Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

In this chapter, we describe emerging theory and evidence on the relational self within the social and personality psychology literatures. As will become apparent, there is some variation in these literatures in how the term “relational self” is used, but broadly speaking, the relational self refers to aspects of the self associated with one’s relationships with significant others (e.g., romantic partners, parents, friends). To preview, in the first part of our chapter, we delineate the key assumptions of our recent, integrative conceptualization of the relational self (Chen et al., 2006). In developing this conceptualization, we have drawn from several theoretical perspectives on the self and significant others, but most notably from a transference perspective (Andersen & Chen, 2002). As such, in the second part of our chapter, we describe this particular perspective in detail, while including a range of supporting evidence.

In the third part of our chapter, we review several other prominent theoretical perspectives on the self and significant others, with an emphasis on their points of convergence with and divergence from the transference perspective and, in turn, our integrative conceptualization of the relational self. First, we describe work on relational schemas and attachment theory, both of which offer unique contributions to our broader conceptualization. We then turn to work on inclusion of others in the self and the relational-interdependent self-construal, two theoretical perspectives that provide a useful point of comparison for our conceptualization of the relational self. In the final part of our chapter, we discuss relations between the relational self and other aspects of identity (e.g., cultural identity), as well as some important directions for future research.

An Integrative Conceptualization of the Relational Self

In a recent review article, we offered an integrative conceptualization of the relational self aimed at bringing together the unique facets of several different theories on the self and significant others (Chen et al., 2006). This conceptualization draws especially heavily not only from social-cognitive work on transference, but also from work on relational schemas and attachment theory. According to our conceptualization, the relational self reflects who a person is in relation to his or her significant others. Put differently, we propose a “self-within-relationships” viewpoint on the relational self, which contrasts with other perspectives, to be discussed in later sections, that take a more “relationships-within-the-self” viewpoint on the self and significant others.

Our integrative conceptualization can be summarized in four key assumptions: (a) the relational self is comprised of knowledge about the self when relating to significant others, where this knowledge is linked in memory to stored information about significant others; (b) the relational self exists at multiple levels of specificity (i.e., relationship-specific, generalized, global); (c) the relational self is capable of being contextually or chronically activated—that is, accessed from memory; and (d) the relational self is composed of self-conceptions as well as a constellation of other self-aspects (e.g., goals, self-regulatory strategies) that characterize the self when relating to significant others (Chen et al., 2006). Below, we elaborate on each of these assumptions.

Linkages Between Stored Self-Knowledge and Significant-Other Knowledge

Our integrative model is grounded in basic social-cognitive theory and research on knowledge representation and use (e.g., Higgins, 1996a). Specifically, we propose that relational selves are composed of people’s stored knowledge about the self in the context of their relationships with significant others (e.g., who I am when relating to my brother) that is distinct from, but linked in memory to, people’s stored knowledge about their significant others.

The view that relational-self and significant-other knowledge are distinct is important because, as noted above, it sets our view of the relational self as the self in relation to significant others apart from theories, to be discussed below, that posit the incorporation of aspects of significant others into the self-concept (e.g., Aron, Aron, Tudor, & Nelson, 1991). To illustrate, a person may be submissive around her mother—in other words, this person’s relational self with her mother may be characterized by submissiveness—but this need not imply that submissiveness is an attribute of the person’s mother that has been incorporated into the person’s self-concept. Indeed, the person’s mother may or may not be submissive.

Multiple Levels of Specificity

Consistent with a large social-cognitive literature indicating that abstract, general social knowledge, as well as highly specific forms of social knowledge, are stored in memory (e.g., Smith & Zárate, 1992), we maintain that most people possess multiple relational selves and that these selves exist at varying levels of specificity. A relationship-specific relational self designates the self in relation to a specific significant other (e.g., the self in relation to one’s father), whereas a generalized relational self is akin to a summary representation of the self in the context of multiple relationships. These relationships may involve either a single, normatively defined relationship domain (e.g., the self when relating to one’s family members) or idiosyncratic groupings of relationships (e.g., the self when relating to one’s book club friends). Finally, people may possess a global relational self which denotes conceptions and aspects of the self in relation to significant others as a general class of individuals.

Although different theoretical perspectives on the self and significant others have tended to examine different levels of specificity of relational selves, results reflecting relationship-specific, generalized, and global working models generally converge, thereby increasing the likelihood of parallels in the responses associated with relational selves at different levels of specificity (e.g., Baldwin, Keelan, Fehr, Enns, & Koh-Rangarajoo, 1996; Mikulincer & Arad, 1999). Nonetheless, maintaining distinctions between different levels of specificity is important insofar as the particular content of a given person’s relational selves may vary widely across different levels of specificity (e.g., a person’s relational self with his spouse may differ sharply from his relational self with significant others in general). Thus, the predictions one can make about a given relational self can only be as specific or broad as the particular relational self in question.

Paralleling the different levels of specificity possible for relational selves, the significant-other knowledge to which relational-self knowledge is linked may also vary in specificity. Thus, relational-self knowledge may be linked in memory to knowledge about a specific significant other (e.g., one’s oldest sister), knowledge abstracted from experiences with multiple significant others to whom the self relates similarly (e.g., all of one’s siblings to whom one relates in a similar manner), or knowledge about significant others in general. Regardless of level of specificity, because relational-self knowledge is linked to significant-other knowledge, when significant-other knowledge is activated, this activation should spread by association to the relevant relational-self knowledge.

Activation of Relational Selves

In terms of when relational selves are activated, our conceptualization draws on the notion of the working self-concept (Markus & Kunda, 1986), or the set of self-aspects that are accessible from memory in the current context. Thus, not all self-knowledge is accessible at the same time; rather, only a subset of one’s overall pool of self-knowledge is accessible in any given context. Our model posits that cues in the immediate context that denote the actual, imagined, or symbolic presence of a significant other, or the relational dynamics between the self and this other, should alter the content of the working self-concept, just as other contextual cues do (e.g., McGuire & Padawer-Singer, 1976). Specifically, such cues should shift the working self-concept toward the relevant relational self. For example, a phone call from a significant other, a whiff of his or her cologne, or facing circumstances (e.g., threat) that call for the other’s support, may all serve as contextual activators of relevant relational-self knowledge.

Alongside these working self-concept assumptions, we adhere to the principle that repeated, frequent activation of any stored knowledge construct increases its baseline or chronic level of accessibility; that is, its activation readiness (e.g., Higgins, King, & Mavin, 1982). The higher the chronic accessibility of a construct, the less contextual cueing is required to activate it (Higgins & Brendl, 1995). Thus, although the activation of relational selves is influenced by immediate, contextual cues, frequent contextual activation of a relational self should result in its chronic accessibility, rendering it more likely to be activated even in the absence of contextual cueing (e.g., Andersen, Glassman, Chen, & Cole, 1995).

Content of Relational Selves

Finally, our integrative model assumes that relational selves are composed of attribute-based (e.g., fun-loving) and role-based (e.g., authority figure) conceptions of the self in the context of the relevant significant others. These attributes and roles are derived idiographically (from unique experiences) as well as from normative, cultural role prescriptions that are similar across people (Clark & Mills, 1979; Fiske, 1992). Importantly, relational selves are also thought to contain affective material, goals and motives, self-regulatory strategies, and behavioral tendencies. That is, they include positive and negative evaluations of the self in relation to a given significant other, the affect one experiences when relating to the other, the goals one pursues in the relationship with the other, the self-regulatory strategies one uses in interactions with the other, and the behaviors one enacts toward the other. Thus, for example, a person’s relational self with his critical mother may be composed of conceptions of the self as inferior, feelings of rejection, and the goal to please.

To summarize, drawing from various perspectives on the self and significant others (to be reviewed below), we propose that relational selves refer to conceptions and aspects of the self specifically in the context of relationships with specific or multiple significant others, or with significant others in general (Chen et al., 2006). Due to linkages between relational-self and significant-other knowledge, relational selves can be made accessible through activating significant-other knowledge. Reflecting working self-concept notions, relational selves are activated by contextual cues, although the chronic accessibility of a relational self increases its likelihood of being activated across contexts. When a relational self is activated, a person not only conceives of and evaluates himself or herself as when relating to the relevant significant other(s), but he or she also exhibits associated affective, motivational, self-regulatory, and behavioral responses.

