22 Pages


Course Number: PSY 394, Fall 2009

College/University: University of Texas

Word Count: 21403


Document Preview

Psychological Bulletin 2001, Vol. 127, No. 5, 651-672 Copyright 2001 by the American Psychological Association, Inc. OO33-29O9/O1/S5.OO DOI: 10.1037//0O33-29O9.127.5.651 The Social Construction of the Personal Past and Its Implications for Adult Development Monisha Pasupathi University of Utah This article examines conversational recounting about experiences as a potential mechanism by which people socially...

Unformatted Document Excerpt
Coursehero >> Texas >> University of Texas >> PSY 394

Course Hero has millions of student submitted documents similar to the one
below including study guides, practice problems, reference materials, practice exams, textbook help and tutor support.

Course Hero has millions of student submitted documents similar to the one below including study guides, practice problems, reference materials, practice exams, textbook help and tutor support.

Bulletin Psychological 2001, Vol. 127, No. 5, 651-672 Copyright 2001 by the American Psychological Association, Inc. OO33-29O9/O1/S5.OO DOI: 10.1037//0O33-29O9.127.5.651 The Social Construction of the Personal Past and Its Implications for Adult Development Monisha Pasupathi University of Utah This article examines conversational recounting about experiences as a potential mechanism by which people socially construct themselves and their worlds over the life span and the resulting implications for understanding adult development. Two principles governing conversational recounting of past events are proposed: coconstruction (the joint influences of speakers and contexts on conversational reconstructions of past events) and consistency (the influence of a conversational reconstruction on subsequent memory). Operating together, the principles provide an account for how autobiographical memory is socially constructed. In addition, the principles may illuminate how conversations about the past can influence the development of identity in adulthood. Conversation is to be thought of as creating a social world just as causality generates a physical one. (Harr6, 1983, p. 65). We live our lives immersed in talk, providing others with stories of what happens to us and ideas about what we think our experiences mean. We all remember times when a conversation made a big difference in how we understood an experience, where our listener provided new insights or perspectives, or, in telling, we realized something about ourselves. We may be less aware of more subtle influences of social context on our everyday talk about experiences or of the ways that conversations we have about our experiences shape our memory for our own past (see Vorauer & Miller, 1997). In this article, I explore conversations about experiences as a way in which we socially construct our past and our identities. On a metatheoretical level, my goal is to accommodate both social constructivist and more traditional approaches within psychology while developing an account of autobiographical remembering and adult development. The resulting framework is rooted in research on memory and language (H. H. Clark, 1996; Engel, 1999; Fivush, 1991; Neisser & Fivush, 1994); social cognitive, social constructivist, and discursive approaches to cognition (P. B. Baltes & Staudinger, 1996; Edwards, 1997; Schank & Abelson, 1995; Van Langenhove & Harre, 1993; Weldon, in press), and narrative approaches to the study of personality (Bruner, 1990; McAdams, 1996; Thorne, 2000). It complements recent trends in life span developmental research, which focus on the social contexts in which development unfolds over adulthood (see, e.g., M. M. Baltes & Carstensen, 1998; Berg, Meegan, & Deviney, 1998; Edwards, 1997; Gould & Dixon, 1993; Hess & BlanchardFields, 1999; Staudinger & Baltes, 1996; Thorne, 2000). In this article, I bring together insights from these different areas in considering the social construction of adult development through talk about the past. The model I present is based on the notion that much learning and development begins in social practices and gradually becomes represented internally (Edwards & Middleton, 1986a; Fivush, 1991; Rogoff, 1990; Vygotsky, 1978), and I focus on the social practice of recounting past events in conversation. Briefly stated, people tell events to their listeners in ways they hope will engage and interest them, and listeners contribute their own insights and reactions to the retelling. Telling is a joint product of the speaker and the audience and thus influences the way we subsequently remember the told event. This means that the responses of audiences have long-term implications for how tellers remember their experiences. Such implications include the idea that conversational remembering is one process by which people's social worlds influence their development in adulthood by shaping both what they remember and how they think of themselves. Consider a concrete example. Thomas tries out technical rock climbing for the first time. He enjoys it to some extent but finds it difficult, exhausting, and somewhat anxiety provoking. In telling Matthias, he represents the experience as exciting (justifying telling Matthias) and the activity as one that interests him and perhaps downplays his anxiety. Suppose Matthias supports this version of events by showing interest and enthusiasm. According to the model I will propose below, Thomas will remember his experience with greater detail than if he had not told it, will remember it as more positive because of the way he told it, and will be more likely to think of himself as being interested in and enjoying the activity of rock climbing. His conversation with Matthias thus shapes his 651 This project benefited from the support of multiple institutions and individuals, including fellowship support from the American Psychological Association, the Andrus Foundation, and the Center for Lifespan Psychology at the Max Planck Society for Human Development. Acknowledgments are due to Lisa M. Stallworth, Cindy Berg, Susan Bluck, Laura L. Carstensen, Susan Charles, Frank Drews, Paul Florsheim, Alan Fogel, Alexandra Freund, Tilman Habermas, Quinn Kennedy, Sarah Lucas, Elizabeth Marsh, Mara Mather, Fred Rhodewalt, Carol Sansone, Cecilia Wainryb, and the rest of the Carstensen Lab, who contributed in various ways to the construction of this article. Any errors and flaws, are, of course, my own responsibility. Correspondence concerning this article should be addressed to Monisha Pasupathi, Department of Psychology, University of Utah, 390 South 1530 East BEH-S 502, Salt Lake City, Utah 84112. Electronic mail may be sent to 652 PASUPATHI memory of his introduction to rock climbing and, perhaps, the likelihood that he will climb in the future. In elaborating the model, I first examine evidence for the coconstruction of conversational event memories with listeners and evidence that such coconstructions influence later memory. I then consider adult development, beginning with the idea that conversations about the past can influence the development of identity. However, because conversational remembering over adulthood occurs in the context of other developmental changes, I also review how those other changes might modify conversational recounting. Finally, I compare the proposed model with other current models of memory and consider its utility in explaining existing phenomena and generating new research questions. First, however, let me be more specific about conversational recollections of the past. What Counts as Conversations About the Past? My focus is restricted to conversations with other individuals about past events and does not include internal or imagined dialogues. By past events, I refer to personal memory (Brewer, 1986), which includes both specific temporally bounded events like a car accident or a high school graduation ceremony as well as more temporally extended events, such as making the decision to pursue a particular career or going to high school (Linde, 1993). I include both significant and mundane occurrences. I use the term past subjectively, in that something is past if the experiencer views it as having just happened but not if it is viewed as ongoing. Not all talk is talk about the past. People also talk about the future, about repeated events ("every time we visit grandma"), about other people ("he is such a jerk"), and despite all good advice to the contrary, about politics, religion, and sex. Talk of past events is often embedded within these other conversational contexts. Thus, talk about past experiences is a specific sort of talk, albeit one that may occur in the context of other activities. Talk about the past is also not equivalent to narratives about the past, although such talk is often studied by looking at narratives about past events in conversation (Gergen & Gergen, 1988; Labov & Waletsky, 1967; Manier & Hirst, 1996; Polanyi, 1985). Such narratives are viewed as having particular structural properties such as chronological order, narrative structure, goal-action sequences, and patterns of rising and falling action (Labov & Waletsky, 1967; McCabe & Peterson, 1984; Polanyi, 1985). Moreover, narratives typically include both factual details that could be confirmed by someone also present at the event and interpretive details that reflect an individual's subjective experience (Polanyi, 1985). It is notable that many conversations about past events do not meet any traditional structural criteria for narratives but would be labeled as stories by the participants (Ervin-Tripp & Kuentay, 1995). Thus, talk about the past is talk about personal memory for significant or mundane events that have happened in the subjective past, may take narrative or more abbreviated forms, and includes both facts and interpretations of those facts. 2000; Schacter, 1996). Strong versions of social constructivism imply that memory reconstructions are highly plastic (Davies & Harre, 1990) and provide little reason to believe that reconstructions would be similar across different remembering contexts. However, memory research shows that reconstructions of events become relatively stable after an initial period of malleability (Johnson & Chalfonte, 1994; Koutstaal, Schacter, Johnson, Angell, & Gross, 1998; McClelland, 1994). This stability means that the social construction of event memories in conversations could have developmental significance. Thus, recent approaches to memory imply two principles that govern conversational recounting: coconstruction and consistency. Below, I first examine how memories retold in conversation are jointly produced by speakers and listeners (coconstruction). I then review how such coconstructed memories may have long-term implications for the way events are subsequently recalled (consistency). The Principle of Coconstruction The principle of coconstruction is that any autobiographical recollection in conversation is the product of both the speaker and the context,1 as depicted in Figure 1. As shown there, qualities of the speaker and listener influence the reconstruction of events in conversation, as do the larger context within which the reconstruction takes place and prior conversational tellings (through the principle of consistency). In what follows, I limit my attention to speakers and listeners. The coconstruction of events characterizes every occasion on which we talk about an experience. People choose (or are asked) to talk about certain types of events and interpret those events in ways consistent with their own character in order to achieve particular aims with respect to particular listeners. Our listeners dispute details, complete our sentences, enthusiastically reinforce a good story, and clearly express boredom, disapproval, or confusion over a bad one. Coconstruction influences both which events are talked about and what sorts of interpretive statements, details, and emotions are connected with the event as it is discussed. The notion that remembering is a social construction is not new. It has been advanced by social constructivists (e.g., Davies & Harre, 1990; Gergen & Gergen, 1988), social cognition researchers (Hardin & Higgins, 1996; Karniol & Ross, 1996; Schank & Abelson, 1995), and psycholinguists (H. H. Clark, 1996). Empirical work has emphasized speakers' construction of stories about their past rather than the way that listeners influence such stories or the mutual construction of stories in conversations (for exceptions, see Edwards, 1997; Edwards & Middleton, 1988; Fivush, 1991; Manier & Hirst, 1996; Miller, Mintz, Hoogstra, Fung, & Potts, 1992). Speakers influence the way events are recalled. Speakers are part of the social context in which telling occurs, and existing research gives them the primary role in shaping conversational reconstructions. It is not surprising that there is ample evidence In some sense, it is inaccurate to separate speakers from their contexts because they are part of the context just as much as any other element of it. However, because the emphasis in this article is in part on what contextual influences change speakers' behaviors across multiple occasions, there is good reason to treat them as interacting systems rather than parts of a single whole. 