. Follow-up analyses revealed a weaker impact of both costs and benefits on willingness in older adults (bs = 0.26 and 0.08, respectively) than in the young group. In addition, a significant Benefits Costs interaction specific to older adults, b = 0.04, t(986) = 2.32, p = .02, appeared to drive the 3-way interaction. To further explore this interaction, we examined the impact of costs at representative levels of benefits (±1 SD from the mean) within the older group. Consistent with expectations, levels of willingness were relatively high (a = 6.15) and unaffected by costs (b = 0.02, p = .68) at high levels of benefits in older adults, whereas overall levels were reduced (a = 5.40) and negatively associated with costs (b = .13, p = .003) when benefits were low. In other words, older adults were willing to engage resources in social interactions when the costs where high if they also perceived high benefits, supporting H6.6

In our next set of analyses, we substituted positive and negative affect ratings for benefits and costs to test alternative hypotheses based in SST, with specific interest in the extent to which affect differentially predicted willingness across age groups. As seen in Table 3, positive affect positively predicted willingness. In addition, we obtained significant Age Positive affect and Age Importance Positive affect Negative affect interactions. Comparisons across age groups revealed a stronger impact of positive affect in the young than in the old (bs = .43 vs. .25, respectively), which would again contrast with the SST-based prediction that positive affect more strongly determines social engagement in older adults. The four-way interaction did not appear to reflect any meaningful age-related variation. We subsequently re-ran this analysis using benefits and costs as covariates to see if the effects associated with affect would drop out. The effect of positive affect and its interaction with age remained significant, but the impact of positive affect was greatly reduced (b = .23, p = .0007).

We attempted to examine whether the nature of the relationship with the social partner made a difference in determining interaction willingness. We ran the same MLM but separately for family relationships, which are presumably more obligatory, and nonfamily ones, which are presumably less obligatory. The results generally did not provide consistent or systematic evidence in support of our primary findings or those based on the predominant theory, SST. In the family context, we found the same positive main effects and Benefits x Costs interaction. However, relationship importance mattered more for young adults’ willingness than for that of older adults, and costs had a stronger influence on willingness in relationships high in importance in comparison to those low in importance. Willingness levels still remained quite high for low importance relationships in the latter case. In addition, if true, the former would actually suggest that older adults exhibit less selectivity in family relationships than do younger adults, contradicting both SET and SST. In the nonfamily context, we also observed a stronger impact of benefits on willingness in younger adults, again inconsistent with both perspectives. We consequently exercise caution in considering or discussing these results further.

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 42–51

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

Table 3. MLM results for prediction of willingness from ratings of affect Effect

p <.0001

Intercept

5.2760

0.1116

47.27

0.4298

0.0631

986

6.81

<.0001

0.0274

0.0577

986

0.47

0.6351

Importance

0.4707

0.0409

986

11.50

<.0001

Age group

0.5204

0.1514

3.44

0.0010

PA NA

0.0530

0.0311

986

1.70

0.0888

PA Importance

0.0375

0.0255

986

1.47

0.1416

PA Age group

0.1766

0.0853

986

2.07

0.0387

NA Importance

0.0008

0.0252

986

0.03

0.9747

NA Age group

0.0002

0.0885

986

0.00

0.9983

Importance Age group

0.1683

0.0544

986

3.09

0.0020

PA NA Importance

0.0145

0.0098

986

1.47

0.1412

PA NA Age group

0.0316

0.0409

986

0.77

0.4408

PA Importance Age group

0.0024

0.0353

986

0.07

0.9466

NA Importance Age group

0.0271

0.0480

986

0.57

0.5719

PA NA Importance Age group

0.0352

0.0154

986

2.29

0.0221

Centrality

0.2015

0.0393

986

5.13

<.0001

Note. PA = Positive affect, NA = negative affect. The model without benefits and costs as covariates explained 59% and 54% of the Level 2 and Level 1 variance, respectively. Controlling for benefits and costs, an additional 4% of the Level 1 variance was explained. Only the positive affect and negative affect slopes were random in this model because of convergence issues.

In addition, the four-way interaction was no longer significant (p = .081). Thus, as suggested in our preliminary analyses (and supportive of H3), a more general assessment of benefits and costs appears to capture a considerable amount of the affective influence on willingness.