The Relational Self: A Transference-Based Perspective

As noted, our conceptualization reflects an integration of several theories on the self and significant others, but draws the most from theory and research on transference. As such, in this section we review the transference perspective on the relational self in detail.

Transference refers to the phenomenon whereby aspects of past relationships re-surface in encounters with new others. Clinicians (Freud, 1958; Sullivan, 1953) view transference as a tool that therapists use to encourage clients to transfer their thoughts, feelings, and habitual patterns of behavior with significant others onto the therapist, as a means of fostering client insight into, and improvement of, maladaptive relational patterns. In contrast, the social-cognitive model of transference (Andersen & Glassman, 1996; Chen & Andersen, 1999, 2008) specifies the cognitive structures and processes that account for the occurrence of transference in everyday social interactions. Below we describe this model, flesh out the role of the relational self within it, and offer a sampling of evidence in support of the transference approach to the relational self.

The Social-Cognitive Model of Transference

The social-cognitive model of transference (Andersen & Glassman, 1996; Anderson & Cole, 1990; Chen & Andersen, 1999, 2008) assumes that mental representations of past and current significant others—such as a parent, romantic partner, close friend, or sibling—are stored in memory. These representations are akin to warehouses of knowledge about these important individuals, including beliefs about their physical and personality attributes, as well as their internal states, such as their thoughts and feelings (Andersen, Glassman, & Gold, 1998; Chen, 2003). Given that many of our goals, thoughts, and feelings hinge on significant others, representations of these individuals are laden with affect and motivation.

According to the social-cognitive model of transference, transference occurs when a perceiver’s representation of a significant other is activated in an encounter with a new person—for example, due to the person’s physical resemblance to the significant other, or the overlap of his or her personality attributes or role vis-à-vis the self with those of the significant other. Upon such activation, the perceiver interprets the new person in ways consistent with the knowledge stored in his or her representation of the relevant significant other, and responds to the person in ways derived from his or her relationship with the specific other. Although the activation and use of significant-other representations reflect nomothetic processes, or processes that operate similarly across people, the content and meaning of significant-other representations is thought to be idiographic.

Research has shown that significant-other representations are chronically accessible, indicating that they are in a constant high state of activation readiness, or readiness to influence social perception, judgment, and behavior (e.g., Higgins & King, 1981). Nonetheless, transient or temporary cues in the environment—such as cues based on priming specific features of one’s significant other—further increase the accessibility of these representations (Andersen et al., 1995). In each case, the “match” between stored knowledge and the present cues heightens the likelihood of knowledge activation and use (Higgins, 1996a). In most research on transference, the to-be-interpreted stimulus person has been described as having some of the attributes of the relevant significant other as a way of increasing the accessibility of the corresponding significant-other representation (e.g., Andersen & Baum, 1994; Chen, Andersen, & Hinkley, 1999). In other words, the attribute-based resemblance of the stimulus person to the significant other elicits transference.

The research paradigm that has been used to empirically examine transference involves two sessions. In the first, pretest session, participants name a significant other (e.g., parent, romantic partner), and then generate descriptors about this person. Several weeks later, in an ostensibly unrelated session, participants are led to believe there is another participant with whom they will later have an interaction (in most studies, the other participant does not actually exist). Participants then go through a learning phase in which they are presented with descriptors allegedly about their upcoming interaction partner. For participants in the “Own Significant Other” condition, some of these descriptors are derived from those that they previously generated about their significant other. That is, their interaction partner is described as resembling their own significant other, thus eliciting transference. Participants in the “Yoked Significant Other” condition, on the other hand, are shown descriptors about the significant other of another participant in the study, and thus transference is not elicited. “Own” and “Yoked” participants are randomly paired on a one-to-one basis so that the descriptors used across “Own” and “Yoked” conditions are identical, but they are significant only to participants in the “Own” condition.

After the learning phase, transference is assessed using one or both of two standard measures. One is a recognition-memory test that measures representation-derived inferences about the interaction partner. Such inferences are indexed by participants’ confidence that they learned descriptors about the partner that are true of their significant other, but that were not actually presented in the learning phase. Thus, this measure taps how much participants use stored knowledge about their significant other to make inferences about the partner. The other measure, which asks participants to evaluate their interaction partner based on what they have learned about them in the learning phase, assumes that the positive or negative affect associated with significant others is elicited upon the activation of a significant-other representation (Fiske & Pavelchek, 1986). Evidence for transference on this measure takes the form of “Own” participants evaluating the interaction partner significantly more positively (negatively) when the partner resembles their own positively (negatively) evaluated significant other, with no such difference observed among “Yoked” participants. In short, this measure taps how much participants’ evaluations of their partner are influenced by evaluations of their significant other.

Nearly two decades of research has documented transference using these measures. Included in this body of work is evidence that transference may occur automatically (e.g., Glassman & Andersen, 1999). In other words, perceivers need not consciously draw analogies between significant and newly encountered people for transference to occur. Indeed, given the chronically high activation readiness of significant-other representations, it is likely that transference typically occurs automatically (Andersen, Reznik, & Glassman, 2004; Chen, Fitzsimons, & Andersen, 2006). Importantly, in addition to inferences and evaluations of new others derived from significant-other representations, transference elicits a myriad of consequences for the self. In the following section, we summarize transference effects reflecting people’s responses when relating to their significant others—in other words, responses reflecting relational selves.

Transference and the Relational Self

Expanding on the social-cognitive model of transference, Andersen and Chen (2002) presented a theory of the relational self in which they posited that every individual possesses a repertoire of relational selves, each reflecting aspects of the self when relating to a particular significant other. Moreover, as reflected in our broader conceptualization (Chen et al., 2006), we argued that significant-other representations and relational selves are linked in memory by knowledge reflecting the typical patterns of relating to the significant other. Because of such linkages, when a significant-other representation is activated, this activation spreads to the relevant relational self. In working self-concept terms, transference elicits a shift in the self-concept toward the person one is when relating to the significant other. Thus, just as non-significant-other contexts (e.g., the office) elicit shifts in the self-concept toward aspects of the self relevant to the particular context (e.g., job-related aspects of the self), significant-other contexts elicit shifts toward relational aspects of the self.

As in our integrative conceptualization, Andersen and Chen (2002) maintain that relational-self knowledge includes both attribute- and role-based aspects of the self with significant others. That is, relational selves include the positive and negative self-evaluations, affect, goals, self-regulatory strategies, and behaviors that are typically experienced or exhibited in relation to significant others. Finally, like significant-other representations, the content of relational selves is not only thought to be largely idiographic, unique to each individual, but also includes socially shared facets, such the role enacted with significant others (e.g., parent).

Evidence for the Relational Self in Transference

Below we present a sampling of findings from research on transference and significant-other representations more generally that provide support for the transference perspective on the relational self (for additional evidence, see Chen & Andersen, 2008; Chen et al., 2006).

Self-definition and self-evaluation. As noted, the transference perspective predicts that when a significant-other representation is activated, the working self-concept should be infused in part with associated relational-self knowledge. Thus, the perceiver should define and evaluate his/her self in line with the relational self that has been activated. For example, participants in one study were asked to perform five different feature-listing tasks in the pretest session of the transference research paradigm described above (Hinkley & Andersen, 1996). First, they were asked to list features to describe themselves as a baseline self-concept measure. They then listed features to describe both a positively and negatively regarded significant other, after which they listed features to describe themselves when relating to each significant other, as a baseline measure of each relational self. In the ostensibly unrelated second session of the transference paradigm, participants were presented with descriptors about a new person who either did (“Own” condition) or did not (“Yoked” condition) resemble the positively or negatively evaluated significant other they described in the pretest session. They then listed descriptors to characterize themselves at that moment—as a measure of their working self-concept—and classified each descriptor as positive or negative—as a measure of their self-evaluation.

To assess shifts in the working self-concept toward the relational self, Hinkley and Andersen (1996) first calculated the overlap between the features listed in participants’ baseline working self-concept and in each relational self. Controlling for this pretest overlap, participants in the “Own” condition, for whom transference should be elicited due to the new person’s resemblance to one of “Own” participants’ significant others, showed a greater shift in their working self-concept toward the relevant relational self, relative to “Yoked” participants—a finding that held for both positively and negatively evaluated significant others. Turning to self-evaluation, Hinkley and Andersen summed the positive and negative classifications that participants ascribed to those features listed in the second session that overlapped with their pretest relational self. “Own” but not “Yoked” participants evaluated these overlapping descriptors more positively when the new person resembled their positive, rather than negative, significant other. Hence, when transference occurs, both self-definition and self-evaluation shift to reflect the relevant relational self.

