You are on page 1of 21

Review of General Psychology Copyright 2000 by the Educational Publishing Foundation

2000, Vol. 4, No. 2, 155-175 1089-2680/00/$5.00 DOI: 10.1037111089-2680.4.2.155

The Internal Working Models Concept: What Do We Really Know

About the Self in Relation to Others?

P a u l a R, P i e t r o m o n a c o Lisa Feldman Barrett

University of Massachusetts at Amherst Boston College

The internal working models concept is the foundation for understanding how attach-
ment processes operate in adult relationships, yet many questions exist about the
precise nature and structure of working models. To clarify the working models concept,
the authors evaluate the empirical evidence relevant to the content, structure, operation,
and stability of working models in adult relationships. They also identify 4 theoretical
issues that are critical for clarifying the properties of working models. These issues
focus on the central role of affect and goals in working models, the degree to which
working models are individual difference or relational variables, and the definition of
attachment relationships and felt security in adulthood.

Each individual builds working models of the world "felt" security (cf. Bretherton, 1985; Sroufe &
and of himself in it, with the aid of which he perceives Waters, 1977). The internal working models
events, forecasts the future, and constructs his plans. In
the working models of the world that anyone builds a concept, as a mediator of attachment-related
key feature is his notion of who his attachment figures experience, is the cornerstone of attachment
are, where they may be found, and how they may be theory.
expected to respond. Similarly, in the working model Bowlby (1979) claimed that mental represen-
of the self that anyone builds a key feature is his notion
of how acceptable or unacceptable he himself is in the tations of the self and others, formed in the
eyes of his attachment figures. (Bowlby, 1973, p, 203) context of the c h i l d - c a r e g i v e r relationship,
carry forward and influence thought, feeling,
Attachment theory (Bowlby, 1969, 1973, and behavior in adult relationships. Further
1979, 1980) has profoundly influenced research work (Hazan & Shaver, 1987) elaborated on
and theorizing about the nature o f human rela- this idea, highlighting the parallels between the
tionships across the life span. The primary as-
c h i l d - c a r e g i v e r relationship and the relation-
sumption o f attachment theory is that humans ship between romantic partners. Since Hazan
form close emotional bonds in the interest of
and Shaver' s article, a growing number of stud-
survival. These bonds facilitate the develop-
ies have examined attachment patterns in adult
ment and maintenance of mental representa-
relationships. Many of these studies have at-
tions of the self and others, or "internal working
tempted to examine aspects o f internal working
models," that help individuals predict and un-
models, the hypothesized mechanism through
derstand their environment, engage in survival-
which attachment behaviors are transferred to
promoting behaviors such as proximity mainte-
different relationship partners throughout the
nance, and establish a psychological sense of
life span. Although the working models concept
is the foundation for understanding how attach-
ment processes operate throughout the life
The concept of attachment trajectory was developed in course, many unanswered questions about the
discussions with Wayne Bylsma and John Wall. Preparation nature and structure o f working models remain.
of this article was facilitated by National Science Founda- The purpose o f this article is to evaluate exactly
tion Grant 9727896. what is known about the working models con-
Correspondence concerning this article should be ad-
dressed to Paula R. Pietromonaco, Department of Psychol- cept in attachment between adults and to iden-
ogy, Tobin Hall, Box 37710, University of Massachusetts, tify critical areas that remain to be clarified.
Amherst, Massachusetts 01003-7710, or to Lisa Feldman Although our focus is on the working models
Barrett, Department of Psychology, Boston College, 427 concept as it functions in adult relationships, we
McGuinn Hall, Chestnut Hill, Massachusetts 01246. Elec-
tronic mail may be sent to or draw on the literature examining children's working models when appropriate.