Discussion In the current study, we attempted to explore age-related variations in social engagement patterns in adulthood using SET (Hess, 2014) as a guiding framework. From this perspective, both costs – reflective of the burden placed on personal resources to interact – and benefits – reflective of the enjoyment and satisfaction derived from a relationship – should work together to determine the willingness to engage with different social partners. Greater benefits should result in higher levels of willingness, and costs should lead to lower levels, especially for older adults. The most pronounced age differences should also occur in the least important and central relationships, where the costs of interacting outweigh the relatively lower perceived benefits. Results based on data from a social-mapping and partnerrating task generally support our expectations in terms of how individuals conceive of social partners and how these conceptualizations affect willingness. Individuals of all ages viewed close relationships as less costly, more satisfying, and more important than less close ones. Somewhat GeroPsych (2020), 33(1), 42–51

surprisingly, however, compared to young adults, older adults attributed fewer costs to less important relationships as well as reported lower negative affect and similar levels of positive affect across all relationships. These findings do not necessarily contradict our initial expectations or those of SST; rather, they could simply highlight the selective nature of older adults’ networks. People tend to characterize their relationships as less problematic and ambivalent as they age (Fingerman, Hay, & Birditt, 2004), while at the same time report feeling increasingly more positive, less negative, and more satisfied about them (for review, see Luong, Charles, & Fingerman, 2011). Perhaps as a function of the “social pruning process,” people may also become more discriminating in terms of choosing partners who place fewer demands on their personal resources. An alternative explanation would also suggest that the accumulation of social expertise (e.g., Hess, Osowski, & Leclerc, 2005) may make older adults more adept than young adults at identifying problematic partners, avoiding confrontation, and regulating their socioemotional experiences. For instance, studies showed that, although older adults prefer disengagement strategies (Birditt, Fingerman, & Almeida, 2005), these strategies can actually facilitate the management of interpersonal problems (BlanchardFields, Mienaltowski, & Seay, 2007). In other words, people build a repertoire of strategies throughout their lifetimes on which they can draw in order to effectively navigate difficult social situations, which can subsequently lead to the perception of fewer costs. More specifically in this case, Ó 2020 Hogrefe

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

despite potentially greater costs associated with interacting with unfamiliar and negative partners in older age, individuals may develop strategies that minimize the impact of these costs. In sum, the lower levels of costs and negative affect associated with less central members in the older group may reflect the effects of prior selection via social pruning, an age-related increase in the ability to navigate difficult interactions and relationships, or a combination of both. We also found support for SET, in that low costs and high benefits resulted in greater willingness, with the effects of centrality and affect largely captured by these two factors. The fact that each of these effects were generally weaker in older adults may again reflect prior selection effects, whereby partners selected into older adults’ networks vary less in terms of their demandingness and satisfaction ratings. Nevertheless, older adults showed the expected reduction in willingness in relationships with low benefits and high costs. These results are, to our knowledge, the first to extend SET into the social domain. They also highlight how aging-related adaptive processes are tethered to the pursuit of social goals. The costs associated with enduring negative and cumbersome relationships may lead to disengagement in situations where they outweigh the potential rewards, especially later in life because of limitations in personal resources. These costs may be overcome, for example, through the use of certain strategies (e.g., conflict resolution), if the relationships are salient and important parts of one’s social network and promote the attainment of personal goals. Taken together, the findings from the current study, reflective of social selectivity in aging, are perhaps best understood through a lens that combines two complementary perspectives, SST and SET. Both theories overlap in their general account of the maintenance of everyday functioning and emotional well-being in later life: As we age, we become more discerning about our engagements because of age-related reductions in cognitive ability (SET) or perception of future time (SST), redistributing our resources and making determinations about engagement based on the (emotional) meaningfulness or – more broadly – the benefits. Much research supports these ideas from an SST approach (e.g., English & Carstensen, 2014; Fredrickson & Carstensen, 1990; Fung et al., 2001), finding that older adults prefer familiar partners whom they also tend to rate positively and who allow them to feel emotionally fulfilled. However, at least two questions remained unanswered: What factors besides the affective attributes of social partners influence selective engagement? And under what conditions can we observe more or 7

less selectivity? In the current study, we attempted to answer these two questions and build on existing work by using SET as a guide since it allowed us to make specific predictions about the impacts of two factors on social engagement, the costs and benefits associated with social interaction. Our results offer new insights to suggest that affect makes up part of a broader set of factors that influences engagement.