Other researchers have documented self-evaluative processes associated with relational selves simply by activating a significant-other representation. For example, research has shown that activating a significant-other representation leads people to stake their self-worth in domains they believe are valued by the significant other (Horberg & Chen, 2010). As a result, successes and failures in these domains lead to rises and drops, respectively, in people’s momentary self-esteem. Importantly, these self-esteem effects only occur for people who desire closeness to the significant other, presumably because linking one’s self-esteem to one’s performance in domains valued by a significant other is ultimately in the service of maintaining the relationship.

Expectations of acceptance or rejection. Numerous theories emphasize the role that people’s expectations of significant others’ acceptance and rejection play in their relationships (e.g., Downey & Feldman, 1996). Like any other aspect of significant-other relationships, such expectations should be stored in the linkages binding relational-self and significant-other knowledge. Hence, when a significant-other representation is activated, these expectations should play out in interactions with new others. Indeed, research on transference has shown that participants in the “Own” but not “Yoked” condition expect more acceptance from an upcoming interaction partner when the partner resembled a positively, rather than negatively, evaluated significant other (Andersen, Reznik, & Manzella, 1996).

In another study examining physically or psychologically abusive family members, Berenson and Andersen (2006) arranged for female participants with and without an abusive parent to anticipate an interaction with a partner who did (“Own” condition) or did not (“Yoked” condition) resemble this parent. The results showed that “Own” but not “Yoked” participants with an abusive parent expected more rejection from the upcoming interaction partner than did their counterparts without an abusive parent, once again demonstrating that acceptance/rejection expectations held by the relational self are activated in transference encounters.

Goals and motives. Significant others enable people to satisfy the fundamental need for belonging (Andersen, Reznik, & Chen, 1997; Baumeister & Leary, 1995). This desire to connect with others should be stored in the linkages binding relational selves and significant-other representations, and therefore activated in transference. Indeed, several transference studies have shown that “Own” but not “Yoked” participants were more motivated to be emotionally open with, and not distant from, a new person who resembled a positively rather than negatively evaluated significant other (e.g., Andersen et al., 1996; Berk & Andersen, 2000).

Research has also explored self-evaluative motives, such as self-verification motives, pursued when relational selves are activated. Self-verification refers to the desire to have others view the self in a manner consistent with one’s pre-existing self-views (Swann, 1990). Self-verification theory argues that when others see us like we see ourselves, this bolsters our sense of prediction and control by assuring us that we hold sensible beliefs about ourselves and that others’ expectations of us are appropriate—and thus that interactions with others should proceed smoothly (Swann, Stein-Seroussi, & Giesler, 1992).

Traditionally, research on self-verification has focused on people’s efforts to seek verification of their global self-views. However, some recent studies have examined self-verification strivings directed at more contextualized self-views, such as people’s views of themselves in the context of their relationships—in other words, people’s relational self-views (Swann, Bosson, & Pelham, 2002). Given research suggesting that people seek self-verifying appraisals from significant others (e.g., Swann, De La Ronde, & Hixon, 1994), Kraus and Chen (2009) hypothesized that self-verification motives are activated along with relational selves in transference encounters. Supporting this, a transference study showed that, relative to “Yoked” participants, “Own” participants wanted an upcoming interaction partner to evaluate them more in a manner that verified their core relational self-views (i.e., self-views that they rated as highly defining of the relevant relational self), regardless of the positivity or negativity of the self-views. Another study showed that the transient activation, or priming, of a significant-other representation, relative to priming a representation of an acquaintance, led participants to provide more favorable ratings of feedback that verified the relevant relational self, compared to other forms of feedback (e.g., self-enhancing).

Other researchers have similarly found evidence for goal-related elements of relational selves by activating a significant-other representation. For example, one study showed that subliminal exposure to the name of a significant other leads people to behave in line with goals associated with this other (Fitzsimons & Bargh, 2003). Moreover, activation of a significant-other representation increases the accessibility of the goals associated with the other (e.g., achievement goals), as well as goal commitment and persistence, especially when participants are close to this other and believe he or she values the goal (Shah, 2003a). Finally, significant others’ expectations about one’s goal attainment affect one’s own appraisals of the difficulty of goal attainment (Shah, 2003b).

Elicitation and disruption of affect. According to Andersen and Chen (2002), the affect-laden nature of significant-other representations means that the emotions associated with significant others should be elicited as part of the relational self in transference. In one test of this, participants’ facial expressions of emotion while reading each descriptor about an upcoming interaction partner in the learning phase of the transference paradigm were covertly videotaped (Andersen et al., 1996). Judges’ ratings of the pleasantness of participants’ facial expressions showed that “Own” but not “Yoked” participants expressed more pleasant facial affect when the representation of their positively rather than negatively evaluated significant other had been activated. Thus, transference elicits the affect associated with the significant other.

Affect reflecting the overall emotional tone of a relationship tends to be chronically experienced by the self in the context of the relationship. However, perceivers’ external and internal circumstances may disrupt this affect, which should be detectable in transference. To illustrate, results of one transference study indicated that negative affect is elicited when the representation of a positively regarded significant other is activated with a new person whose role in relation to the self is incongruent with the significant other’s role; for example, the new person is in the role of a “novice” whereas the significant other’s role is typically that of an “authority figure” (Baum & Andersen, 1999). Such role violations disrupt the positive affect typically enjoyed in positive significant-other relationships, presumably because they signal that the goals one pursues in the significant-other relationship are not likely to be met (e.g., Martin, Tesser, & McIntosh, 1993).

Self-regulation. Two forms of self-regulation have been studied in transference contexts. The first pertains to efforts to meet significant-other-related standards, and the second pertains to strategic responses aimed at defending the self and one’s relationship in the face of threat (see also Gregg, Sedikides, & Gebauer, Chapter 14, this volume).

An example of the first form of self-regulation comes from recent work drawing on self-discrepancy theory (Higgins, 1987). Self-discrepancy theory maintains that people are aware of the standpoints of significant others on their actual, ideal, and ought selves (in addition to their own standpoints). The actual self is comprised of attributes that the individual believes he/she actually possesses, whereas the ideal and ought selves are comprised of, respectively, attributes that the individual wishes to possess and attributes he/she feels it is his/her duty or obligation to possess. Significant-other standpoints on these different selves are likely to be stored as part of relational selves. As such, the activation of a relational self should activate the ideal and ought self-guides held by the relevant significant other. To the degree that actual-ideal discrepancies exist, dejection-related affect should ensue, whereas actual-ought discrepancies should elicit agitation-related affect.

To test these predictions, Reznik and Andersen (2007) asked participants with either an ideal or ought self-discrepancy from the standpoint of a parent learn descriptors about an upcoming interaction partner who did (“Own” condition) or did not (“Yoked” condition) resemble this parent. Activating the parent representation should activate the associated relational self, including the ideal or ought self-discrepancy from the parent’s standpoint. Indeed, ideal-discrepant participants in the “Own” but not “Yoked” condition reported more dejection-related affect, whereas their ought-discrepant counterparts reported more agitation-related affect.

Regulating the self with respect to ideal standards reflects a promotion regulatory focus (a focus on attaining positive outcomes), whereas self-regulation in the service of ought standards reflects a prevention focus (a focus on preventing negative outcomes) (e.g., Higgins, 1996b). If activating a parent representation activates the self-discrepancy from this parent’s standpoint, the regulatory focus with respect to this significant other should also emerge in transference. In the research described above, then, ideal-discrepant participants in the “Own” condition should show greater-approach tendencies toward their partner, whereas their ought-discrepant counterparts should show more avoidance. Supporting this, ideal-discrepant “Own” participants reported less motivation to avoid their partner in anticipation of meeting him or her, relative to after learning the meeting would not occur (at which point promotion was no longer relevant). In contrast, ought-discrepant “Own” participants reported more avoidance motivation before relative to after learning the meeting would not occur (at which point prevention was no longer relevant). Such results were not observed among “Yoked” participants.