1 The Social Construction of the Personal Past in Conversation Growing evidence suggests that memories are dynamic, fluid, and situationally bound constructions that are influenced by the context in which they are produced (Conway & Pleydell-Pearce, SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT 653 Mocrosocto' Context Speaker Qualities Personality Traits Goals Knowledge Behavior Prior Conversationa Recollections Conversational Recollection of Personal Experience Listener Qualities Personality Traits Goals Knowledge Behavior Selection of events to discuss Insertion of specific details Interpretive statements Emotional assocotions Figure 1. The principle of coconstruction: Both listeners and speakers influence the way events are recounted in conversation; however, this process is constrained by the influence of prior tellings of the event. that speakers' characteristics are reflected in their reconstructions of past experiences, although this evidence has largely focused on enduring features of speakers that transcend specific contexts. Moreover, much evidence is based not on conversational tellings per se but on other assessments of memory. Enduring qualities of speakers, such as their personality characteristics, theories about development, gender, and cultural background, clearly influence the way they recall events. In interviews about autobiographies, agentic individuals tell agentic stories, communal individuals tell interpersonally oriented stories, and generative individuals tell characteristic types of stories (McAdams, Diamond, de St. Aubin, & Mansfield, 1997; McAdams, Hoffman, Mansfield, & Day, 1996). Independent individuals find independence-related events more memorable, and individuals with high self-esteem recall experiences consistently with their self-views and motivations (Barclay & Subramaniam, 1987; Thorne & Michaelieu, 1996; Woike, Gershkovich, Piorkowski, & Polo, 1999; Woike, 1995). People reconstruct the past in ways consistent with their notions of development (Karney & Coombs, 2000; Ross, 1989). People from collectivist cultures may recall experiences with less detail and elaboration than those from individualistic cultures (Han, Leichtman, & Wang, 1998), and across many studies, women's memories have been found to be more elaborative and detailed than men's (P. J. Davis, 1999; Fivush, 1998; Ross & Holmberg, 1990). The bulk of this work is correlational and therefore subject to multiple interpretations. Enduring qualities of the speaker probably influence the way experiences are encoded and retrieved or reconstructed. However, these same enduring qualities may arise, in part, from people's recollections of the past and their actual past experiences (Barclay & Subramaniam, 1987; Fivush, 1998). An additional limitation is that many of these studies elicit memories in an interview context, where the interviewer's goal typically is to exert as little influence as possible on the speaker's construction of specific events of his or her life story. This leaves speakers free to reconstruct the past in ways that are consistent with their general views of themselves.2 Finally, enduring qualities of the speaker operate across multiple contexts. In contrast, the goals speakers bring to an interaction may be quite situation specific. People may have numerous goals when they recall an event in a particular moment (Baumeister & Newman, 1994; H. H. Clark, 1996; Conway & Pleydell-Pearce, 2000; Kihlstrom, 1996; Parrott & Sabini, 1990; Ross, 1997; Tversky & Marsh, 2000; Webster, 1997), and these goals may determine which events they recall. For example, people engage in solitary recall of particular memories to support specific self-views, to regulate emotions (Bahrick, Hall, & Berger, 1996; W. M. Klein & Kunda, 1993), or to fulfill identity-related goals (Sanitioso, Kunda, & Fong, 1990). People may recall particular experiences in conversation to establish or maintain intimacy with their listeners (Norrick, 1997) but also to present or construct a particular self-image in relation to that listener (Greenwald, 1980; W. M. Klein & Kunda, 1993; Ross, McFarland, & Fletcher, 1981; Sanitioso et al., 1990; W. J. Swann, 1987; Tesser, 1988). People may engage in selective remembering of events to maintain positive views of their relationship (Karney & Coombs, 2000; Murray & Holmes, 1993, 1994). Finally, people may remember past experiences to select goals or to solve problems (Pasupathi, Lucas, & Coombs, 2001; Pillemer, 1998). Different goals may also result in different ways of recalling the same events. Consider a student who fights with a roommate. If she talks to another friend as an attempt to make herself feel better, she may focus on details of the roommate's wrongdoing (e.g., "she called me a slob, but she knew I had no time to clean"). If her talk is an effort to resolve the conflict and prevent further problems (social and directive functions), different interpretations and details are required--those that acknowledge her own contributions to the problem. 2 Interviewers may be more or less successful in this regard, but training of interviewers often emphasizes a kind of neutral encouragement. 654 PASUPATHI Such effects occur whether the intent to communicate is present prior to or following exposure to information. In everyday life, we sometimes intend prior to an experience to report on it (e.g., telling an absent friend what happened at the party). In other situations, we think during or immediately after an event, "I have to tell X about this!" Therefore, people who expect to talk to others about their experiences may initially encode those experiences with greater coherence, structure, and detail. Listeners thus exert their impact prior to entering a conversational scene. However, listeners also influence what can be talked about, and how it is recounted, during conversations. Whether an event is evoked by a conversation at all can be viewed as one aspect of coconstruction. Speakers and listeners must jointly agree that a speaker can relate an anecdote; such agreements are negotiated in conversations through devices like, "Did I ever tell you about the time .. .." Thus, even the possibility of talking about an experience is something that is coconstructed on the basis of goals and knowledge of listeners and speakers in a particular moment (H. H. Clark, 1996). This means that the coconstruction of our past extends implicitly to those events that are not discussed in conversation. Reasons for not talking about events may range from their insignificance (Rime, 1995) to their distressing or socially undesirable content (Wegner & Lane, 1995). What does not get told is partly a function of the listener's anticipated disinterest or disapproval. Once an event is admitted into a conversation, listeners provide a special remembering context that differs from other contexts. Listeners elicit more opinions, evaluations, and summaries than are included in solitary written recollections (Edwards & Middleton, 1986b; Edwards & Middleton, 1988; Gould & Dixon, 1993; Pasupathi, Stallworth, & Murdoch, 1998). In conversations, stories also have parts that are optional in written narratives. These include items such as a preface, in which the speaker proposes to tell a story (e.g., "I have a great story about that"), and evaluative content, with which speakers give a point or meaning to what they have just related (H. H. Clark, 1996). These items serve to justify the telling of a particular story as one that is not known to the listener or to clarify the reason for telling the story. Consequently, recollections to a listener often contain more interpretive content. Listeners also differ from one another in the kinds of stories they elicit, depending on their characteristics. For example, both men and women report preferences for women as recipients of disclosure, especially emotional disclosure, and this preference is already present by early adolescence (R. A. Clark, 1994). Expertise may also influence the types of stories a listener elicits. Speakers make inferences about a listener's expertise based on prior knowledge and on the listener's feedback. They then formulate utterances in ways that match the listener's needs (H. H. Clark, 1996; Isaacs & Clark, 1987; Krauss, 1987). Reminiscing about a professional women's basketball game with the person I sat next to proceeds very differently than doing so for the friend who could not come. Even young children say more about unshared experiences than shared experiences in conversation, reflecting the different demands of expert and naive listeners (Reese & Brown, 2000). Differences in recall of stories to experimenters and nonexperimenters--nonexperimenters elicit more opinions and evaluations or more complete accounts--may also reflect speakers' assumptions that experimenters already know the material (Ad- Much remembering in social settings might be conceived of in terms of persuasion goals. In two studies, participants were asked to produce a biased account of a set of events by playing a partisan role, such as posing as one party's lawyer (McGregor & Holmes, 1999; Tversky & Marsh, 2000). In both of these studies, participants were asked to write a biased recollection of the events to persuade a particular audience (judges, administrators) to take their partisan view. Speakers' goals for persuading a hypothetical audience about what they should believe determined the way they wrote their accounts and, in particular, which events were retold (Higgins, McCann, & Fondacaro, 1982; McGregor & Holmes, 1999; Tversky & Marsh, 2000). These studies involved hypothetical listeners and nonautobiographical experiences. Thus, participants were free to reconstruct events in ways that they felt would achieve their goals and no actual listener was present. Thus, speakers' goals can influence what is remembered and how it is remembered. However, existing studies leave much about memory in conversational settings unknown. In everyday settings memories may be reconstructed in rather nonnarrative ways and for different purposes than in laboratory settings. Further, everyday settings are likely to simultaneously involve multiple goals (Cohen, 1998; Hyman & Fanes, 1992; Marsh & Tversky, 2001). Focusing on the conversational context of remembering suggests additional questions that may be pursued both in laboratory settings, through manipulations that induce multiple, potentially conflicting goals, and in naturalistic settings, by means of diary or experience-sampling studies. Moreover, most findings on how speakers' goals influence their recall of events are drawn from studies of remembering in particular and unusual social contexts--solitary writing, standardized interviews, or social psychology experiments (W. M. Klein & Kunda, 1993; McAdams et al., 1996; McGregor & Holmes, 1999; Sanitioso et al., 1990; Tversky & Marsh, 2000). These contexts are not typical conversational settings, and they often permit participants to reconstruct events in goal-congruent or self-consistent fashion. Such freedom may be a luxury. Normally, speakers reconstruct memories in conversation in ways that simultaneously take into account their own objectives and those of their audiences. Consequently, the audience also plays an important role in conversational remembering. Listeners influence the conversational recall of events. Everything from expectations about having a listener to the behavior of that listener while he or she listens can influence remembering in conversation (H. H. Clark, 1996; Krauss & Chiu, 1998). Below, I address the following issues in turn: (a) expecting a listener, (b) having a listener versus not having a listener, and (c) specific characteristics and behaviors of listeners. People encode and interpret ongoing experiences prior to talking about them, and this initial encoding is likely to influence the way they will talk about the event later. It is interesting that simply anticipating talking to others could influence that encoding (Guerin & Innes, 1989; Zajonc, 1960). For example, people have more structured, more unitary, and sometimes more detailed and accurate impressions when they anticipate telling others about those impressions (Guerin & Innes, 1989; Higgins et al., 1982; Zajonc, 1960). People are more likely to believe in the position they are transmitting and retain that belief over time when they believe that their audience will have received a similar message (Boninger, Brock, Cook, Gruder, & Romer, 1990; Tetlock, 1991). SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT ams, Smith, Pasupathi, & Vittolo, 2001; Hyman, 1994; Takahashi, 1994). Listeners communicate their comprehension, goals, and interests during recounting, in part through nonverbal behavior. Speakers depend on this feedback; speakers who are deprived of listeners' nonverbal feedback or who are given feedback that is improperly timed are more disfluent in producing single utterances (Krauss, 1987). Speakers recounting longer stories are also more disfluent, speak for shorter times, and include less detailed information when they talk to a distracted listener (Pasupathi et al., 1998; see also Bavelas, Coates, & Johnson, 2000; Tatar, 1998). Similar findings have emerged for uncooperative or dissenting listeners (Dickinson & Givon, 1995). When faced with a listener who does not display appropriate emotional responses at key points in their story, speakers find it difficult to end the tale coherently (Bavelas et al., 2000). Speakers experience distracted listeners as uninterested in their story and respond by producing shorter, less detailed, and less coherent stories. Many of the studies mentioned above required speakers to tell a story to a listener who was performing a dual task (Bavelas et al., 2000; Pasupathi et al., 1998; Tatar, 1998). Both autobiographical (Bavelas et al., 2000; Tatar, 1998) and nonautobiographical stories (Pasupathi et al., 1998) have been used. Theoretically, the dual task (typically, counting words beginning with t or th) eliminates the listeners' capacity to respond to the meaningful content of speakers' stories. The everyday analogy is the situation of talking to someone who is engaged in another demanding task, such as reading the newspaper. This analogy, however, can be criticized. In sum, anticipating telling during an event, having the opportunity to tell, the presence of a listener, and specific characteristics of listeners (gender, expertise, and nonverbal responses) all influence whether experiences get recounted and how they are retold. The influence of listeners is present at multiple points in the process and includes more coherent encoding of the experience, greater detail and different types of detail (including details interjected by the listener) in telling, and more interpretive content (including interpretive content generated by the listener). However, this literature is also incomplete. For example, the mechanisms underlying some listener effects are poorly understood. How does the intent to transmit information change encoding? Although distracted listeners have clear impacts on what speakers say and how well they manage to say it, is it the presence of distracted behaviors, the absence of attentive behaviors, or the timing of nonverbal signals that contributes to this effect? Further, distracted and attentive do not make up the full range of potential types of listeners; other types include overbearing or dissenting listeners. Very little research addresses the ways in which events are not told and how those depend on audiences that either have refused to listen or have been presumed to be uninterested in listening. Further, the evidence presented above is drawn primarily from memory for nonautobiographical experiences. Autobiographical events may involve different levels of motivation to preserve one's initial sense of what is true, as well as different levels of motivation and expertise on the part of listeners. Also, listener effects typically have been examined with stranger interactions, despite the fact that much of our conversational recollection takes place with familiar, intimate partners. I return to this issue below. Finally, and perhaps most important, many studies isolate either speakers' goals or personality, or listeners' goals or behavior. The 655 intent is to reveal the goals of the speaker unaffected by the listener (e.g., the interview context) or to examine effects of listener behavior on speakers who presumably are the same. But coconstruction cannot be adequately understood without examining the interplay of speakers and listeners as they reconstruct the past together. This collaborative reconstruction of experiences is perhaps best studied using approaches in which the distinction between speakers and listeners is not easily made. Speakers and listeners in interaction. Most studies that fit this description examine collaborative recall performance. Such studies ask individuals to recall material together. Recall of personal experiences, general knowledge, and more standard tasks such as story recall have been used in these studies. The question is whether collaborative recall results in better memory performance, with a resulting emphasis on recall of factual details. The findings are clear. Collaborative recall of nonautobiographical stories produces richer and more detailed accounts than typical individual recall (see Andersson & Ronnberg, 1995; Weldon & Bellinger, 1997). This effect is