Limitations Several caveats should be considered when interpreting these findings. First, this study represents an initial attempt to examine SET in the social domain, with observed effects based on a relatively small sample.7 Future research should aim to replicate these findings in a larger and more representative sample. Second, the cross-sectional nature of our design does not allow us to make strong conclusions about how conceptualizations of social partners and their impacts on willingness change over time, and whether these changes influence the observed reductions in networks in later life. Second, we did not measure the total size of participants’ social networks, which precludes us from determining how network size varies systematically with age. This information could help contextualize the agerelated reduction in perceived costs. Smaller compared to larger social networks might correspond to a reduction both in general and relationship-specific costs associated with social engagement. Third, by exploring the feasibility of SET to explain agerelated variation in social selectivity, we intentionally employed broad measures of costs and benefits. This provided fruitful evidence to suggest that factors such as affect, centrality, and importance make up part of a broader, more general construct associated with benefits and costs. Future studies, however, could identify other characteristics reflective of benefits and examine their contribution to network assignments and social willingness. For instance, early work has implicated shared interests as one trait underlying adults’ characterizations of potential social partners (Fredrickson & Carstensen, 1990). The extent of shared interests may accordingly contribute to the benefits of engagement in unique ways. Consideration of other types of benefits that capture functional characteristics of individuals’ networks may also contribute to a more nuanced understanding of selectivity in the social setting. Relationships with similar levels of benefits but that serve different functions (e.g., provision versus receipt of social support) may have different implications for willingness

Although the field has yet to reach a consensus on determining power in complicated MLM contexts, recent simulation studies provide useful guides for future power analyses (Arend & Schäfer, 2019; Mathieu, Aguinis, Culpepper, & Chen, 2012).

Ó 2020 Hogrefe

GeroPsych (2020), 33(1), 42–51

to engage. Future work could similarly inform the current perspective by refining the assessments of costs to address more specifically those associated with physical demands, cognitive demands, and psychological demands. Finally, an examination of how physical or mental ability, in addition to social partner characteristics, influences willingness to engage in social interactions could further offer meaningful insights.

Conclusions The extent to which individuals selectively attend to some relationships over others can depend on shifts in social goals. Our findings indicate that an examination of the balance between potential benefits and demands placed on personal resources may also further our understanding of age differences as well as changes in social relationships and interactions over time. Future research should aim to further elucidate these findings by identifying the specific benefit- and cost-related factors that matter most in determining engagement as well as the degree to which the emphasis placed on these factors changes across the lifespan.

References Antonucci, T. C. (1986). Hierarchical mapping technique. Generations: Journal of the American Society on Aging, 10, 10–12. Arend, M. G., & Schäfer, T. (2019). Statistical power in two-level models: A tutorial based on Monte Carlo simulation. Psychological Methods, 24, 1–19. https://doi.org/10.1037/met0000195 Birditt, K. S., Fingerman, K. L., & Almeida, D. (2005). Age and gender differences in reported reactions to interpersonal tensions: A daily diary study. Psychology and Aging, 20, 330–340. https://doi.org/10.1037/0882-7974.20.2.330 Blanchard-Fields, F., Mienaltowski, A., & Seay, R. B. (2007). Age differences in everyday problem-solving effectiveness: Older adults select more effective strategies for interpersonal problems. Journal of Gerontology: Psychological Sciences, 62B, P61–P64. https://doi.org/10.1093/geronb/62.1.p61 Carstensen, L. L., Fung, H. H., & Charles, S. T. (2003). Socioemotional selectivity theory and the regulation of emotion in the second half of life. Motivation and Emotion, 27, 103–123. https://doi.org/10.1023/A:1024569803230 Carstensen, L. L., Gross, J., & Fung, H. H. (1997). The social context of emotional experience. In K. W. Shaie & M. P. Lawton (Eds.), Annual review of gerontology and geriatrics: Focus on emotion and adult development (Vol. 17, pp. 325–352). New York, NY: Springer. Carstensen, L. L., Isaacowitz, D. M., & Charlels, S. T. (1999). Taking time seriously: A theory of socioemotional selectivity. American Psychologist, 54, 165–181. https://doi.org/10.1037/ 0003-066X.54.3.165 Carstensen, L. L., Pasupathi, M., Mayr, U., & Nesselroad, J. R. (2000). Emotional experience in everyday life across the adult life span. Journal of Personality and Social Psychology, 79, 644– 655. https://doi.org/10.1037/0022-3514.79.4.644