Research on the second form of self-regulation, which involves strategic responses aimed at defending the self or one’s relationship in the face of threat, has documented both self- and relationship-protective responses. For example, regarding the research on shifts in self-definition and self-evaluation toward the relational self, the reader may recall that participants learned about an upcoming partner who did or did not resemble a positively or negatively evaluated significant other (Hinkley & Andersen, 1996). Afterward, “Own” participants described themselves with features that overlapped with ones they listed earlier to describe the relational self with this significant other, and evaluated these relational-self features in line with their evaluation of the significant other. In addition, however, “Own” participants in the negative significant-other condition evaluated the non-relational-self features of their working self-concept more favorably than participants in all other conditions—a finding that could be interpreted as a defensive response to the threat incurred by the shift toward the negative relational self.

Regarding relationship-protective self-regulation, in the study assessing facial affect (Andersen et al., 1996), regardless of whether their upcoming partner resembled a positive or negative significant other, participants were shown both positive and negative descriptors about him or her. Being confronted with negative descriptors about an upcoming partner that were derived from a positive significant other poses a threat to participants’ positive views of this other, and thus should prompt a self-regulatory response aimed at curbing this threat. Indeed, “Own” participants responded to negative descriptors about their partner with more pleasant facial affect relative to participants in any other condition and, moreover, relative to positive descriptors about this same significant other. Hence, “Own” participants buffered themselves against the negative descriptors about their positive significant other by expressing more positive facial affect, in line with the evaluative tone of the significant-other representation—a finding suggesting a relationship-protective response (for related findings, see Murray & Holmes, 1993).

Interpersonal behavior. As noted above, Andersen and Chen (2002) posit that when the relational self is activated, this includes expectations regarding the significant other’s acceptance or rejection. Such expectations should have implications for behavior in transference encounters. Indeed, research on the self-fulfilling prophecy shows that perceivers’ expectations about a target person are often fulfilled by virtue of perceivers’ tendency to act in line with these expectations and by the target’s tendency to respond in kind (e.g., Snyder, Tanke, & Berscheid, 1977). Such a self-fulfilling cycle has also been demonstrated in transference.

In this work, participants (perceivers) were exposed to descriptors about another participant (target) with whom they then had an audiotaped conversation (Berk & Andersen, 2000). The target resembled the perceiver’s own (or a yoked participant’s) positively or negatively evaluated significant other. The pleasantness of the affect expressed in participants’ conversational behavior was coded. It was hypothesized that the relational self associated with the positive or negative significant other should be activated in transference in such a way that people behave in line with their positive or negative assumptions, respectively, thus eliciting confirmatory behavior in the target. Indeed, the target expressed more pleasant affect when he or she resembled the perceiver’s own positive rather than negative significant other; no such effect was seen in the “Yoked” conditions.

Another example of the behavioral consequences of activated relational selves in transference comes from a set of studies examining affiliative behavior (Kraus, Chen, Lee, & Straus, 2010). Participants were exposed to descriptors about their upcoming partner; in the “Own” condition, the partner was described as resembling the participants’ own positively evaluated significant other, whereas in the “Yoked” condition the partner resembled someone else’s positive significant other. Across “Own” and “Yoked” conditions, the partner was described as either an in-group or out-group member (e.g., liberal or conservative). Affiliative behavior was measured as the distance participants chose to move their chair to the one that they thought was going to be occupied by their upcoming partner (e.g., Burgoon, Buller, Hale, & DeTurck, 1984). Regardless of the group status of the partner, “Own” participants pulled their chair closer to the chair of their anticipated interaction partner than did “Yoked” participants, reflecting the kind of affiliative behavior they presumably exhibit in relation to their positive significant other.

Summary

In sum, the transference perspective on the relational self (Andersen & Chen, 2002) formed the primary basis for our integrative conceptualization of the relational self (Chen et al., 2006). This social-cognitive perspective on transference and the relational self maintains that when a significant-other representation is activated in an encounter with a new person, a broad constellation of processes and phenomena come to reflect the self one is when relating to the relevant significant other—an assertion supported by a growing body of research not only on transference, but also on significant-other representations more generally. As noted from the outset, however, in addition to work on transference, our conceptualization was also grounded in several other prominent perspectives on the self and significant others, to which we now turn.

Other Theoretical Perspectives on the Self and Significant Others: Points of Convergence and Divergence

In this section, we describe four other prominent theoretical perspectives on the self and significant others. For each, we will highlight points of convergence with and divergence from the transference perspective on the relational self, and then discuss ways in which the perspective adds to or serves as a point of comparison for our broader conceptualization. Two of the perspectives, relational schemas and attachment theory, converge with the transference perspective in a number of respects, yet at the same time add unique elements to our integrative model. The other two perspectives, inclusion of others into the self and the relational-interdependent self-construal, offer a viewpoint on relational selves that diverges from our conceptualization. We will discuss this divergence, and also suggest ways that these perspectives may offer complementary rather than conflicting insights.

Relational Schemas

According to Baldwin (1992), a relational schema consists of three components: schemas of the self and the significant other in the self-other relationship, and an interpersonal script. The script consists of if–then contingencies of interaction between the self and significant other—for example, “If I seek support, then my mother will provide it.” Such if–then contingencies embody expectations about how significant others will respond to the self, built on the basis of past experiences with these individuals. Baldwin (1992, 1997) further argues that people derive rules of self-inference and self-evaluation from repeated exposure to if–then contingencies of interaction. To illustrate, the contingency “If I make a mistake, then others will criticize and reject me” may give rise to the self-inference rule “If I make a mistake, then I am unworthy” (Baldwin, 1997, p. 329).

Considerable research on relational schemas has focused on the self-evaluative outcomes that result when a relational schema is activated. For example, when the relational schema associated with a critical or disapproving significant other is activated, people exhibit more self-critical responses (e.g., Baldwin, Carrell, & Lopez, 1990; Baldwin & Holmes, 1987). Such responses reflect the self-evaluations that have presumably been derived from repeated experience with critical or disapproving feedback from the relevant significant other. Another strand of research on relational schemas has focused on the particular types of relational schemas associated with individual differences such as self-esteem. For example, research has shown that people who are low but not high in self-esteem tend to possess relational schemas in which success is associated with acceptance, whereas failure is associated with rejection (Baldwin & Sinclair, 1996).

How does the relational-schema approach relate to the transference perspective and our broader model of the relational self? The relational-schema perspective is compatible with the transference perspective on the relational self in several key respects. First, the self-schema component of relational schemas fits with the transference view that relational-self knowledge reflects knowledge about the self when relating to significant others, which is distinct from knowledge about significant others. In addition, the interpersonal script component of relational schemas fits the assumption that linkages exist in memory between relational-self and significant-other knowledge. Once again, positing such linkages is important because it distinguishes the kind of “self-within-relationships” viewpoint on the relational self put forth by transference researchers and our broader conceptualization from the “relationships-within-the-self” viewpoint put forth by other researchers to be described below. Whereas the former views the relational self in terms of the self in relation to significant others, the latter views the relational self in terms of the incorporation of aspects of significant others into the self-concept.

Theory and research on relational schemas are also compatible with transference work in that relational-self knowledge is thought to be activated by either transient or chronic sources of accessibility. That is, it has been argued that, when immediate cues in the environment activate the significant-other schema component of a relational schema, this in turn activates associated if–then rules that shift one’s views of the self toward self-conceptions in the relevant relationship (Baldwin, 1997). Research has also shown that relational schemas may be chronically accessible (e.g., Baldwin et al., 1996), implying that transient cueing is not always needed to activate these schemas and their associated self-elements.

Despite the considerable compatibility of the relational-schema and transference approaches to the relational self, studies based on the two approaches have tended to differ in emphasis and methodology. For example, although both assume that relational-self knowledge arises from repeated patterns of interaction with significant others, relational-schema research offers particular precision regarding the mechanisms underlying the formation of this self-knowledge. Specifically, as noted above, self-inferences and self-evaluations are thought to be derived from the repeated use of if–then rules, which refer to procedural knowledge structures that dictate the self-inferences and self-evaluations that follow from particular responses from significant others (Baldwin, 1997). Such if–then rules can, however, be readily incorporated into the transference view of the relational self and, in turn, our broader model. Specifically, when a significant-other representation is activated, if–then self-inference rules (derived from repeated interactions with the relevant significant other) are activated, thus leading to a shift toward relevant relational self-aspects.