stronger for natural pairs (friends, spouses) than unacquainted pairs, suggesting that the relationship between collaborators matters (Andersson & Ronnberg, 1995; Dixon & Gould, 1996). Although groups enhance performance relative to the individual, this enhancement is smaller than simply summing the memories of two individuals (e.g., Andersson & Ronnberg, 1995; Weldon & Bellinger, 1997), perhaps because groups converge on a version of events that interferes with the retrieval of additional details (Weldon & Bellinger, 1997). Thus, socially shared memories are held with greater confidence and certainty than those that are not shared (Weldon, in press; Weldon & Bellinger, 1997). In addition to looking at memory performance, such studies can also look at the process of collaborative remembering. For example, dating pairs and spouses appear to coordinate the storage and retrieval of knowledge, letting one person be the expert responsible for some types of knowledge in a transactive memory system (Hollingshead, 1998; Wegner, Erber, & Raymond, 1991). When couples are permitted to work together on learning information or on tests of general knowledge, they can outperform stranger pairs with equivalent individual knowledge because they are better able to coordinate their responses (Hollingshead, 1998; Wegner et al., 1991). Moreover, once couples have coordinated knowledge, they are able to retain some of their coproduced knowledge in subsequent, individual testing (Hollingshead, 1998). Of importance is the fact that pairs collaborate less well if deprived of face-to-face contact, highlighting the importance of verbal and nonverbal interchange in the collaborative process (Hollingshead, 1998). Other work has looked at collaborative remembering of autobiographical experiences in dating and spouse pairs. Collaborative remembering of relationship events (first date, argument) produces more acknowledgement of forgetting by husbands and wives, presumably because in coconstructing the events, spouses provide new information that the other spouse no longer recalls (Ross & Holmberg, 1990). Further, the modes within which spouses coordinate recollection of personal experiences and stories vary as a function of how long they have been married (Dixon & Gould, 1996; Gould & Dixon, 1993). Younger couples recalling a vacation not only produce less subjective and more itinerary-like versions of their trip, they also are more responsive as listeners and share their telling more equally. In comparison, older couples 656 PASUPATHI dition, however, the listener is presented as a person who dislikes self-enhancing individuals. Participants who have been threatened disregard their listeners' preferences and make highly positive self-statements anyway (Gollwitzer & Wicklund, 1985). Such work does not focus on autobiographical remembering, and because the listeners in these studies are not actually present, speakers receive no direct feedback from them. As such, these findings provide only a suggestion for how conflicting goals get resolved. Similarly, when the listener's disinterest is strong, he or she may shut down attention to the speaker. Empirical support here is even less direct but can be inferred from findings on stonewalling, or unresponsive listening during intensely negative sequences of marital conflict discussion (Gottman & Levenson, 1988). Gottman and Levenson speculated that stonewalling functions to regulate emotion by shutting down attention to the aversive conversation. It is notable, however, that stonewalling provokes more intense negative responding from the speaking spouse and is thus an ineffective response over the long term. Clearly, both speakers and listeners shape the way that events are narrated in conversation. The enduring qualities of speakers (personality, gender, culture) have received more attention in the literature than any other facets of coconstruction, including listener effects and collaborative remembering. Examining how listeners influence the reconstruction of events would provide rich information about the social shaping of autobiographical memory, a topic that has been neglected (Schacter, 1996). Moreover, collaborative memory work shows that listeners are not reducible to other variables, such as changes in speakers' goals or the addition of information to a speaker's account. Limiting examination of listeners to manipulations of speaker goals or to the presence of additional information also precludes examining conflicting agendas on the part of speakers and listeners. Finally, examination of collaborative remembering has to expand to include situations in which remembering occurs in the context of pursuing other goals. Collectively, the evidence clearly supports a view of conversational remembering as a coconstruction of the past in collaboration with listeners. But coconstruction would be of little developmental interest if it did not have more lasting implications for how events are remembered. Such tellings become more important because of the principle of consistency, which refers to the fact that a collaborative telling on one occasion will influence subsequent recall of the event in another situation. Next, I consider the evidence for consistency. exchange longer monologues and provide fewer listener responses (Gould & Dixon, 1993). Perhaps younger couples are still establishing who is responsible for carrying which portion of a memory. Other studies have focused on parent-child collaborative remembering, which consists partly of instructing and aiding children in learning how one remembers in conversation (Edwards & Middleton, 1986a; Fivush, 1991; Miller et al., 1992). Although adults do not instruct one another explicitly in how to remember experiences, the adult work suggests they probably do so implicitly, by reminding one another of forgotten events or details and negotiating interpretations. Recent work on how the collaborative recall of family experiences gets coordinated among adults supports this notion (Hirst, Manier, & Apetroaia, 1997; Manier, Pinner, & Hirst, 1996). This work has revealed different roles in group remembering, such as the narrator (who speaks the most and contributes the most detail to the group recollection), the mentor (who facilitates remembering), and the monitor (who interjects to prevent perceived inaccuracies and to ensure that important details are not left out). Mentors and monitors might be conceived of as listeners who exert clear influence over the group's final recollection. Note that people may take on different roles for narrating different events; the roles are not stable individual differences, but rather reflect actions of particular individuals within specific social contexts. Taken together, these findings suggest that collaborative recalls are shaped by the joint action of the contributors. The ways in which individuals coordinate the remembering of particular events, such as the distribution of roles taken, clearly demonstrate that both speakers and listeners contribute to the production of a shared recollection. Individuals' solitary recalls might have been quite different from the eventual collaboratively recalled experience (Hirst et al., 1997; Manier et al., 1996). For general knowledge, laboratory stimuli, and personal events, collaboration is also better than individual remembering in terms of the amount of detail recalled, though it falls short of being better than combining individuals' memory performance. This implies that collaborative remembering is truly collaborative--that is, different than simply adding up individual actions. Importantly, collaborative rememberings are viewed as more true than individual ones. Finally, collaborative remembering functions differently for people who know one another, which suggests that listener studies focusing on stranger pairs suffer from a serious shortcoming. Collaborative remembering work examines situations where all parties share a common goal: that of recalling an event, story, or fact. The real-life context of talking about the past, in contrast, often involves different, nonmemory goals. For example, people may recall the past in the service of solving a problem or establishing a shared conception of someone else (Pasupathi et al., 2001). Further, speakers and listeners can have conflicting goals. If a speaker wants to produce a detailed telling and a listener does not want to hear it, what may happen? One possibility, suggested by research on symbolic selfcompletion, is that the person with the stronger motivation will "win." In such studies, key features of the participants' selfconcept are threatened, often by informing the person that their personality does not match their desired career goal. The participant is then (in an ostensibly different portion of the experiment) given an opportunity to say positive things about themselves with regard to fhe threatened self-conception to a listener. In one con- The Principle of Consistency The principle of consistency means later recollections of events will be consistent with earlier recollections. Consistency is not the same as accuracy or veridicality, which are defined with reference to some objective standard--what really happened. However, because of the constructive nature of autobiographical remembering (Conway & Pleydell-Pearce, 2000; Schacter, 1996), assertions about consistency are complex and problematic. Consistency can be understood as implying that recollection of an event produces small, incremental changes in the nature of an autobiographical knowledge base, consistent with associative network models of memory (J. R. Anderson & Bower, 1973; McClelland, 1994). These changes probabilistically increase the likelihood that a later reconstruction of a memory will resemble earlier reconstructions SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT in terms of factual and interpretive content because they strengthen associations between those details and the notion of that event. However, because assessments of memory typically focus on what people reproduce and do not directly assess changes in the strength of associations of the sort implied by consistency, assessments of consistency in memories over time are indirect. This raises several issues, among them discrepancies between what is reported and what is recalled and differences between the capacity to recall and the ability to recognize events and associated details. On any given occasion when people talk about events, they may deliberately omit some information (Marsh & Tversky, 2001). This is particularly true when they talk to different listeners across occasions (e.g., friends vs. parents). In such cases, people recall more than they tell; thus, estimates of consistency based on what they tell will be inaccurate. In other cases people may be unable to recall some information but would accurately recognize omitted details as being true. Studies of autobiographical recognition often show extremely high recognition accuracy for experiences, even if those experiences were not produced under free recall conditions (Barclay & DeCooke, 1988). Because consistency is evaluated in terms of the extent to which the same information is repeatedly reproduced when an event is remembered, discrepancies between what is remembered, reported, and recognized can reduce apparent consistency. Another problem for assessing consistency is that many autobiographical events are hierarchically organized and complex (Hyman & Loftus, 1998). Consider an all-day picnic. In telling about this daylong event, an individual may focus selectively on the specific subevent of opening the basket and finding the cheese melted everywhere. On another occasion, he or she may tell about some other subevent, such as playing frisbee with the dog after eating. These sorts of inconsistencies belong to the class of errors called time slice errors (Hyman & Loftus, 1998). If an individual is not focused on the same specific portion of an experience, apparent inconsistency will result. Conceptually, some types of inconsistency (those plausibly due to time slice errors) may need to be evaluated differently than others (those that contradict earlier content). Further, omission of previously recalled information may require cautious interpretation, because of both time slice errors and discrepancies between what is remembered and what is retold. The picnic example further reveals that consistency can be evaluated on many levels, ranging from the gist of memories (e.g., "we did some kind of picnic") to interpretations (e.g., "that was a very nice day") to specific details (e.g., "you packed the cheese badly and it melted all over everything"). Consistency across multiple telling occasions can be addressed only at comparable levels. Moreover, at these different levels different degrees of consistency might be expected. For example, on the basis of research on errors and accuracy in autobiographical remembering, we might expect substantial consistency in gist and in some interpretations, such as overall evaluations of events as having been positive or negative (Barclay & DeCooke, 1988). Alternatively, in comparing across remembering occasions, we might expect substantial fluidity in the recall of particular details, as some evidence suggests that details in conversational recall may be primarily used to support the conversational point being made (Hyman, 1994). And, events can clearly be reinterpreted in more complex respects throughout life, in light of new information or new perspectives--as when people reinterpret stressful life events 657 as having provided benefits (C. G. Davis, Nolen-Hoeksema, & Larson, 1998; McAdams, 1993; Thome, 2000). Finally, consistency at one level probably influences consistency at others. For example, if interpretive parts of memories include schematic "story skeletons" that will aid later memory (Schank & Abelson, 1995), then consistency at the interpretive level may also lead to some consistency at the level of details. Unfortunately, most studies examine only one level (typically factual details). Because of these complexities, researchers must be (a) clear about how they are addressing discrepancies in reporting, recalling, and recognition at both conceptual and methodological levels and (b) specific about how levels of consistency are examined. Despite these thorny issues, however, there is ample theoretical and empirical ground for expecting consistency. Below, I begin by reviewing some of the cognitive, social, and motivational reasons to expect consistency in memories over time. I focus heavily on memory research because of the direct relationship between such work and the issue of memory consistency over time. Other reasons to expect consistency stem from more general accounts in the social and motivational literatures, which are of particular importance in the context of conversational recounting. Finally, I examine selected empirical support for consistency, focusing on studies in which the memories examined were coconstructed at least in some sense. Memory mechanisms. Established processes and phenomena from traditional research on memory provide multiple reasons to expect consistency over time. Among these are rehearsal, story schematization, retrieval-induced forgetting, and source memory. Applying this work to conversations about past experience means viewing such conversations as rehearsals of events involving retrieval of information about those events from memory. Frequent rehearsal of memories renders them relatively stable and resistant to change, distortions, and deletions (Johnson & Chalfonte, 1994; Koutstaal et al., 1998; McClelland, 1994). For example, flashbulb memories, or memories for the time and place in which one learned surprising, consequential news, are well retained and vivid. This may be because such memories are often rehearsed, both mentally and conversationally (Brown & Kulik, 1977; Conway, 1995; Rubin & Kozin, 1984). Most studies of flashbulb memories do not separate covert mental rehearsals like thinking about an experience and overt conversational rehearsals.3 Analyzing flashbulb memories also reveals that they are sometimes quite inaccurate (Conway, 1995). This suggests the ways that rehearsal can introduce errors that then persist over time (Belli & Loftus, 1996; Roediger, McDermott, & Goff, 1997). These errors may occur in storage, as in partial degradation accounts, but also in retrieval, because rehearsed items are more accessible than initially experienced but unrehearsed items. The result is that people are more likely to recall the rehearsed (and incorrect) version of events than their original experience. Story-schema accounts of memory also imply that memories of events will be quite consistent over time (R. C. Anderson & Pichert, 1978; Bartlett, 1932/1995; McGregor & Holmes, 1999; Schank & Abelson, 1995; Sulin & Dooling, 1974; Tversky & 3 In fact, Conway (1995) notes that rehearsal may operate differently for flashbulb memories than for other memories and that rehearsal has been inconsistently linked to flashbulb memories. 