GeroPsych (2020), 33(1), 42–51

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

English, T., & Carstensen, L. L. (2014). Selective narrowing of social networks across adulthood is associated with improved emotional experience in daily life. International Journal of Behavioral Development, 38, 195–202. https://doi.org/10.1177/ 0165025413515404 Ennis, G. E., Hess, T. M., & Smith, B. T. (2013). The impact of age and motivation on cognitive effort: Implications for cognitive engagement in older adulthood. Psychology and Aging, 28, 495–504. https://doi.org/10.1037/a0031255 Fingerman, K. L., Hay, E. L., & Birditt, K. S. (2004). The best of ties, the worst of ties: Close, problematic, and ambivalent social relationships. Journal of Marriage and Family, 66, 792–808. https://doi.org/10.1111/j.0022-2445.2004.00053.x Fredrickson, B. L., & Carstensen, L. L. (1990). Choosing social partners: How old age and anticipated endings make people more selective. Psychology and Aging, 5, 335–347. https://doi. org/10.1037/0882-7974.5.3.335 Fuller-Iglesias, H. R., Webster, N. J., & Antonucci, T. C. (2015). The complex nature of family supports across the lifespan: Implications for psychological well-being. Developmental Psychology, 51, 277–288. https://doi.org/10.1037/a0038665 Fung, H. H., Carstensen, L. L., & Lang, F. R. (2001). Age-related patterns in social networks among European Americans and African Americans: Implications for socioemotional selectivity across the lifespan. International Journal of Aging and Human Development, 52, 185–206. https://doi.org/10.2190/1ABL-9BE5M0X2-LR9V Fung, H. H., Carstensen, L. L., & Lutz, A. M. (1999). Influence of time on social preferences: Implications for life-span development. Psychology and Aging, 14, 595–604. https://doi.org/ 10.1037/0882-7974.14.4.595 Growney, C. M., & Hess, T. M. (2019). Longitudinal relationships between resources, motivation, and cognitive engagement. Manuscript submitted for publication. Hess, T. M. (2014). Selective engagement of cognitive resources: Motivational influences on older adults’ cognitive functioning. Perspectives on Psychological Science, 9, 388–407. https://doi. org/10.1177/1745691614527465 Hess, T. M., Growney, C. M., O’Brien, E. L., Neupert, S. D., & Sherwood, A. (2018). The role of cognitive costs, attitudes about aging, and intrinsic motivation in predicting engagement in demanding and passive everyday activities. Psychology and Aging, 33, 953–964. https://doi.org/10.1037/pag0000289 Hess, T. M., Osowski, N. L., & Leclerc, C. M. (2005). Age differences in sensitivity to diagnostic cues and the flexibility of social judgments. Psychology and Aging, 20, 447–459. https:// doi.org/10.1037/0882-7974.20.3.447 Hess, T. M., Smith, B. T., & Sharifian, N. (2016). Aging and effort expenditure: The impact of subjective perceptions of task demands. Psychology and Aging, 31, 653–660. https://doi.org/ 10.1037/pag0000127 Huxhold, O., Fiori, K. L., & Windsor, T. D. (2013). The dynamic interplay of social network characteristics, subjective wellbeing, and health: The costs and benefits of socio-emotional selectivity. Psychology and Aging, 28, 3–16. https://doi.org/ 10.1037/a0030170 Lang, F. R., & Carstensen, L. L. (1994). Close emotional relationships in later life: Further support for proactive aging in the social domain. Psychology and Aging, 9, 315–324. https://doi. org/10.1037/0882-7974.9.2.315 Lang, F. R., Rieckmann, N., & Baltes, M. M. (2002). Adapting to aging losses: Do resources facilitate strategies of selection, compensation, and optimization in everyday functioning? Journal of Gerontology: Psychological Sciences, 57B, 501–509. https://doi.org/10.1093/geronb/57.6.P501