As another example of differences in emphasis, given that transference refers to the re-surfacing of prior relationships in interactions with new others, research on the phenomenon has tended to rely on attribute-based cues in a new person that match the attributes of a perceiver’s significant other to activate a significant-other representation and its associated relational self. In other words, the activation cues used in research on transference emanate directly from new people themselves. Because the new person’s resemblance to the significant other is relatively minimal, the activation of transference, as we discussed above, is fairly implicit. By contrast, although subliminal exposure to significant-other faces has been used to activate relational schemas (e.g., Baldwin et al., 1990), most relational-schema research has had participants consciously visualize that they are interacting with an actual significant other (e.g., Baldwin & Sinclair, 1996). In other words, relational schemas have been activated by procedures that refer directly to significant others, rather than by cues in a resembling new person. These differences are primarily procedural, not conceptual, in nature—but it might nonetheless be worthwhile to examine the consequences of activated relational schemas in interactions with new people.

Of interest, research on relational schemas has also shown that novel cues (e.g., auditory tones) that are repeatedly paired with elements of relational schemas can activate these schemas (e.g., Baldwin & Main, 2001). If–then contingencies can also serve as activation cues in that harboring expectations about an interaction partner’s responses (Pierce & Lydon, 1998), or being exposed to an interaction pattern that resembles if–then dynamics with a significant other (Baldwin, Fehr, Keedian, Seidel, & Thompson, 1993), can activate relational schemas. Applied to transference, such studies suggest that, in daily social encounters, transference may be elicited not only by new people who resemble a significant other, but also by cues incidentally associated with a significant other or cues reflecting the dynamics of the relationship.

In sum, theory and research on relational schemas largely converges with the transference perspective on the relational self. Differences between the two bodies of work—namely the greater precision with which the relational-schema approach specifies the mechanism by which relational-self knowledge is formed, and the types of cues that have been used to activate relational schemas and their associated self-elements—suggest ways to expand the transference perspective on the relational self and, in turn, are useful additions to our broader model of the relational self.

Attachment Theory

Attachment theory is a broad theoretical framework with evolutionary, ethological, and cognitive underpinnings. The theory is used by psychologists spanning several sub-disciplines, including developmental, clinical, personality, and social psychology. It was originally developed to understand variations in the emotional bond formed between infants and their caregivers (Ainsworth, Blehar, Waters, & Wall, 1978; Bowlby, 1982, p. 2), but in recent decades has been used to study the bonds formed in any relationship, such as relationships formed between adults (e.g., romantic relationships, friendships).

A core assumption of attachment theory, and one that is particularly relevant to the present discussion, is that people develop internal working models of themselves and others in the course of early interactions with attachment figures, such as one’s mother (Bowlby, 1982). Caring and responsive attachment figures foster the formation of a model of the self as competent and worthy of love, and of others as caring and available. Attachment figures who are inconsistently responsive or are neglectful, on the other hand, give rise to insecure models—for example, a model of the self as unworthy of love and of others as uncaring. Once formed, internal working models of attachment are thought to shape people’s thoughts, feelings, and behaviors. For instance, a person with an insecure model of attachment may be more likely to interpret an ambiguously rejecting response from a romantic partner as, in fact, rejection, given his/her prior relationship experiences.

Most comparable to research on the other theoretical perspectives discussed in this chapter is research on adult attachment conducted by personality and social psychologists. This literature on adult attachment is vast and continues to grow rapidly (for a recent review, see Cassidy & Shaver, 2008). In broad strokes, the primary focus of this literature has been on the ways in which attachment working models of the self and others predict a wide array of intrapersonal and interpersonal responses. To give just a handful of examples, research has examined how people with different attachment working models regulate their emotions (e.g., Mikulincer, 1998a, 1998b), give and receive social support in their relationships (e.g., Simpson, Rholes, & Nelligan, 1992), approach interactions with strangers (e.g., Feeney, Cassidy, & Ramos-Marcuse, 2008), and make attributions about their relationship partners’ behaviors (e.g., Collins, Ford, Guichard, & Allard, 2006).

How does the attachment perspective compare and contrast with the other theoretical perspectives on the self and significant others? Several points of convergence and divergence exist between the attachment-theoretical and transference views of the self and significant others. In terms of points of convergence, early infant and adult attachment research focused on attachment figures, defined as individuals who serve a specific set of functions (e.g., significant others who provide the individual with a “secure base” from which to explore the world). However, more recent work on adults has shown the utility of applying attachment theory to a broader circle of significant others (e.g., Baldwin et al., 1996), whether or not they meet all of the criteria of attachment figures per se. Thus, working models of the self can reflect attachment figures or significant others more generally, which fits the focus of the transference theory of the relational self on the impact of significant others, attachment figures, or otherwise, on the self.

As another point of convergence, attachment theory maintains that working models of the self and of others are complementary and intertwined (e.g., Bowlby, 1973; Collins & Read, 1994), implying that they exert their effects in tandem. This fits well with the transference perspective that linkages exist between relational-self and significant-other knowledge, although most attachment research does not explicitly refer to such linkages. Exceptions are studies conceptualizing individual differences in attachment working models in terms of relational schemas, which are explicitly composed of self and significant-other schemas bound together by linkages embodying the typical if–then dynamics between self and other. In such studies, differences in attachment working models are conceptualized in terms of differences in the nature of the if–then contingencies stored in relational schemas (Baldwin et al., 1993).

Consistent with both transference and relational-schema findings, attachment working models can be activated by transient or chronic sources of accessibility (e.g., Mikulincer & Arad, 1999). In fact, methods for activating attachment working models often overlap with methods for activating significant-other representations (e.g., Saribay & Andersen, 2007a). However, attachment theory is unique in positing that physical or psychological threats in the environment activate the attachment system, and thereby activate working models (Bowlby, 1982). A key function of attachment figures is to provide a safe haven. Thus, people should seek proximity to these figures in the face of threat. Indeed, research has shown that threat-related, semantic stimuli (e.g., separation) increase the accessibility of representations of attachment figures among those who are securely attached (Mikulincer, Gillath, & Shaver, 2002). More pertinent to the activation of working models of the self, threat (e.g., failure feedback) has been shown to polarize the chronic self-evaluations of insecurely attached individuals such that those with negative self-evaluations evaluate themselves even more negatively, whereas those with positive self-evaluations evaluate themselves even more positively (Mikulincer, 1998a).

As another unique facet, most adult attachment research has treated attachment as an individual differences variable (e.g., Hazan & Shaver, 1987). As a result, working models of the self are often treated as though they reflect the self-concept as a whole, and the relational origins of these models recede into the background. For example, some research has used global self-esteem as a measure of attachment working models of the self (e.g., Griffin & Bartholomew, 1994). Although some attachment experiences may become so internalized that it may be suitable to treat them as general trait characteristics, the transference perspective on relational selves focuses on self models that derive from interactions with significant others and that designate the self in relation to specific significant others. Consistent with this focus, a growing body of research suggests that people possess both general and relationship-specific attachment models (e.g., Klohnen, Weller, Luo, & Choe, 2005; Overall, Fletcher, & Friesen, 2003; Pierce & Lydon, 2001). Overlap may exist across levels, but general and relationship-specific working models may predict different outcomes—for example, relationship outcomes may be predicted only by the corresponding relationship-specific model, and not more generalized models (Klohnen et al., 2005). In short, there is evidence to support both the attachment emphasis on generalized models and the transference emphasis on specific models.

That people can have more generalized conceptions of significant others and of relational selves, as suggested by attachment research, suggests ways in which the scope of the transference phenomenon can be widened. Namely, a new person may activate a more generalized significant-other representation (e.g., of family members), thus shifting the working self-concept toward self-aspects experienced with multiple family members. Research on transference can also inform adult attachment work. For example, recent findings suggest that transference may constitute a mechanism by which attachment working models arise in current encounters and manage to persist over time (e.g., Brumbaugh & Fraley, 2006, 2007). That is, attachment working models may persist not only because they are activated in interactions with attachment figures themselves, but also because they are activated in encounters with new people who resemble these figures.

In sum, there are several points of convergence, as well as divergence, between the attachment-theoretical and transference perspectives on the self and significant others. However, rather than suggesting a fundamental incompatibility between the two bodies of work, we maintain that the differences that exist between attachment and transference perspectives suggest ways to extend both literatures and, moreover, add to our broader model of the relational self. For example, our assumption that relational selves and their associated significant others vary in specificity was derived largely from research suggesting that attachment working models of the self and others vary in specificity.