658 PASUPATHI certainly, the negative behavior of listeners can be experienced as a punishment (Pasupathi et al., 1998; Tatar, 1998). In addition, people are probably motivated to maintain consistent recollections of coconstructed past experiences. Research on dissonance (Festinger, 1957), accountability (Tetlock, 1991), conformity (Asch, 1952), persuasion by social proof (Cialdini, 1988), self-presentation (Rhodewalt, 1998), self-verification motives (W. J. Swann, 1987), social comparison (Festinger, 1954), social representations (Moscovici, 1988), and shared reality theory (Hardin & Higgins, 1996) all imply that once we construct a socially accepted version of the past, our subsequent recollections will be consistent with that version. This is analogous to the idea that social groups, from romantic couples to larger societies, engage in a kind of reality monitoring as they determine what is believable and true in remembering (Johnson & Mather, in press). Much of social psychology interprets pressures for consistency within the framework of cognitive dissonance (Festinger, 1957). The notion is that conflicting cognitions--in this case, conflicting representations of events--produce unpleasant internal arousal. Consequently, people seek to reduce the dissonant cognitions, even when doing so means discounting positive and potentially rewarding information (W. J. Swann, 1987). Presenting the past conversationally in ways that are discrepant with prior recollections could induce dissonance, especially because such presentations are usually voluntary and always public. Finally, because much of what we believe about the world is not objectively verifiable but rather gets affirmed through social consensus, the coconstruction of the past also permits us to see our version of events as true (Festinger, 1954; Hardin & Higgins, 1996; Moscovici, 1988; Stallworth, 1995). These social and motivational contributors to consistency may apply more to consistency at the interpretive or schematic level. That is, at the level of social pressures for consistency, it is less crucial to get the details of the party correct and more important to recall oneself being shy or outgoing at the party. Further, these approaches suggest that social consensus about experiences will carry particular truth weight in relation to other knowledge sources. As reviewed, there is substantial reason to expect conversational recollections of past events to influence subsequent memory for those events. Below, I examine evidence that recollections that are arguably conversational do in fact influence later memory for events. In doing so, I consider evidence drawn from studies in which participants engaged in a rehearsal of information that was influenced by social features of their rehearsal context--in other words, studies involving coconstruction. This includes studies involving nonautobiographical material and studies in which the rehearsal context was only metaphorically conversational. Evidence for consistency over time. Perhaps the most famous evidence that memory for an event in one context affects later memory comes from research on eyewitness memory. Numerous studies have shown that when individuals are asked about events in misleading ways, they produce a biased account. They subsequently recall the same events in ways influenced by the postevent misinformation or leading question, remember a big crashing noise or a beard when none was present (Belli & Loftus, 1996; Loftus, 1992; Zaragoza et al., 1997). Such findings are primarily based on recognition tests, but recall methods have also demonstrated strong misinformation effects (Belli & Loftus, 1996). A distorted version Marsh, 2000). Initially, many of the details of the original events may be available and accessible. Formulating events in story form, however, involves the construction or application of a schema about the event; such schemas render many details comprehensible and meaningful at the expense of excluding schema-inconsistent information. This process may be of particular importance when talking about experiences with others because of culturally shared schemas for structuring narratives of personal experience (Habermas & Bluck, 2000; Miller, 1995; Schank & Abelson, 1995). Once established, schemas function as recall aids, ensuring that the recall is consistent with the schema and decreasing the likelihood that schema-inconsistent features of the events will be included. In fact, schemas can result in schema-consistent errors, that is, memories for details that fit the schema but did not occur (Tversky & Marsh, 2000). Note that schemas can be viewed as interpretations of an event (Schank & Abelson, 1995), and thus consistency in interpretive recall may lead to consistency in recall of factual details. Active suppression of some aspects of memory when retrieving information can result in retrieval-induced forgetting. In retrievalinduced forgetting, aspects of events that are inhibited from recall on one occasion are subsequently less likely to be recalled (M. C. Anderson, Bjork, & Bjork, 1994). Recollections are thus selective rehearsals that strengthen memory for the recounted features of an event but that may simultaneously weaken memory for other features. Rehearsals of events in conversation also connect memory for facts and interpretations of those events to two different sources: the original experience and the telling. Source memory, or recollection for the place and time that information was encountered, is more quickly forgotten than memory for the information itself (Johnson, Hashtroudi, & Lindsay, 1993; Schacter, Norman, & Koutstaal, 1998; Zaragoza, Lane, Ackil, & Chambers, 1997). It may thus become difficult for people to distinguish between a recollection and the initial event unless they can accurately recall the circumstances under which telling occurred (the source information). Thus, source memory errors may displace a memory for "how it actually was" by one for "how I told it to my husband" (Dickinson & Givon, 1995; Hyman, 1999; James, 1890; Schank & Abelson, 1995). On subsequent occasions, reconstructions of the event are likely to be consistent with the earlier telling. In sum, conversational tellings are likely to influence subsequent remembering because those tellings function as a rehearsal that is selective, is schematic, and can lead to source memory confusions. However, because conversational recollections occur in social contexts, there are other reasons to expect them to be consistent across time. Social and motivational contributors to consistency. From social and motivational perspectives, conversational recollections of the past are viewed as performances of one's history for an audience. This view implies the applicability of research on rewards, accountability, and dissonance to the issue of whether a recollection of the past at one time will influence subsequent recollections. First, rewarded behavior will tend to be repeated (Skinner, 1953). A version of events that wins social approval may well be turned into an anecdote that is pulled out of the drawer whenever possible. Versions that do not win such approval may be discarded; SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT of events that is collaboratively produced by the participant (speaker) and experimenter (listener or interviewer who introduces misleading information or bias) later may be remembered consistently with the distorted version. Another compelling illustration involves the building of false memories (Hyman & Loftus, 1998; Loftus, 1997). In a typical study, participants are told that they had or may have had a specific childhood experience, and are asked to remember it (often, relatives are named as the source of the experimenter's knowledge). Most paradigms in this field involve two key elements: (a) communicating the idea that people can in fact retrieve early memories and (b) encouraging participants to generate plausible event content, including imagery. Many involve production of a narrative about the imagined event in an interview setting (a specific type of conversation). The findings suggest that some individuals come to believe memories for events that never occurred, depending on the plausibility of the event and, notably, the extent to which the individual is prone to source monitoring errors (Hyman, 1999; Hyman & Loftus, 1998).4 Research on retractors of recovered memories supports the idea that narrative coconstruction with the therapist is a critical process in the building of false memories (de Rivera, 1997). For generalizing to more everyday memory talk, these findings suffer from several limitations. These include their focus on forensic and clinical settings; the fact that not all individuals in such studies appear susceptible to memory distortion effects; the nature of the events examined; the nature of the power and expertise imbalance between the misinformation source and the participant; and ultimately the extent to which interviews, leading questions, and questionnaire-provided misinformation adequately reflect conversations. Fortunately, other findings broaden the evidence base for consistency effects. Above, I reviewed two studies in which participants wrote accounts of a series of events to persuade an audience to take a particular viewpoint (McGregor & Holmes, 1999; Tversky & Marsh, 2000). As noted, the resulting letters were biased in the events and details they included. After writing their accounts, in both studies, participants' subsequent memory for the events was reassessed. Tversky and Marsh (2000) had participants engage in a distractor task, following which they were asked to remember the events again, this time without any particular bias. Their delayed memory for the events was biased by the initial account they had written. The memory bias was evident using free recall or recognition measures. McGregor and Holmes (1999) assessed delayed memory on two occasions following participants' initial exposure to the events and biased recall. Their results confirm that an initially biased account has long-term effects and further suggest that there are different time courses for different aspects of storytelling influences on memory (McGregor & Holmes, 1999). Specifically, writing the initial account biased subsequent retrieval of facts from the events, but participants also developed what the authors call a heuristic, or gist, version of the events (interpretation) that directly biased later judgments about those events, such as which character was most to blame. The heuristic bias was evident at all delays evaluated, whereas the bias in memory for event details was evident only at midrange time periods. This may have been due to a floor effect in recall of details after the midrange delay (McGregor & Holmes, 1999). 659 Work on person perception also suggests that versions of events developed for one audience will become difficult to adapt for a new audience over time (Higgins et al., 1982). People communicating impressions of a hypothetical person adapted to different audiences when the delay between exposure to information and communication was brief (within a single session). But over longer delays, they were no longer able to change their story for a new audience. These studies provide support for consistency in memory for events across different remembering contexts. However, the memories were not autobiographical and the audiences were hypothetical, limiting the applicability of these findings. Four other studies looked more explicitly at actual conversations and subsequent memory for events or objects. A series of dissertation studies showed that individuals who conversed about abstract paintings converged on a set of dimensions for representing the paintings (Stallworth, 1995). Subsequently, when participants wrote individual descriptions, they used the dimensions they had developed in conversations. Further, those pairs who expected to work together in the future were more likely to do so than pairs who did not expect further contact. Manier and colleagues (Manier et al., 1996) asked individuals to read a news story and write down what they remembered from it (pretest). Subsequently, some individuals remembered the news story to the experimenter, whereas others engaged in group recall. The groups interjected, discussed, argued, and negotiated memory. These discussions presumably allowed some content to be added to what participants had initially encoded, similar to postevent information in eyewitness memory studies. Finally, following a distractor task, participants provided another written recall (posttest). Comparisons of posttest memory (number of facts recalled) revealed that those who engaged in a collaborative recall had significantly better memory at the posttest than those who recounted the event to the experimenter. Later memory incorporated details from the collaborative discussion, thus demonstrating some consistency with the interim recall and improving overall performance. Two studies of memory for film clips reveal similar effects. In these studies, participants watched film clips and provided initial written recalls of the clips. These initial recalls ensured that participants had initially remembered similar amounts of information from the clips. They then told a listener about these clips. Listeners were assigned to be attentive, distracted, or uncooperative (Dickinson & Givon, 1995; Pasupathi et al., 1998). Participants subsequently provided a second written recall of the film clips. Recalls were scored against a set of details extracted from the film clips. Attentive and engaged listeners elicited more detailed and elaborative versions of the events during recounting, leading in turn to more detailed memory at the subsequent posttest. Both studies showed that attentive listeners facilitated long-term memory, whereas distracted or uncooperative listeners did not. Such findings imply consistency (more recall at Time 1 leads to more recall at Time 2) but do not directly demonstrate it. However, in one study, consistency was directly examined by tracking the retention of specific details across immediate, conversational, and delayed The role of imagery in producing false memories (Hyman, 1999) is important for understanding false memory construction but less germane to the issue of how conversations can induce such memories. 