Ó 2020 Hogrefe

E. L. O’Brien & T. M. Hess, Selective Social Engagement in Adulthood

Lang, F. R., Wagner, J., Wrzus, C., & Neyer, F. J. (2013). Personal effort in social relationships across adulthood. Psychology and Aging, 28, 529–539. https://doi.org/10.1037/a0032221 Lansford, J. E., Sherman, A. M., & Antonucci, T. C. (1998). Satisfaction with social networks: An examination of socioemotional selectivity theory. Psychology and Aging, 13, 544–552. https:// doi.org/10.1037/0882-7974.13.4.544 Li, M., & Dong, X. (2018). Is social network a protective factor for cognitive impairment in US Chinese older adults? Findings from the PINE Study. Gerontology, 64, 246–256. https://doi.org/ 10.1159/000485616 Litwin, H. (2003). Social predictors of physical activity in later life: The contribution of social-network type. Journal of Aging & Physical Activity, 11, 389–406. https://doi.org/10.1123/japa.11. 3.389 Luong, G., Charles, S. T., & Fingerman, K. L. (2011). Better with age: Social relationships across adulthood. Journal of Social & Personal Relationships, 28, 9–23. https://doi.org/10.1177/ 0265407510391362 Mathieu, J. E., Aguinis, H., Culpepper, S. A., & Chen, G. (2012). Understanding and estimating the power to detect cross-level interaction effects in multilevel modeling. Journal of Applied Psychology, 97, 951–966. https://doi.org/10.1037/a0028380 Queen, T. L., & Hess, T. M. (2018). Linkages between resources, motivation, and engagement in everyday activities. Motivation Science, 4, 26–38. https://doi.org/10.1037/mot0000061 Reed, A. E., & Carstensen, L. L. (2012). The theory behind the agerelated positivity effect. Frontiers in Psychology, 3, 1–9. https:// doi.org/10.3389/fpsyg.2012.00339 Rook, K. S., Luong, G., Sorkin, D. H., Newsom, J. T., & Krause, N. (2012). Ambivalent versus problematic social ties: Implications for psychological health, functional health, and interpersonal coping. Psychology and Aging, 27, 912–923. https://doi.org/ 10.1037/a0029246 Sharifian, N., Manly, J., Brickman, A. M., & Zahodne, L. B. (2019). Social network characteristics and cognitive functioning in ethnically diverse older adults: The role of network size and composition. Neuropsychology, 33, 956–963. https://doi.org/ 10.1037/neu0000564 Shiovitz-Ezra, S., & Litwin, H. (2012). Social network type and health-related behaviors: Evidence from an American national survey. Social Science & Medicine, 75, 901–904. https://doi. org/10.1016/j.socscimed.2012.04.031

Ó 2020 Hogrefe

Smith, B. T., & Hess, T. M. (2015). The impact of motivation and task difficulty on resource engagement: Differential influences on cardiovascular responses of young and older adults. Motivation Science, 1, 22–36. https://doi.org/10.1037/mot0000012 Ware, J. E. Jr. (1993). SF-36 health survey. Boston, MA: The Health Institute, New England Medical Center. Wrzus, C., Hänel, M., Wagner, J., & Neyer, F. J. (2013). Social network changes and life events across the life span: A metaanalysis. Psychological Bulletin, 139(1), 53–80. https://doi.org/ 10.1037/a0028601 History Received July 29, 2019 Accepted November 29, 2019 Acknowledgments The research procedures, including data collection, and preparation of this manuscript were supported by National Institute on Aging grant R01 AG 05552 awarded to Thomas M. Hess. Preparation of the manuscript was also partially supported by National Institute on Aging grant T32 AG049676 awarded to The Pennsylvania State University (PI: David Almeida). Conflict of Interest The authors declare no conflict of interest. Author Note Preliminary findings from this manuscript were presented in 2018 at the Cognitive Aging Conference in Atlanta, GA. ORCID Erica L. O’Brien https://orcid.org/0000-0002-4863-9159 Erica L. O’Brien Center for Healthy Aging The Pennsylvania State University 428 Biobehavioral Health Building University Park, PA 16802 USA eml5781@psu.edu

GeroPsych (2020), 33(1), 42–51

A concise guide to the assessment and treatment of insomnia for busy professionals “This book provides an excellent concise introduction to insomnia and its treatment with cognitive behavioral therapy. A great addition to any therapist’s bookshelf!” Philip Gehrman, PhD, CBSM, Associate Professor, Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA

William K. Wohlgemuth / Ana Imia Fins

Insomnia (Series: Advances in Psychotherapy – Evidence-Based Practice – Volume 42) 2019, viii + 94 pp. US $29.80 / € 24.95 ISBN 978-0-88937-415-7 Also available as eBook About 40% of the population experiences difficulty falling or staying asleep at some time in a given year, while 10% of people suffer chronic insomnia. This concise reference written by leading experts for busy clinicians provides practical and upto-date advice on current approaches to assessment, diagnosis, and treatment of insomnia. Professionals and students learn to correctly identify and diagnose insomnia and gain hands-on information on how to

www.hogrefe.com

carry out treatment with the best evidence base: cognitive behavioral therapy for insomnia (CBT-I). The American Academy of Sleep Medicine (AASM) and the American College of Physicians (ACP) both recognize CBT-I as the first-line treatment approach to insomnia. Appendices include useful resources for the assessment and treatment of insomnia, which readers can copy and use in their clinical practice.