Including Others in the Self

The inclusion-of-other-in-the-self (IOS) approach (Aron et al., 1991) is part of a larger conceptual framework known as the self-expansion model (Aron & Aron, 1986; Aron et al., 2004). The self-expansion model assumes that a fundamental human motivation is expansion of the self—that is, people are motivated to acquire resources, perspectives, and identities as a means of increasing self-efficacy and their ability to achieve goals. The IOS approach builds on this assumption by positing that one way that people can expand the self is to incorporate the resources, perspectives, and identities of close others into the self-concept. Thus, people enter into and maintain relationships in part out of a desire to expand the self via including aspects of relationship partners into the self-concept. For example, a person may treat his/her relationship partner’s possessions as his/her own, exhibit the same cognitive biases as the partner, and view the self as possessing the same attributes as the partner. According to the IOS approach, the closer a relationship is, the more the relationship partner has been included in the self. In social-cognitive terms, the IOS approach maintains that closeness in relationships leads to the merging of, or overlap in, representations or schemas of the self and of significant-other representations.

The IOS approach has been supported by various forms of evidence. In terms of the inclusion of close others’ resources, research has shown, for example, that people treat close others’ resources (e.g., money) as if these resources are their own such that allocations of resources to the self versus a close other are more similar than resource allocations to the self versus a less close other (the latter allocations favor the self over the other) (Aron et al., 1991). Importantly, this effect occurs even when participants were led to believe that the other would not know that they were responsible for the allocations. In terms of the inclusion of close others’ perspectives, studies have shown, for instance, that the attributional biases that people typically exhibit with regard to others but not the self (e.g., blaming negative actions of others on their internal states) are less apparent when others are close. In particular, the actor–observer difference, whereby people tend to make dispositional attributions for others’ behavior but situational attributions for their own behavior, is lessened when the other is a close other (e.g., Aron et al., 1991; Sande, Goethals, & Radloff, 1988).

Finally, research examining the inclusion of close others’ identities has typically shown that people confuse the attributes of close others as their own (e.g., Aron et al., 1991; Aron & Fraley, 1999). Specifically, people are faster to make “me/not me” judgments, and make fewer errors in these judgments, for attributes on which the self matches a close other (i.e., the attribute is either true or not true of both the self and other) than for mismatching attributes (i.e., the attribute is true of the self but not the other or vice versa). In other words, for mismatching attributes it takes longer to sort out and properly indicate that an attribute is true of the self and not a close other (or vice versa) because representations of the self and the close other are so merged or overlapping.

How does the IOS approach compare and contrast with other perspectives on the self and significant others? The IOS approach distinguishes self-knowledge from significant-other knowledge. However, its core assumption that closeness leads to overlap, or the merging of, self- and significant-other knowledge sets this approach apart not only from the transference perspective on the relational self, but also the relational schema and attachment perspectives. The transference and relational-schema perspectives and, in turn, our broader conceptualization, are especially explicit about treating relational-self and significant-other knowledge as linked but separate, reflecting the view that the relational self designates how the self relates to, rather than incorporates, significant others.

Research adopting a relational-schema approach provides a useful illustration of the above distinction by showing that people’s self-construals assimilate to their relationship partners on some dimensions (e.g., affiliation), but complement their partners on others (e.g., control; Tiedens & Jimenez, 2003). Thus, self-conceptions may be similar to or different from conceptions of significant others, but what matters is linkages between self- and significant-other knowledge—that is, how the self relates in interactions with significant others. The IOS approach, on the other hand, focuses on assimilated or overlapping aspects of the self and significant others to the exclusion of complementary ones, which may be equally or more relevant to the relational self.

Other differences between the transference and IOS approaches become apparent when one considers the instrument most commonly used to measure the degree to which others are included in the self. This measure consists of seven pairs of circles, with one circle in each pair designating the self and the other circle designating a significant other (Aron, Aron, & Smollan, 1992). The degree of overlap between the circles varies, with more overlap indicating greater inclusion of the other in the self. Respondents are asked to indicate the circle pair that best reflects their relationship. Although this measure is usually administered with respect to a specific significant other, there is ambiguity as to which “self” is being assessed. To illustrate, research shows that entering a new relationship yields self-concept expansion, due partly to the inclusion of aspects of the relationship partner into the self-concept (Aron, Paris, & Aron, 1995). In this research, participants were asked to describe themselves with instructions that did not explicitly refer to the relationship. Thus, it is unclear whether the “self” here refers to self-conceptions in the context of the relationship or to global conceptions of the self. In fact, IOS theorizing is relatively silent on whether contextual variations, relationship or otherwise, have implications for how much others are included, whereas variations in the relational context are central to the transference view of the relational self and, in turn, to our broader model.

In sum, IOS researchers assume that significant others influence the self by being incorporated into the general self-concept, whereas transference and relational-schema researchers maintain that significant others prompt the formation of self-aspects reflecting the self when relating to these others. Put another way, as noted in prior sections, whereas our broader conceptualization, along with the theories in which it is grounded, put forth a “self-within-relationships” viewpoint on the relational self, the IOS approach is a prominent example of a “relationships-within-self” viewpoint on the self and significant others. The IOS approach, then, may or may not make predictions about how an individual will respond to significant or new others, whereas relational selves as conceptualized from the other perspectives discussed thus far provide a clear basis for making such predictions.

Importantly, although we are drawing a distinction between perspectives like the IOS approach that take a “relationships-within-self” viewpoint on the self and significant others and perspectives like ours that take a “self-within-relationships” viewpoint, we are not saying that one of these types of perspectives is more “correct” than the other. Instead, we believe both viewpoints may characterize the nature of the self and significant others, and may even do so at the same time, within the same individual. For example, it is certainly possible for a person to interact with significant others (or new others who resemble a significant other) in ways derived from these others themselves—that is, how the self relates to others may include some aspects of significant others, suggesting these aspects have been included in the self—but at the same time these relational selves need not be derived solely from significant others. As another example, perhaps the degree to which aspects of a significant other have been included within one’s sense of self determines the extent to which aspects of the relational self become integrated with and infused into one’s general self-concept. In short, the IOS perspective on the self and significant others and broader perspectives like ours can, and likely do, co-exist.

Relational-Interdependent Self-Construal

Finally, the relational-interdependent self-construal (Cross, Bacon, & Morris, 2000) is another recent and prominent perspective on the self and significant others. Individuals who hold such a self-construal are thought to define the self primarily in terms of their close relationships. Cross et al. (2000) developed the relational-interdependent self-construal (RISC) scale to index individual differences in this self-construal primarily among respondents from North American cultures (see below for a discussion of cultural differences in self-construal). In this sense, the RISC construct is in a different category than our construct of the relational self, which is not focused on assessing individual-level variation. However, comparing and contrasting the RISC perspective and other conceptualizations of the relational self, including ours, is nonetheless useful.

Broadly speaking, the thoughts, feelings, motives, and behaviors of people who score high on the RISC scale—that is, highly relational individuals—are more colored by their close relationships than those of individuals who score lower on the RISC scale. For example, some key findings in the literature on the RISC construct include evidence that, relative to low scorers on the RISC scale, high scorers are more likely to consider the needs and opinions of significant others in their decision-making, and are judged as more open and responsive by interaction partners after a getting-acquainted interaction (Cross et al., 2000). High-RISC people also have more elaborate cognitive networks of close relationships, and have been shown to selectively attend to, and thus better recall, relational information about other people (Cross, Morris, & Gore, 2002). As a final example, high-RISC individuals are more apt to pursue goals for reasons that take into account their close relationships (e.g., Gore, Cross, & Kanagawa, 2009).

How does the RISC construct compare and contrast with the other perspectives described above on the self and significant others? Despite some basic points of convergence, such as the assumption that self-knowledge and significant-other knowledge are jointly activated from memory, several key differences exist between the RISC construct and the transference perspective. As noted earlier, a fundamental difference is that the RISC approach was specifically developed to index individual differences in defining the self in relational terms, whereas the transference approach simply assumes that all people possess aspects of the self that are linked to significant others, and that these aspects are influential for all individuals. Variations undoubtedly exist in the content and number of relational selves that people possess, but these variations have thus far not been the focus of empirical work emerging from either the transference perspective or, more broadly, our integrative model.