4 660 PASUPATHI emotional and meaningful. Such qualities provide justification for telling an audience. More emotional and meaningful autobiographical memories are often better remembered over the long term, and this greater memorability has been attributed to more frequent rehearsal in thought and conversation (Conway, 1995; Rubin & Kozin, 1984). At the same time, not all traumas are admissible in conversations; some involve shameful or troubling experiences that may be less amenable to conversational telling. Third, speakers' and listeners' goals may diverge when highly significant events are being discussed. Conversations about significant events entail high stakes, especially when the events are negative. Speakers may be highly sensitive to the listeners' reaction for such important experiences; listeners, in turn, may have difficulty in responding well to tales of true distress or suffering. In fact, listeners may simultaneously experience a desire to be helpful and a desire to avoid having to hear a tragic story (Barclay, 1996; Pennebaker, 1997). Ambivalent responses from listeners may lead speakers to not tell, to minimize the experiences or shorten their tellings, and to consequently remember the experience differently. However, such ambivalence may be greater for some types of traumatic events than for others. The end consequence may be less detailed, rich, and accessible memories for inadmissible traumas than for admissible ones, although this is a speculation awaiting empirical confirmation. Finally, the availability of existing schemas for representing experiences may be challenged by some highly significant life experiences (Staudinger & Dittmann-Kohli, 1992). In these cases, conversations may be even more influential in making sense of events than for more mundane, easily understood experiences. Once again, not all traumas are equal in this respect. Some traumatic experiences have a widely available array of representational schemas, whereas others do not. Further, history can witness the emergence of such schemas over time as autobiographies, biographies, films, and novels about particular types of experience proliferate. To summarize, the importance of the event being recounted may influence the initial encoding of the event, the likelihood of talking about it, the goals listeners and speakers bring to the situation, and the availability of existing schemas for representing the experience. Thus, the principles are applicable to both mundane and significant events, but the specific ways in which they play out may be moderated by the type of event being talked about. Further, a simple distinction between significant and mundane events breaks down in light of considering social aspects of remembering. For conversational recountings, there is great variability among highly significant experiences, along both valenced and other dimensions. For example, flashbulb memories represent a set of highly significant experiences with great admissibility in conversation. It is widely acceptable to relate how one learned about O. J. Simpson. In contrast, memories of sexual assault are relatively inadmissible and may be talked about only in very special, private, and intimate circumstances. In the former case, rehearsal and positive or engaged responses from listeners are likely to support a vivid, detailed memory, which would be consistent with some of the literature on such memories (Conway, 1995). In the latter case, ambivalent listener responses may aid in creating a more repressive response to the experience, with all the attendant consequences (Conway & Pleydell-Pearce, 2000; Wegner & Lane, 1995). However, not all traumatic experiences are inadmissible; recalls. The consistency of the delayed recall with the retold version of the film clip was greatest when the listener had been attentive (Pasupathi et al., 1998). Taken together, these studies suggest that conversational recollections of events at one point in time do influence subsequent memory, in terms of the factual information retained across time, and the interpretations that individuals have of past events. Further, they point to meaningful differences in the impact of different types of conversational recollection. Collaboratively, agreeably coconstructed recollections are related to better long-term memory for details and greater influence of tellings on subsequent memory. Disagreement, noncooperation, disinterest, and lack of support from a listener is related to poorer memory in the long term. Such findings lend support to the principle of consistency. However, these studies also suffer from a number of important limitations and may overestimate the power of the consistency principle. First, not all studies examined consistency directly by looking at the retention of specific details or interpretations across multiple remembering occasions. Rather, consistency was indirectly inferred from other effects (e.g., total amount of information recalled). Second, most studies examined the impact of a conversational (or quasi-conversational) recollection on subsequent written recollections. It is less clear whether one could expect consistency in subsequent conversational recollections of the same event to a different listener, although the person perception work reviewed above suggests this will depend on the time between those recollections (Higgins et al., 1982). Third, and most important, none of these studies examined autobiographical memory. Autobiographical recollections involve higher stakes with respect to self-presentation. Individuals may vary more in retelling events across situations when the consequences are significant, justifying the omission of some information or the embellishment of other details. And, as noted above, autobiographical events are often more complex than the events examined in these studies, with greater potential for inconsistency depending on the particular facets of the event (e.g., time slice, facts vs. interpretations) being remembered. Thus, the consistency of conversational recountings remains relatively poorly researched. Significant Versus Mundane Events: Does the Type of Event Matter for Consistency or Coconstruction? Thus far, I have not distinguished between mundane and significant events, although this distinction is clearly important in the field of autobiographical memory and for memory more generally (Barclay & DeCooke, 1988; Christianson, 1992). There are, however, four potentially relevant differences between highly significant events and more mundane ones. The first is that there may be neurological and cognitive advantages in the encoding of traumatic events (Christianson, 1992; Conway & Pleydell-Pearce, 2000). Such differences include a narrowing of attentional focus to central details at the expense of peripheral ones; these differences may also apply to significant positive experiences. As a consequence, speakers recounting significant experiences would have improved access to central details but poorer access to peripheral information. Such differences might serve to constrain potential interpretations of the events. Second, highly significant events may be more frequently rehearsed than trivia) and mundane events because they are more SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT 661 consider the mountaineering disaster reported in Into Thin Air (Krakauer, 1997). This is a trauma with a willing and interested audience numbering in the millions. Socially Constructing the Personal Past Taken together, the principles of coconstruction and consistency imply that what we tell certainly influences, and may become, what we "know" about our own past. Existing limitations in the literature need to be addressed, however. Relatively few studies examine how coconstructed events may or may not influence subsequent memories for those events; rather, most of the work reviewed thus far focuses either on consistency or on coconstruction but not on both. Of those that do address both principles (e.g., Pasupathi et al., 1998), virtually none (on adults) examined actual autobiographical events, where people may have more of a vested interest in a particular version of the past (see, e.g., Greenwald, 1980; W. M. Klein & Kunda, 1993; Ross et al., 1981; Sanitioso et al., 1990). Many studies, particularly those not motivated by memory-related questions, do not distinguish among different stages in memory processes.5 For example, generative individuals may tell generative stories because they encoded their experiences in terms of generative issues, because they and their audiences have constructed generative stories of those experiences in conversations, or because they use their self-conception as a cue in free recall (Barclay & DeCooke, 1988). Those studies examining listener effects have emphasized veridicality and retention of details (Dickinson & Givon, 1995; Manier et al., 1996; Pasupathi et al., 1998) rather than thematic consistency and have used stimuli that do not easily permit looking at interpretations in recollection. Perhaps most important is that few studies that address both coconstruction and consistency have examined familiar social partners, even though it is our friends and family with whom we are most likely to recount the past. All intimate relationships provide a context for conversational recall that has distinctive qualities when compared with recall with a stranger, and findings from collaborative memory studies bear this out (Andersson & Ronnberg, 1995; Dixon & Gould, 1996). First, accountability to friends and family is much higher than accountability to experimenters encountered in a laboratory setting. Experimenters have no knowledge of what actually happened and often attempt to display no investment in particular versions of their participants' past experiences. There is little cost for distorted recollections of the past in the laboratory--nobody comes along two weeks later and says "that was a lie!" In contrast, friends and family are likely to have great investment in particular versions of the past, sometimes because they were present at the event and have their own understanding of what happened and what it meant (and may thus serve as mentors and monitors for a narrator). Intimate relationships thus exert stronger pressures for constructing versions of events that are agreed on by audiences and speakers, as well as stronger pressures for consistency. There may also be differences in coconstruction and consistency within different established relationships. Different relationships may involve different expectations, histories, and accountability to one another (Tetlock, 1991). In addition, just as individuals reconstruct the past in ways that reflect their general self-conceptions (McAdams et al., 1997; McAdams et al., 1996), relationships may involve themes that structure experiences both as they occur and when they are recalled in conversation (De La Ronde & Swann, 1998). Different relationships may be better or worse audiences for our accounts of the past. Duck, Rutt, Hurst, and Strejc (1991) found some evidence that people felt less positively about conversations with lovers than friends, although their focus was not explicitly on conversations about past experiences. Specific relationships are contexts that vary in the event recountings they elicit and the way events get recalled. At present, however, little is known about how talk about the past might vary as a function of the relationship between the speaker and listener. Our autobiographical memories are a product of coconstruction and consistency in conversations across our lifetimes. This social shaping of the past, in turn, may both influence and be influenced by the unfolding of the future. Put differently, adult development can be shaped by the social construction of the past, and the social construction of the past may change over adulthood. The next section of this article takes up these issues in more detail. The Social Construction of the Past Across Adulthood: Developmental Implications If conversational recounting of the past is both coconstructed and consistent, it is a process by which other people necessarily influence the contents of our autobiographical memory. Because memory and identity are intimately connected, the social shaping of memory may also be a process by which the self is socially shaped. However, conversational recounting also takes place in the context of a developing person, and developmental changes can influence the way that coconstruction and consistency take place. Below, I consider these two developmental implications. First, can the social construction of the personal past influence identity development during adulthood? In other words, does what we tell become what we know and what we believe about ourselves? Second, does development in adulthood change the way that coconstruction and consistency operate? Constructing the Past in Conversation: Implications for Identity Stability and Change Who we are has been said to be a function of where we have been (Barclay & Smith, 1992; Hirst et al., 1997; Neisser, 1988),6 and a sense of self may be a prerequisite for remembering autobiographical experiences (Howe, Courage, & Peterson, 1994). I believe that the social construction of the past in conversation opens the door for maintaining stability in one's identity or incorporating change. This can be illustrated with two facets of identity: the life story and motivational aspects of identity. 5 My gratitude goes to Elizabeth Marsh, who pointed this out to me after reading an earlier draft of this article. 6 Identity and personal history may be related, but there are dissociations between semantic knowledge of the self like that assessed in selfconception research and episodic knowledge of the self, or autobiographical memory (Kihlstrom & Klein, 1994; Rhodewalt, 1998). In fact, one amnesic patient showed highly accurate trait judgments about herself despite the absence of memory for specific episodes of her life. Thus, recollections of what we have experienced do appear somewhat distinct from knowing who we are, even if the latter are initially based on, or grounded in, the former. 