Second, unlike relational selves from the transference and our perspective, knowing a person’s score on the RISC scale does not provide any information on the precise content (e.g., attributes, goals, evaluations) of the person’s selves in relation to his or her significant others, nor then does it allow one to predict which of the person’s relational self-aspects are likely to be elicited when a particular significant-other representation is activated.

Finally, like the notion of including others in the self, the RISC construct connotes the incorporation of significant others into the self, or puts forth a “relationships-within-self” viewpoint, rather than the kind of “self-within-relationships” viewpoint put forth by the other perspectives discussed in this chapter. That is, for high-RISC people, “representations of important relationships and roles share the self-space with abstract traits, abilities, and preferences” (Cross et al., 2000, p. 791). On a somewhat related note, the RISC construct refers to a global, higher-order self-structure below which lower-order, more specific self-schemas exist (Cross et al., 2002). Relational selves in the transference approach and, in turn, our broader conceptualization, are more akin to lower-order self-schemas than to the central organizing structure that the RISC construct—with its explicit focus on indexing individual-level variation in self-construal—is thought to be.

In sum, like the IOS approach discussed above, the relational-interdependent self-construal offers a different perspective on the self and significant others compared to the other perspectives discussed in this chapter, including our broader conceptualization. But once again, we argue that both kinds of perspectives—the IOS and RISC perspectives, on the one hand, and the remaining perspectives, on the other hand—can and likely do co-exist. That is, one kind of perspective is not more “correct” than the other; both are potentially accurate characterizations of the self and significant others.

Moreover, it is not difficult to posit potential avenues of integration between the RISC construct and the other perspectives, such as the transference and relational-schema approaches. In fact, one recent set of studies examined the effect of activating a significant-other representation, and presumably the associated relational self, on perceivers’ “self-confidence”—that is, perceivers’ confidence in and comfort with who they are (Gabriel, Renaud, & Tippin, 2007). Gabriel et al. found that individuals scoring high on the RISC scale, or who were manipulated to hold a relational-interdependent self-construal, reported greater self-confidence after the priming of a significant other. Gabriel et al. concluded that, because high-RISC individuals (or those manipulated to hold such a self-construal) define themselves largely in terms of their relationships, bringing to mind relational selves by activating a significant-other representation should increase their self-confidence.

Another possible point of intersection is a suggestion by Cross et al. (2002) that relational schemas—and by implication, the relational self-component of these schemas—may be activated more often among high-RISC individuals. In other words, relational schemas may be more chronically accessible among high-relative to low-RISC individuals, and thus more likely to color high-RISC individuals’ thoughts, feelings, motives, and behaviors.

The Bigger Picture: Relations to Other Aspects of Identity and Future Directions

One question we have yet to address is how relational selves from the perspective of our integrative conceptualization (Chen et al., 2006) are related to other aspects of the self and identity. In this section, we consider the role of relational selves in people’s cultural and gender identities, drawing on existing evidence that speaks to this question. Finally, we discuss possible directions for future research on the relational self.

Relational Selves and Culture

In considering connections between relational selves and cultural identity, readers familiar with the cross-cultural literature are likely to wonder how relational selves are related to Markus and Kitayama’s (1991) distinction between independent and interdependent self-construals (see also Smith, Chapter 11, this volume). Markus and Kitayama argued that the traditions, institutions, and practices of North American cultures promote an independent self-construal, a view of the self as a separate, autonomous, and bounded entity. In contrast, East Asian traditions, institutions, and practices foster an interdependent self-construal, a view of the self as interconnected with others. Seemingly reminiscent of our relational-self construct, Markus and Kitayama (1991) defined the interdependent self-construal as the “self-in-relation to specific others in specific contexts” (p. 227). Moreover, they argued that this self-construal influences a wide array of psychological processes and outcomes. Does this mean that one can equate the relational self and the interdependent self-construal?

We see several reasons why the answer to this question should be no. First, although both our view of the relational self and the interdependent self-construal refer to the self in relation to others, they differ in terms of who is specified as the “others.” According to Markus and Kitayama (1991), interdependence includes an awareness of one’s part in a larger social unit, which can include both significant-other relationships and group memberships (see Brewer & Chen, 2007). Thus, minimally, the conceptualization of the interdependent self-construal is broader in scope than that of the relational self, which focuses almost exclusively on the role of significant others (for an exception, see Saribay & Andersen, 2007b, described below).

Second, the interdependent self-construal derives from self-regulatory tasks mandated in East Asian cultures, particularly that of Japan. In Japan, “one’s behavior is determined, contingent on, and, to a large extent organized by what the actor perceives to be the thoughts, feelings, and actions of others in the relationship” (Markus & Kitayama, 1991, p. 227). Indeed, Japanese individuals are expected to know their place, fit in, and engage in socially appropriate action (but see Arnett Jensen, Jensen Arnett, & MacKenzie, Chapter 13, this volume). It is this active process of attending and adjusting to others that defines the self in relation to others, and success leads to feelings of being a good cultural member (Markus & Kitayama, 1994). We are not arguing that only the Japanese adjust themselves according to whom they are with—in fact, our perspective on the relational self assumes that, across cultural contexts, different self-aspects are activated with different significant others. However, the emphasis of “adjustment” may vary across different cultures. Specifically, whereas adjustment in Japan emphasizes consideration of others’ thoughts, feelings, and needs, adjustment in North American culture, for example, may be relatively more self-focused, whereby people adjust themselves with different significant others as part of self-oriented tasks such as defining, evaluating, or presenting the self.

Lastly, although relational selves have goal elements, as described in prior sections, the relational self does not assume any single, overarching motive. In contrast, challenging the idea that the need for positive regard is a human universal, so robustly demonstrated among North Americans, Heine and colleagues have suggested instead that self-criticism and self-improvement are chief self-evaluative motives among the Japanese (Heine, Lehman, Markus, & Kitayama, 1999; cf. Sedikides, Gaertner, & Vevea, 2005; for a review, see Boucher, 2010). That is, the Japanese are hyper-vigilant to their flaws, continuously seek to improve themselves, and persevere at whatever tasks they undertake. It is through such tendencies that people with an interdependent self-construal promote unity within and commitment to their relationships and groups. That is, being aware of one’s shortcomings informs the individual where improvement efforts need to be directed so as to secure approval from others and, by implication, to maintain relationship and group harmony. Overall, then, despite surface similarities, our view of the relational self is distinct from Markus and Kitayama’s interdependent self-construal in several fundamental respects.

There are, however, potential connections between relational selves and cultural identity. For example, English and Chen (2007) recently examined differences in the consistency of relational selves across different contexts, as well as within the same context over time, among Asian-Americans versus European-Americans. They found that, consistent with theory and evidence suggesting that individuals of East Asian descent (i.e., Chinese, Korean, and Japanese descent) exhibit lower self-concept consistency (e.g., Cousins, 1989; Suh, 2002) and are especially likely to tailor the self to different relationships, East Asian Americans showed less consistency in their self ratings across different relationship contexts relative to European-Americans. In other words, East Asian Americans are especially likely to form distinct relational selves. Importantly, however, when consistency in the self was defined in terms of consistency within the same context across time, rather than consistency across different contexts, East Asian and European Americans showed similarly high levels of consistency. Hence, by examining the consistency of relational selves across and within contexts over time, English and Chen (2007) provided a more nuanced understanding of cross-cultural differences in consistency in the self.

Relational Selves and Gender

Relational selves may also be related to a person’s gender identity (see Bussey, Chapter 25, this volume). There is wide-ranging evidence, based largely on North American samples, for gender differences in the relational-interdependent self-construal, such that women tend to define themselves in terms of their close relationships more so than men (e.g., Cross & Madson, 1997; Gabriel & Gardner, 1999; Josephs, Markus, & Tafarodi, 1992). For example, women tend to score higher than men on the RISC scale (Cross et al., 2000). Earlier we described findings showing that the activation of relational selves via the priming of a significant-other representation led to higher self-reported self-confidence among high-relative to low-RISC individuals (Gabriel et al., 2007). The implication here, then, is that the activation of relational selves is likely to have a greater impact on women’s overall self-confidence than on men’s overall self-confidence.

In a somewhat related vein, researchers have examined the implications of gender differences in the degree to which the self is defined in terms of relationships for self-verification purposes. Specifically, Chen, English, and Peng (2006) hypothesized that because women define themselves in terms of their close relationships more so than men, women should be more likely to seek verification of their relational self-views. Supporting this hypothesis, Chen et al. found that, whereas both men and women favored verifying feedback over non-verifying feedback about a global self-view, only women favored verifying feedback over non-verifying feedback about a relational self-view.

Future Directions

Finally, we discuss several potential directions for future research on the relational self, each prompted by some initial, suggestive findings in the literature.

Moderators of transference. The research that we reviewed on transference and the other perspectives on the self and significant others suggests that relational selves exert a major influence in people’s daily interpersonal lives. Nonetheless, some recent work has identified moderating variables that make transference and other phenomena associated with the activation of significant-other representations more or less likely to occur, and these moderators provide insight into possible future areas of inquiry on the relational self.

Specifically, researchers have shown that transference effects are more likely to occur when participants are tested during times of “circadian mismatch”—that is, when people who prefer daytime activity are tested in the evening and when people who prefer nighttime activity are tested in the morning—relative to times of “circadian match” (Kruglanski & Pierro, 2008). This circadian difference presumably reflects reduced cognitive resources during times of circadian mismatch, and hence a greater likelihood of reliance on existing schemas—in this case, representations of significant others. In a related vein, Pierro and Kruglanski (2008) demonstrated that individuals who score high on the need for cognitive closure, or the desire for “a firm answer to a question, any firm answer as compared to confusion and/or ambiguity” (Kruglanski, 2004, p. 6), are more likely to exhibit transference effects, reflecting the pronounced tendency of such individuals to “seize and freeze” on judgments derived from highly accessible schemas—again, in this case, significant-other representations.

Other work has documented moderators of the effects of activating significant-other representations on goal-related processes. As described earlier, there is evidence that activating a significant-other representation elicits the pursuit of goals associated with the relevant significant other (e.g., Fitzsimons & Bargh, 2003; Kraus & Chen, 2009; Shah, 2003a). Does it matter whether the goals in question reflect perceivers’ own personal goals, whether they reflect the goals that significant others hold for perceivers, or both? Morrison, Wheeler, and Smeesters (2007) found that, when people share the goal that a significant other holds for them, then activating the representation of this significant other led to pursuit of this goal. In contrast, when people do not share their significant other’s goal for them, activating the representation of this significant other led to pursuit of the other’s goal only among individuals who are high in their motivation to respond to social cues (i.e., high in self-monitoring or the need to belong).

Overall, such evidence for moderators of transference and related significant-other effects is quite useful, as it adds complexity and precision to relational-self effects. More broadly, these initial moderator findings pave the way for additional research focused on identifying new dispositional and situational variables that render relational-self effects more or less likely.

Relational selves and social identity. Another possible direction for future research on the relational self has to do with connections between relational selves and social identities, where social identities refer to people’s membership in and sense of belonging to different social groups (e.g., ethnicity, religion, sexual orientation; see Spears, Chapter 9, this volume). Some initial research on transference suggests that information about social identities may be stored as part of relational selves—for example, the fact that one shares the same social identity as a significant other may be stored as part of the relational self with this important other—and are therefore activated along with the relevant significant-other representation in transference (Saribay & Andersen, 2007b). In Saribay and Andersen’s work, participants in the “Own” condition (i.e., those anticipating an interaction with a partner who resembled one of their own significant others) assumed that their upcoming partner possessed the same ethnic group identity as the relevant significant other. Moreover, “Own” participants showed bias against other ethnic groups if the relevant significant other lacked an ethnically diverse friendship circle, relative to “Yoked” participants (i.e., those anticipating an interaction with a partner who resembled another participant’s significant other) and to “Own” participants whose significant others had more diverse social networks. This inter-group bias finding—that people tend to discriminate in favor of their in-group and against out-groups (e.g., Tajfel & Turner, 1979; Spears, Chapter 9, this volume)—suggests that perceivers’ social identities are activated in transference contexts involving a significant other who shares their same group identity.

In a different but related vein, some initial research has begun to examine whether the positive expectations, attitudes, and behaviors that characterize relational selves associated with positively evaluated significant others can be harnessed in inter-group interactions to improve inter-group relations. For example, Mikulincer and Shaver (2001) found that, relative to control participants, those exposed to positive significant-other relationship primes evaluated out-group members more positively, presumably because those participants who were reminded of positive relational experiences felt safe and secure enough to be welcoming toward members of other groups. In related transference work, “Own” participants engaging in a transference encounter involving a positively evaluated significant other evaluated a target person positively—in line with the positive tone of the relevant significant-other representation—even when the target belonged to a different social group than the significant other and the participant (Kraus et al., 2010; cf. Saribay & Andersen, 2007b). Overall, this research suggests that positively toned aspects of relational selves can be used in social identity contexts as tools for reducing negative biases toward members of other social groups (see Moshman, Chapter 39, this volume; Spears, Chapter 9, this volume).

Relational selves and well-being. Finally, given the ubiquity of the relational self, an interesting question involves the possibility of using relational selves to promote well-being. We have recently conducted some work in this area under the rubric of self-affirmation theory (Steele, 1988). According to this theory, people defend themselves from the impact of information that threatens feelings of self-worth by emphasizing an unrelated but important aspect of self (see Gregg et al., Chapter 14, this volume). For example, someone may defend against the possibility of having made a bad choice, and the feelings of incompetence that ensue, by thinking of his or her promising career as a scientist (Steele, Hoppe, & Gonzales, 1986). We argue that relational selves can serve as a self-affirmational resource in the same way, by deflecting a threat that is delivered in an unrelated domain, especially for those for whom relational selves are important (Chen & Boucher, 2008). Supporting this prediction, both women and people who score high on the relational-interdependent self-construal (RISC) scale—for whom relational aspects of the self are especially self-defining, as described in prior sections—were more likely to emphasize relational self-aspects after receiving failure feedback on academic competence tests relative to men and low-RISC scorers; that is, they were more likely to spontaneously affirm relational selves in the face of threat.

Furthermore, the series of studies that we have conducted has demonstrated that relational self-affirmation repairs the blow to self-esteem that occurs after threat. Specifically, in one study, high- and low-RISC participants received threatening feedback (or not) and were induced to affirm a relational self-aspect (or not). Threatened high-RISCs had higher implicit self-esteem (i.e., non-conscious or automatic evaluation of the self) if they were induced to affirm a relational self-aspect, relative to their low-RISC counterparts; indeed, their implicit self-esteem was higher than that of high-RISCs who were not threatened at all. But, even threatened low-RISCs who were induced to affirm a relational self-aspect scored relatively highly on the implicit self-esteem measure (especially compared to threatened low-RISCs who did not affirm), although the boost they received did not match that of their high-RISC counterparts. This research is important because it suggests that, although low-RISCs may not spontaneously affirm relational selves after threatening feedback in the same way as high-RISCs, exhorting even low-RISCs to do so could serve a self-esteem repair function.

On a more basic level, one relatively straightforward way to promote well-being is to encourage the activation and use of representations of positively evaluated significant others. As described in earlier sections, when such significant-other representations are activated, not only are self-evaluations more positive (e.g., Hinkley & Andersen, 1996), but perceivers’ expectations about, and responses toward, others are similarly positively toned (e.g., Andersen et al., 1996).

Summary and Concluding Remarks

In this chapter, we described our recent, integrative conceptualization of the relational self (Chen et al., 2006), which puts forth the view that the relational self captures aspects of the self specifically in relation to—that is, in the context of interactions with—significant others. Our conceptualization was grounded heavily in social-cognitive theory and research on transference and the relational self (Andersen & Chen, 2002), which we reviewed in detail. We then described several other perspectives on the self and significant others, emphasizing ways in which they are compatible with and add unique facets to our broader model, or ways in which they offer a distinct but nonetheless useful viewpoint on the link between the self and significant others. Finally, we considered relational selves in a broader context—namely in relation to cultural and gender identities—and discussed several potential directions for future inquiry on relational selves. To conclude, judging from theory and findings to date, and across different conceptualizations, relational selves constitute an important component of an individual’s identity, and one that is likely to be a topic of great interest for self and identity researchers for decades to come.