662 PASUPATHI The life story. The life story is the facet of identity with the greatest complexity, contextual variability, and potential for reflecting change and development (Bruner, 1990; Fivush, 1991; Freeman, 1993; Habermas & Bluck, 2000; Hermans, 1996; Linde, 1993; Mankowski & Rappaport, 1995; McAdams, 1996; Miller et al., 1992; Neisser, 1988; Thome, 2000). This aspect of identity is believed to emerge in adolescence and early adulthood (Habermas & Bluck, 2000; Thome, 2000). Because the life story is constructed from significant memories that are connected into a coherent, ongoing narrative, autobiographical memory is the raw material from which identity is constructed as a life story. If, as argued above, the contents of autobiographical memory are at least partly a product of social construction, life story identity is also shaped by the processes of coconstruction and consistency. telling In someone their life story or parts of it, people select certain episodes for narration. Many of these episodes will be events that are canonical parts of a life story in our culture (Habermas & Bluck, 2000; Strube, Gehringer, Ernst, & Knill, 1985), whereas other episodes are selected to illustrate a particular quality or theme in our lives (Singer & Salovey, 1993) and still others may be selected or not selected because of the audiences to which we are speaking. For example, authority figures may elicit and tolerate fewer tales of wild youth than peers. Some friends have greater interest and tolerance for work-related talk and experiences than others. Thus, the life story itself may be shaped in ways that reflect coconstruction and consistency, as audiences reject or accept certain events as "good" parts of the story (Linde, 1993). Much of this shaping probably occurs in what has been termed "emergent adulthood," or the years between 20 and 30 (Amett, 2000, p. 469). By this time, individuals have the capacity to construct life stories (Habermas & Bluck, 2000) but still show notable instability in the events they include in that story across occasions (Thome, 2000). Middle-aged adults may show greater stability in their selection of memories (Thorne, 2000). Note that one of the changes from emergent adulthood to midlife is an increasing stability in the audiences we turn to when recalling the past. The potential to revise the life story in light of new experiences and new audiences remains an avenue for change or stability throughout adulthood. Depending on both ourselves and the responses of our audiences, we may reformulate our life stories to permit change (McAdams, 1993) or be trapped in our existing identities (W. B. Swann, 1996). In sum, the life story represents a facet of identity that is directly subject to coconstruction and consistency pressures and directly offers the potential for stability (keeping the same events, interpretations, and themes) or change (adding or removing events, interpretations, or themes). Motivational facets of identity. Motivational constructs such as goals and self-efficacy beliefs compose an additional aspect of identity, one that includes concrete goals, such as losing weight or learning a new sport; more generalized strivings, such as ideal or hoped-for selves; and specific conceptions of skills and talents. These aspects of identity have direct implications for adult development in their connection to the activities and contexts people select (e.g., Heckhausen & Schulz, 1995). The relationship of memories for the past and current goals and strivings has been recently articulated in rich detail by Conway and Pleydell-Pearce (2000). They show that goals facilitate selective, relevant retrieval from an autobiographical knowledge base and that goals in turn are grounded in that knowledge base (see also Karniol & Ross, 1996). This latter point implies that the social construction of memories can influence goals by changing the autobiographical knowledge base. However, motivational facets of identity also are socially constructed directly in the context of memory talk. Memories often invoke goals and self-conceptions to explain or make sense of events (Hilton, Mathes, & Trabasso, 1992; Miller et al., 1992; Thome, 2000). Consequently, memory talk serves as a rehearsal of existing beliefs or permits the introduction of new self-related beliefs. Tellings make a public commitment to oneself as a particular sort of person and simultaneously achieve social support for that vision (Tetlock, 1991; Tice, 1992). For example, Pillemer and colleagues (Pillemer, 1998; Pillemer, Picariello, Law, & Reichman, 1996) reviewed how people remember specific events that resulted in changed goals--for example, conversations with professors that resulted in shifting views of a person's ability and shifting career goals. Such conversations probably involve the coconstruction of beliefs about a person's abilities, often in conjunction with past events (e.g., comments in class, performance on exams). Such beliefs, once established, permit people to reconstruct other experiences in similar terms, rehearsing and strengthening a self-conception. They also, as in the rock-climbing example earlier, promote the seeking out of experiences that can influence self-conceptions. Thus, talking about past experiences in conversation is also a process by which we maintain stability or permit change in our identities, across multiple levels of identity and via multiple pathways. However, one problem with this proposal is that, as noted earlier, our identity also influences the way we encode and recall (in conversation or alone) our experiences (Greenwald, 1980; W. M. Klein & Kunda, 1993; McAdams, 1996; McAdams et al., 1996; McCrae & Costa, 1986; Ross, 1989; Sanitioso et al., 1990; Thome & Michaelieu, 1996; Woike et al., 1999; Woike, 1995). Existing evidence relating self to memory is often correlational and raises the question of whether remembering the past in particular ways can cause changes in aspects of identity as well as in memory. Supportive evidence can be found in studies of reminiscence therapy and in studies of self-presentation and carryover effects. Insights from reminiscence research. At the heart of reminiscence therapy7 for older adults is the notion that the recall, interpretation, and integration of past experience is a key process in successful aging (Bluck & Levine, 1998; Haight, 1991; Thornton & Brotchie, 1987; Webster & Cappeliez, 1993; Wong & Watt, 1991). This assumption has led to a variety of therapeutic techniques aimed at helping older adults to engage in reminiscence or life review.8 In general, the empirical evidence is mixed in quality and in outcome, probably because of the broad variety of inter7 Note that recipients of reminiscence therapy may be individuals with psychopathology, but many studies also include visitors to senior centers and otherwise normal, community-dwelling elderly. 8 These terms are now always synonymous (see Staudinger, 1989, for a detailed review of differences and similarities between reminiscence, life review, and other such processes). However, in terms of therapeutic interventions, all involve the reconstruction of past experience in the present, with the expectation of changes in identity. SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT ventions and outcome measures used (Haight, 1991; Thornton & Brotchie, 1987). Only a few studies have examined how reminiscence therapy affects self-conceptions. Botella and Feixas (Botella & Feixas, 1995) examined the effects of a directed, structured group in which participants recalled experiences and evaluated them with the help of facilitators by administering measures of personal constructs (idiosyncratic ways of thinking about other people) and ideal and current self-conceptions as baseline and outcome indicators. The results suggest that the reminiscence intervention reduced discrepancies between ideal and actual self-conceptions and led to changes in ways of thinking about other people, although the specific nature of these changes was not addressed. Other studies have shown that reminiscence and structured life review interviews can improve life satisfaction and self-esteem (Haight, 1991; Haight & Hendrix, 1995), are related to the achievement of ego integrity (Taft & Nehrke, 1990), and may even be associated with reduced morbidity and mortality (Stones, Rattenbury, & Kozma, 1995). Such findings appear to provide strong support for the idea that the coconstruction of autobiographical memory can produce changes in self-conceptions. However, reminiscence effects may be due to the group nature of the interventions or increased social interaction among the participants rather than to reconstructing the past (Haight, 1991). Further, people of all ages report reconstructing and rethinking past experiences in both conversational and nonconversational contexts (Staudinger, 1989; Webster, 1997), but reminiscence therapy is typically examined in elderly adults and within particular social settings (therapy groups or one-on-one interactions with a counselor). Thus, it is not clear whether reconstructing the past specifically promoted observed changes in selfconceptions, whether the effect would be similar in other age groups, or how much of the effect is due to the influence of counselors or the group on how experiences are remembered. Self-presentation: Carryover effects. Research on self-presentation examines the effects of presenting the self publicly on later privately held self-views (Gergen, 1965; Gergen & Wishnov, 1965; McKillop, Berzonsky, & Schlenker, 1992; Rhodewalt, 1998; Schlenker & Trudeau, 1990; Tice, 1992; Upshaw & Yates, 1968). Generally speaking, studies of this sort first assess self-perceptions of participants. Later, in ostensibly unrelated circumstances, participants present themselves to an audience in a self-discrepant way. The self-presentation may be relatively private or public. Thereafter, participants describe themselves again privately. A typical finding is that self-perceptions change in the direction of the discrepant self-presentations. The effect depends on the publicness of the self-presentation, the response of the audience, and the degree to which participants are concerned with social aspects of identity (Gergen, 1965; Gergen & Wishnov, 1965; McKillop et al., 1992; Tice, 1992; Upshaw & Yates, 1968). Further, changes in self-conceptions due to self-presentation are stronger when the self-presentation is drawn from participants' own past experiences (Tice, 1992). Theoretical explanations of carryover effects invoke two processes: cognitive dissonance and biased scanning (Rhodewalt, 1998). The former involves changing self-beliefs to accommodate self-presentations that are discrepant with existing self-views. Dissonance explanations apply to cases where the self-presentation is both public and perceived as voluntarily chosen. Both are the case 663 with typical reconstructions of past experiences in conversation. Thus, when individuals talk about the past in ways discrepant with their existing self-conceptions, they may experience dissonance and consequently change their existing self-views to accommodate the version of the past just presented. However, because we often recall experiences consistently with our current views of ourselves, not all conversational reconstructions will produce dissonance (Barclay & DeCooke, 1988; Conway & Pleydell-Pearce, 2000; Ross, 1989; W. J. Swann, 1987). On the other hand, self-presentations in the context of talk about past experiences necessarily involve biased scanning. Biased scanning refers to a biased recruitment of past experiences to support a particular self-presentation and has been viewed as causing transient rather than permanent change (Rhodewalt, 1998). However, one implication of Rhodewalt's review is that biased scanning could result in lasting changes both to the knowledge base from which memories can be reconstructed and to identity. This may be especially likely if the social context that elicited biased scanning in support of a particular self-presentation is repeated; many intimate relationships involve precisely that sort of repetition. In sum, audiences play a role in shaping recollections, and in doing so they also play a role in shaping identities. The outcome of this process can be stability in self-conceptions or change, because conversational reconstructions of the past can involve the rehearsal of existing self-conceptions or discarding old self-views and constructing novel ones. This aspect of the proposed model emphasizes one way in which conversational remembering can shape identity. However, conversational remembering does not occur in a static environment. Rather, it is a process subject to influences from other aspects of adult development. Developmental Changes in Conversational Remembering Both the ways in which those memories are coconstructed with listeners and the consistency with which memories are constructed across settings are likely to change with age across adulthood. Features of coconstruction are likely to change because of shifts in goals, storytelling skill, and concern with social partners. Further, and more generally, the balance between the two principles is likely to shift toward favoring consistency as people get older. With increasing age, individuals' goals for engaging in social contact change in predictable ways (Carstensen, 1995; Carstensen, Pasupathi, Mayr, & Nesselroade, 1999). Specifically, older adults are reliably more interested in and motivated by the emotional aspects of social encounters than are younger adults. This change may be reflected in the ways that older speakers tell stories as compared with younger speakers and in the ways that older and younger adults as listeners focus their interest and attention. Older adults appear to remember more of the emotional features of stories and to tell stories with a greater emphasis on meaning and interpretive features, in contrast to replicating propositional content or details (Adams, 1991; Carstensen & Turk-Charles, 1994). In intergenerational get-to-know-you conversations, older adults are more likely to discuss past experiences, including important and emotional ones, with partners (Boden & Bielby, 1983; Collins & Gould, 1994; Giles, Coupland, Coupland, Williams, & Nussbaum, 1992). Older married couples, as noted earlier, recall their vacations differently than younger couples, with a less itinerary- 664 PASUPATHI like focus and a greater emphasis on emotionally meaningful features of the experience (Gould & Dixon, 1993). Finally, older adults appear to emphasize emotional features even of relatively trivial laboratory experiences (Hashtroudi, Johnson, Vnek, & Ferguson, 1994). Importantly, this emotional focus may lead to reconstructing the past in self-justifying ways because it emphasizes the feelings and thoughts that led to a choice rather than the evidence base for that choice (Mather & Johnson, 2000). Older listeners have not been examined but would be expected to show more interest in emotional or emotion-focused stories than younger listeners. Thus, older adults, whether as speakers or listeners, are likely to produce and desire more emotional stories. In addition to developmental changes in general social motives, older adults also report different goals for remembering the past. Specifically, they report remembering in order to teach others and reminiscing in order to prepare for death (Webster, 1997) more frequently than younger adults. This implies that older adults will place greater emphasis on personal meaning and interpretation in recalling events from their past because such interpretive elements better fit these goals than detailed, fact-oriented styles of remembering. Consistent with this, some storytelling findings have suggested that older adults do engage in more interpretive, meaningdriven remembering (Adams, 1991; Adams, Labouvie-Vief, Hobart, & Dorosz, 1990). Changes in storytelling skill over adulthood may also affect the coconstruction of the past in conversation. Older adults may be better storytellers than younger adults (Kemper, Rash, Kynette, & Norman, 1990; Mergler & Goldstein, 1983), that is, better able to structure past experience in narrative forms that are interesting to an audience (Gould & Dixon, 1993). Thus, the stories produced by older adults are likely to be more entertaining and to gamer more positive responses from listeners. A final age change that may affect coconstruction is the level of concern older and younger adults have with the responses of their social partners. Older adults express less concern with what others think (Reifman, Klein, & Murphy, 1989), sometimes conform less than young adults (Pasupathi, 1999), and in happy marriages, exhibit greater tolerance for inattentive listening (Pasupathi, Carstensen, Levenson, & Gottman, 1999). These changes may reflect older adults' increased skills in emotion regulation, both in avoiding becoming distressed by negative behavior in the first place and in managing distress once it occurs (Carstensen, 1995; Carstensen et al., 1999). They also imply that older adults will be less prone to changing stories to please a listener. In fact, most of the research on coconstruction reported above focuses on young adults, a population uniquely concerned with what others think and susceptible to the types of social pressures that listeners may exert (Arnett, 2000; Sears, 1986). If older adults are less likely to change stories in response to their listeners, they may be more consistent in the way they recall events across different remembering contexts. Moreover, there are other reasons to expect increases in consistency with age over adulthood.9 First, older adults have demonstrated greater reliance on schematic remembering (Hess, 1994; Mather & Johnson, 2001). Thus, when older adults recall a memory in a schematic way, this schema may have greater influence on subsequent recall than in younger adults. Schematic reliance in reconstructing memories is increased for both older and younger adults using an affective focus and, as noted earlier, is more likely for older adults (Mather & Johnson, 2001). Second, age-related declines in source memory (Hashtroudi et al., 1994; Mather & Johnson, 2000) may make it more difficult for older adults, as compared with younger adults, to distinguish between the event as recounted and the event as experienced. Thus, recountings have even greater potential to replace memories of the event itself, making subsequent recountings more consistent with earlier ones. These declines are attributed to the tendency of older adults to focus on thoughts and feelings that are associated with both experiences and their tellings rather than perceptual details (e.g., smells, sights, sounds), which might distinguish between experiences and tellings. Recently, F. J. Anderson and colleagues (F. J. Anderson, Cohen, & Taylor, 2000) documented that older adults' memories for personal experiences are, in fact, more consistent over time than the memories of younger adults. They offer three explanations for their findings. One is schematic reliance. Another is the notion that older adults have precompiled versions of important personal memories, versions based in part on frequent rehearsal. Because these memories are being reproduced (in their compiled form) rather than reconstructed, they are less vulnerable to contextual influences on reconstruction and, hence, more consistent over time. Finally, F. J. Anderson and colleagues suggested that unstable self-conceptions in younger adulthood may lead to instability in the reconstruction of the past. Older adults have more stable self-conceptions, increasing the consistency of their memory constructions at any point in time. Notably, all three explanations can be found in some aspect of the present model: schematic reliance as an explanation for consistency of memories over time, frequent rehearsal as a contributor to precompiled versions of experience that are less open for coconstruction, and a more stable selfconcept that provides for consistency in remembering across occasions. Conversation, Memory, and Development: Summing Up Above, I have argued that talking about past experiences is a process by which our autobiographical memories are socially constructed. I proposed that talk about the past in conversation is coconstructed, and that subsequent memories for events talked about in conversation is coconstructed and that subsequent memories for events talked about in conversation are likely to be consistent with that socially constructed version. Thus, the content of autobiographical memory is a result of both experiences and social reconstructions of those experiences. Later, I suggested that conversing about past experiences both influences and can be influenced by adult development. Socially constructing the past may promote either continuity or change in identity across adulthood. Aging may be associated with shifts toward the emotional and interpretive in remembering. Further, aging is associated with an increasing likelihood of consistency in remembering. Thus, even as the social construction of memories can shape identity across adulthood, age can bring changes in both coconstruction and consistency. 9 Note that the principle of consistency does not refer to the extent to which a conversational recollection is shaped by the contributions of the listener or speaker, but rather the extent to which, once constructed, a recollection influences later memories. SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT I noted earlier that this framework touches base with many areas of psychology. It is consistent with the notion that memory be studied in the social contexts in which it is used, as many others have pointed out (Adams et al., 2001; Edwards, 1997; Engel, 1999; Hirst, 2001; Neisser, 1982; Weldon, in press). It also connects with a growing emphasis on the social contexts within which adult development unfolds (Baltes & Carstensen, 1998; Baltes & Staudinger, 1996), with recent work in psycholinguistics (Clark, 1996), and with discursive and constructivist approaches to psychology (Davies & Harre, 1990; Edwards, 1997; Gergen & Gergen, 1988). It is not surprising that several recent proposals about autobiographical memory and the self raise related issues. 665 Comparisons With Other Recent Models Many have argued for the constructive and contextual nature of memory and its connections to the self (Conway & PleydellPearce, 2000; Engel, 1999; Fivush, 1995; Hirst et al., 1997; McAdams, 1996; Schank & Abelson, 1995; Thome, 2000; Woike et al., 1999). Below, I address these theoretical frameworks in three sets: those dealing with self-memory relationships in adults, those dealing with the childhood development of autobiographical remembering, and those dealing with the development of the life story. The present framework is largely complementary in highlighting issues related to the conversational context of remembering and listeners. Self-memory systems. Conway and Pleydell-Pearce (2000) proposed that autobiographical memories are constructed from the interplay of two systems: the self-system, which houses ongoing goals and concerns, and the autobiographical knowledge base, which houses some record of past experiences. As a consequence, memory grounds the pursuit of goals in everyday life, and these same goals determine which memories are retrieved and how they are reconstructed. Their model attempts to explain the constructive nature of autobiographical memory and some of its systematic features. In contrast, the present model is focused specifically on conversational reconstruction of the past and raises some novel and important issues. First, some goals are shared by more than one person, such as maintaining a relationship or collaboratively solving a problem. Second, listeners can induce competing goals within the speaker, such as conflicts between telling the truth and appearing a good or competent person. Third, listeners have their own goals that may conflict with those of speakers. Finally, listeners also have non-goal-related impact. They add information that speakers may not have in their autobiographical knowledge base, call that knowledge base into question by remembering alternative versions of reality, change the focus of speakers' retrieval strategies, and otherwise alter the discursive reconstruction of events and speakers' subsequent knowledge. Consideration of the conversational context of remembering thus raises an important additional set of issues that is not entirely captured by the notion of a self-memory system. Childhood development of autobiographical memory. Child developmental researchers have noted intimate connections between the self and autobiographical remembering (Fivush, 1991; Howe et al., 1994; Miller et al., 1992; Nelson, 1991). In addition, they have accumulated considerable evidence that parents' reconstruction of past events in conversations with children serves to shape children's later styles of remembering, including gendered styles (Fivush, 1998; Haden, Haine, & Fivush, 1997; Kuebli, Butler, & Fivush, 1995; Reese & Fivush, 1993). In fact, gender differences in the ways that parents and children reconstruct experiences conversationally have intriguing connections to gender differences in adult autobiographical memory (Fivush, 1998; Pillemer, 1998). Finally, some, though not all, studies have shown that the way children converse about events on one occasion influences the way they come to recall those events on later occasions (Fivush, 1994; Tessler & Nelson, 1994), and other findings have examined how children make statements about themselves in the context of conversing about their experiences (Miller et al., 1992). The present framework is consistent with these views as applied to adult development. That is, I am arguing that our audiences for reconstructions influence both concurrent and long-term memory and that this process is tied to the construction of self-conceptions. Child developmental research has largely presumed connections between memory and self, most likely because assessment of self-conceptions is problematic in very young children. Although identity and memory are clearly related, in adults they are not the same because individuals may be able to describe themselves accurately or consistently on personality measures even in the absence of access to personal memories (Kihlstrom & Klein, 1994; S. B. Klein, Loftus, & Kihlstrom, 1996). Once self-conceptions are established, they may become independent of the memories and experiences on which they were initially based. In sum, extending the conversational approach to autobiographical memory to adulthood provides needed continuity with child developmental research and permits examination of identity-memory connections more directly than is possible in very young children. Personal storytelling and personal development. Thorne (2000) proposed that personal storytelling is a causal force in personality development. In doing so, she suggested that stories about personal experiences can change people's developmental trajectories and that these stories are influenced by listeners. The present model can be viewed as an extension of these ideas, one that establishes a set of potential processes by which listeners exert their effects, that permits laboratory tests of how listeners change the stories people tell over the short and long term, and that permits examination of the circumstances under which those changes may be lasting. In addition, although the present review necessarily focused on narrative recall because that is what has been studied, a conversational perspective implies consideration of nonnarrative forms of recalling the past in conversation. As noted earlier, event recollections in conversation do not always take the form of stories, although the sort elicited in interviews and studied by psychologists often do. Finally, Thome's review focuses most heavily on speaker contributions to personal storytelling by emphasizing issues such as a speaker's readiness to tell or the need for speakers to reconcile traits and stories. In contrast, the present framework draws attention back to the listener. Implications for Existing Phenomena: The Reminiscence Bump This model also offers new ways to examine the reminiscence bump (Rubin & Schulkind, 1997), or the tendency for individuals freely recalling past experiences to recall an unusually high num- 666 PASUPATHI ories from the bump period and from other periods. The fact that those frequency ratings do not differ could be explained in at least three ways, as noted by Rubin and Schulkind. First, it may be that rehearsal simply does not explain the reminiscence bump. Second, perhaps there is some rehearsal threshold for long-term retention of event memories, and perhaps more of the total pool of bump period experiences are rehearsed above this threshold. Thus, memories from the bump period and from other periods may be equally rehearsed, but more events from the bump period are rehearsed enough to be retained. This account is consistent with the present framework and would be supported by the kind of studies outlined above. Finally, people may be poor at estimating the frequency with which they rehearse experiences, in which case other measures of rehearsal are needed. To recap, the present framework complements existing models by drawing attention to adult development and to listeners' effects on the construction of memories and by providing a set of cognitive and social research findings that may account for how listeners influence memories. It also speaks to new ways of looking at the reminiscence bump. But I believe the greatest strength of this framework lies in new directions for investigation. New Directions for Future Research There are at least four new areas for research raised by this framework. A first set of open questions concerns the dynamic tension between the malleability of memory (evidenced by coconstruction) and its stability over time (consistency). Personal memories eventually become resistant to change (Johnson & Chalfonte, 1994; Koutstaal et al., 1998; McClelland, 1994), suggesting that eventually the principle of consistency outweighs that of coconstruction. The impact of postevent information on memory in eyewitness paradigms also depends on the timing of that information relative to the initial event (Belli & Loftus, 1996), and social cognition findings reveal similar patterns (Higgins et al., 1982; McGregor & Holmes, 1999). What is the time course in which additional conversational retellings can change memories? Does the dynamic interplay between coconstruction and consistency vary for detailed recall, gist recall, or interpretive features of memories? Untangling this set of questions will likely require a combination of diary approaches that assess telling situations something like those used by Marsh and Tversky (2001), as well as laboratory paradigms that permit control of the circumstances of retelling and subsequent recollections. It seems safe to say that initial conversational retellings will have a greater chance for influencing later memory than later ones, particularly in terms of the sorts of facts and details that individuals retain. A second area open for research involves prospective relationships between reconstructing the past and change and stability in self-conceptions or goals. There is considerable evidence that our current self-conceptions and goals influence the way we remember the past (Conway & Pleydell-Pearce, 2000; Woike et al., 1999; Woike, 1995). There is less evidence that remembering the past in particular ways can shape the way we think of ourselves. In part, this will involve examining components of memories like selfrelated or relationship-related statements rather than memory for veridical details. Narratives about past experiences frequently contain self-related statements, as when people say "I'm the type of person who needs to be appreciated" in justifying anger at an ber of experiences that occurred between the ages of 10 and 30. The bump has been explained in terms of the life story; the emergence of enduring dispositions, goals, and relationships during this period of life; and the developmental focus of this age period on identity construction (Conway & Pleydell-Pearce, 2000; Thorne, 2000). The present framework, combined with changes in social networks over the life span, provides a complementary explanation. For those growing up in the industrialized world, the period between 10 and 30 years of life is possibly the most open to new social contacts, both from the point of view of individual social motivations (Carstensen, 1995) and from a simply cultural or institutional viewpoint. It is during this period that people make their first independent choices about schoolwork, leisure pursuits, and friends; go to college, begin their jobs, or both; meet their future spouse and partner; and begin child rearing. All of these changes involve new social contacts that require the repeated rehearsal of one's experiences by way of self-presentation and in order to build intimacy. This process can be seen as one of identity construction (Thorne, 2000)--certainly it requires the capacity to construct a life story (Habermas & Bluck, 2000)--but must also be viewed in the social and cognitive terms implied by the present analysis. In other words, not only are adolescents and young adults more likely to attempt identity construction in conversation by talking about their experiences, their social milieu is one that will require them to do so more often than that of typical middle-aged and older adults. For those past the emergent phase of adulthood (Arnett, 2000), social contacts tend to involve familiar and longterm partners who do not require a life story and are probably invested in the stability of an individuals' existing story (Carstensen, 1995; De La Ronde & Swann, 1998; Lang, Staudinger, & Carstensen, 1998). Of importance, the high social demand for telling in adolescence and emergent adulthood is not inconsistent with other developmental accounts, such as the need for self-construction in this phase of life. It simply points to the fact that what young adults set out to do in the way of identity construction is also elicited by and subject to the responses and interests of their social worlds. This reasoning suggests there is a set of people who have additional bumps--those whose social networks have been disrupted at a later phase, requiring the building of new social connections. One example would be adults who divorce midlife, representing a group with largely stable identities but with the sudden need to seek new social contacts or renew interrupted ones and the need to articulate a new life story (presumably including recent events). Interestingly, such a group probably varies in the degree to which the divorce required many new social experiences, thus providing a within-group potential for comparison of reminiscence bumps. It can also be argued, of course, that such individuals may change their identities as well, but careful studies might be able to examine reminiscence bumps, new social contacts, and identity stability simultaneously to evaluate this prediction. The possibility that only the formative young adult years would be retold in the service of making new contacts cannot be ruled out, and existing evidence, though imperfect, has not supported a rehearsal explanation for the reminiscence bump (Rubin & Schulkind, 1997). Such evidence primarily relies on individuals' ratings of the frequency with which they think about or talk about mem- SOCIAL CONSTRUCTION AND ADULT DEVELOPMENT insensitive partner or "I am normally very cautious" while recounting a careless driving error. Such statements likely produce stability in self-conceptions, either by rehearsing the self-conceptions or by marking the event as an exception that need not change an existing self-conception. Another route involves applying reminiscence methods, and still another fruitful approach involves developing experimental analogues for studying how memory construction influences subsequent memory and self-conceptions. An analog study would involve providing experiences and conversational recollections in the laboratory and assessing self-views relevant to those experiences. Such paradigms provide a kind of simulation of the developmental process proposed to occur in everyday contexts. Third, if telling about experiences may have such important consequences for memory and development, the issue of who tells what to whom, how often, and why arises. Surprisingly little research addresses these questions. Most of it comes from literature on self-disclosure (Monck, 1991; Pennebaker, 1997; Rime, 1995; Wegner & Lane, 1995), which tends to examine emotional, secret, or otherwise special events and the healing power of disclosing these events (Pennebaker, 1997). But we also talk about the mundane, everyday events of life, and this talk may also build our sense of self and relations to others (Barclay & DeCooke, 1988; Duck, 1994). Initial work by Hyman and Faries (1992) suggests that tellings are frequent, involve emotional events, and often serve rhetorical functions or describe or present the self. More recent work in this area (Marsh & Tversky, 2001), reveals some additional interesting patterns. Their data, which were based on students' diaries of event recollections in conversation, provide three conclusions relevant to the framework proposed above: (a) There is substantial distortion in recollections of personal experiences, (b) distortion appears to vary as a function of the audience and (c) there are a surprising number of distortions that people make while continuing to view their recollections as true. This is quite stunning and, given the likelihood that distortions will change later memory, provides good evidence that audiences are shaping the memories these students have of their lives. Further, Marsh and Tversky (2001) also provide evidence for multiple purposes in telling about events and for event characteristics (emotionality) associated with telling. The developmental literature reviewed earlier suggests that who tells what to whom and why will vary as a function of age, and the literature on speakers' enduring qualities points to gender, culture, and enduring motivations and beliefs as other important factors to consider. Fourth, the framework above also raises questions about relationships between emotion and memory. Such relationships are typically examined in terms of emotions at encoding, retrieval, or both (Christianson, 1992), or in terms of memory for emotions experienced during an event (Levine & Bluck, 1997; Thompson, 1998). A conversational approach raises the possibility that emotions at the initial event and emotions experienced at recounting the event could differ in valence and intensity. We may receive great social rewards for an entertaining tale of a near-death experience or an embarrassing accident in a classroom. The initial events were almost certainly emotionally negative, whereas the recounting provides positive emotional associations with those same events. Contrast this with stories about loss or bereavement, which may be far less enthusiastically received even if elegantly and beautifully told (Pennebaker, 1997). This layering of poten- 667 tially contradictory emotions in successive event recollections represents another novel direction for future work one that stems largely from attention to the social context of remembering. A Temporary End to the Story In this article, I have tried to tell a story about conversation, memory, and development. Talk about the past is certainly not the whole story of adult development, but it is an important one that deserves telling. Conversations about our pasts are a way that the social world can migrate from an abstract social context toward a destination in the center of our identity and at the core of our relationships with others. If "developmental change arises in everyday communication" (Fogel, 1993, p. 5), then I hope I have told one version of how our present and future emerge in the social reconstruction of our pasts. References Adams, C. (1991). Qualitative age differences in memory for text: A life-span developmental perspective. Psychology and Aging, 6, 323-336. Adams, C , Labouvie-Vief, G., Hobart, C. J., & Dorosz, M. (1990). Adult age group differences in story recall style. Journal of Gerontology, 45, P17-P27. Adams, C , Smith, M. C , Pasupathi, M., & Vittolo, L. (2001). Adult age-group differences in story recall: Does the listener make a difference? Manuscript submitted for publication, University of Oregon, Eugene. Anderson, F. J., Cohen, G., & Taylor, S. (2000). Rewriting the past: Some factors affecting the variability of personal memories. Journal of Applied Cognitive Psychology, 14, 435-454. Anderson, J. R., & Bower, G. H. (1973). Human associative memory. Washington, DC: Winston. Anderson, M. C , Bjork, R. A., & Bjork, E. L. (1994). Remembering can cause forgetting: Retrieval dynamics in long-term memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1063-- 1087. Anderson, R. C , & Pichert, J. W. (1978). Recall of previously unrecallable information following a shift in perspective. Journal of Verbal Learning and Verbal Behavior, 17, 1-12. Andersson, J., & Ronnberg, J. (1995). Recall suffers from collaboration: Joint recall effects of friendship and task complexity. Applied Cognitive Psychology, 9, 199-211. Arnett, J. J. (2000). Emerging adulthood: A theory of development from the late teens through the twenties. American Psychologist, 55, 469480. Asch, S. E. (1952). Effects of group pressure upon the modification and distortion of judgments. In H. Guetzkow (Ed.), Groups, leadership, and men (pp. 393-401). Pittsburgh, PA: Carnegie Press. Bahrick, H. P., Hall, L. K., & Berger, S. A. (1996). Accuracy and distortion in memory for high school grades. Psychological Science, 7, 265-271. Baltes, M. M., & Carstensen, L. L. (1998). Social psychological theories and their applications to aging: From individual to collective selective optimization with compensation. In V. L. Bengston & K. W. Schaie (Eds.), Handbook of theories of aging (pp. 209-226). New York: Springer. Baltes, P. B., & Staudinger, U. M. (Eds.). (1996). Interactive minds: Life-span perspectives on the social foundations on cognition. Cambridge, England: Cambridge University Press. Barclay, C. R. (1996). Autobiographical remembering: Narrative constraints on objectified selves. In D. Rubin (Ed.), Remembering our past: Studies in autobiographical memory (pp. 94-128). New York: Cambridge University Press. 668 PASUPATHI Collins, C. L., & Gould, O. N. (1994). Getting to know you: How own age and other's age relate to self-disclosure. International Journal of Aging and Human Development, 39, 55-66. Conway, M. A. (1995). Flashbulb memories. East Sussex, NJ: Erlbaum. Conway, M. A., & Pleydell-Pearce, C. W. (2000). The construction of autobiographical memories in the self-memory system. Psychological Review, 107, 261-288. Davies, B., & Harre\ R. (1990). Positioning: The discursive production of selves. Journal for the Theory of Social Behavior, 20, 43-63. Davis, C. G., Nolen-Hoeksema, S., & Larson, J. (1998). Making sense of loss and benefiting from the experience: Two construals of meaning. Journal of Personality and Social Psychology, 75, 561-574. Davis, P. J. (1999). Gender differences in autobiographical memory for childhood emotional experiences. Journal of Personality and Social Psychology, 76, 498-510. De La Ronde, C , & Swann, W. B. (1998). Partner verification: Restoring shattered images of our intimates. Journal of Personality and Social Psychology, 75, 374-382. de Rivera, J. (1997). The construction of false memory syndrome: The experience of retractors. Psychological Inquiry, 8, 271-292. Dickinson, C , & Givon, T. (1995). Memory and conversation: Toward an experimental paradigm. In T. Givon (Ed.), Conversation: Cognitive, communicative, and social perspectives (pp. 91-132). Amsterdam, the Netherlands: John Benjamins. Dixon, R. A., & Gould, O. N. (1996). Adults telling and retelling stories collaboratively. In P. B. Baltes & U. M. Staudinger (Eds.), Interactive minds: Life-span perspectives on the social foundation of cognition (pp. 221-241). Melbourne, Australia: Cambridge University Press. Duck, S. (1994). Steady as (s)he goes: Relational maintenance as a shared meaning system. In D. J. Canary & L. Stafford (Eds.), Communication and relational maintenance (pp. 45-60). San Diego, CA: Academic Press. Duck, S. W., Rutt, D. J., Hurst, M. H., & Strejc, H. (1991). Some evident truths about everyday conversation: All communications are not created equal. Human Communication Research, 18, 228-267. Edwards, D. (1997). Discourse and cognition. London: Sage. Edwards, D., & Middleton, D. (1986a). Joint remembering: Constructing an account of shared experience through conversational discourse. Discourse Processes, 9, 423-459. Edwards, D., & Middleton, D. (1986b). Text for memory: Joint recall with a scribe. Human Learning, 5, 125-138. Edwards, D., & Middleton, D. (1988). Conversational remembering and family relationships: How children learn to remember. Journal of Social and Personal Relationships, 5, 3-25. Engel, S. (1999). Context is everything: The nature of memory. New York: Freeman. Ervin-Tripp, S. M., & Kuentay, A. (1995). The occasioning and structure of conversational stories. In T. Givon (Ed.), Conversation: Cognitive, communicative, and social perspectives (pp. 133--166). Amsterdam, the Netherlands: John Benjamins. Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117-140. Festinger, L. (1957). A theory of cognitive dissonance. Evanston, IL: Row, Peterson. Fivush, R. (1991). The social construction of personal narratives. MerrillPalmer Quarterly, 37, 59-82. Fivush, R. (1994). Young children's event recall: Are memories constructed through discourse? Consciousness and Cognition, 3, 356-373. Fivush, R. (1995). Language, narrative, and autobiography. Consciousness and Cognition, 4, 100-103. Fivush, R. (1998). Gendered narratives: Elaboration, structure, and emotion in parent-child reminiscing across the preschool years. In C. P. Thompson, D. J. Herrmann, D. Bruce, J. D. Read, D. G. Payne, & M. P. Barclay, C. R., & DeCooke, P. (1988). Ordinary everyday memories: Some of the things of which selves are made. In U. Neisser & E. Winograd (Eds.), Remembering reconsidered: Ecological and traditional approaches to th...