The Internal Working Models Concept ceptable and unworthy (for a review, see
Cassidy, 2000). Working models of others are
A central tenet of attachment theory (Bowlby, hypothesized to include expectations about who
1969, 1973) is that people develop mental rep- will serve as attachment figures (i.e., whom to
resentations, or internal working models, that turn to when in need of security), how accessi-
consist of expectations about the self, signifi- ble those figures are, and principally about how
cant others, and the relationship between the they will respond when needed (Main et al.,
two. Working models are thought to include 1985).
specific content about attachment figures and Attachment styles. People show different
the self that is stored within a well-organized attachment styles that reflect their different in-
representational structure (Bowlby, 1980; terpersonal experiences. Work focusing on par-
Bretherton, 1985, 1990; Collins & Read, 1994). ent-child attachment (see Ainsworth, Blehar,
Furthermore, their content is believed to include Waters, & Wall, 1978) has documented differ-
knowledge about the details (e.g., what hap- ent patterns of behavioral responses from chil-
pened, where, and with whom) of interpersonal dren who are separated and then reunited with
experiences as well as the affect (e.g., happi- their mothers. These behavioral patterns are
ness, fear, and anger) associated with those ex- thought to stem from different underlying work-
periences (Bretherton, 1985). Working models ing models of the self and others. Young chil-
also are assumed to involve processes that in- dren who willingly approach their mother after
fluence what information individuals attend to, the separation and are easily comforted are as-
how they interpret events in their world, and sumed to hold working models that reflect se-
what they remember. Furthermore, these pro- curity. Those who resist contact with their
cesses are hypothesized to operate primarily mothers after a brief separation are assumed to
outside of conscious awareness (Bowlby, 1980; hold working models that reflect an avoidant
Bretherton, 1985, 1990; Main, Kaplan, & Cas- form of insecurity. Those who also display an-
sidy, 1985). Because they work on the principle ger by pushing their mother away and are dif-
of assimilation, directing both attention and be- ficult to comfort after a brief separation are
havior, working models tend to remain stable assumed to hold working models that indicate
over time, although they may change under an anxious-ambivalent form of insecurity.
some conditions (Bowlby, 1973). In the sec- In their pioneering efforts, Hazan and Shaver
tions to follow, we elaborate on these assump- (1987) investigated attachment patterns in
tions and evaluate the evidence relevant to adults that corresponded conceptually to the
working models. descriptions of children's attachment behavior
patterns. Secure adults were defined as those
Content of Working Models who appeared to be comfortable with closeness
in their relationships, and they were not partic-
Theory. Bowlby (1973) proposed that peo- ularly worried about others rejecting them; anx-
ple hold working models of the self and others, ious-ambivalent adults appeared to seek exces-
and other theorists have elaborated on this idea sive closeness and were concerned that they
(Bartholomew, 1990; Bartholomew & Horo- would be rejected; and avoidant adults appeared
witz, 1991). Working models of self are thought uncomfortable with closeness and found it dif-
to arise as individuals interact with close others ficult to depend on others. These original de-
(Cooley, 1902; Markus & Cross, 1990; Mead, scriptions of adult attachment style did not dis-
1934). In particular, they are believed to derive tinguish between specific models of the self and
from beliefs about how acceptable the self is in models of others, although working models
the eyes of early attachment figures, as gauged were assumed to be the foundation of the dif-
from the responsiveness of those figures. Chil- ferent styles. The quality of adults' working
dren who have attachment figures who are models was inferred from their self-reports of
readily available, responsive, and reliable are how they perceived their relationships in gen-
assumed to develop a representation of the self eral. This stands in contrast to the developmen-
as acceptable and worthwhile. Those who have tal research, in which the quality of child-
inconsistent or unresponsive attachment figures ren's working models was inferred from their
are assumed to develop a view of self as unac- behavior.

Most researchers who study adult attachment the Bartholomew and Horowitz model is less
now rely on a refined scheme that explicitly clear, however, when self-esteem is reported
identifies styles according to models of self and on-line (Pietromonaco & Feldman Barrett,
other (see Bartholomew & Horowitz, 1991). 1997b). When evaluations of the self were made
This scheme yields four attachment styles that immediately after specific everyday interac-
result by combining a positive or negative tions, preoccupied individuals reported lower
model of the self with a positive or negative self-esteem than did secure individuals, thus
model of others. Research suggests the follow- showing a pattern similar to the one found in
ing generalizations about people who conform self-report studies. However, the findings for
to each of the four attachment prototypes (see avoidant individuals were less consistent with
Bartholomew & Horowitz, 1991; Griffin & Bar- the patterns found in self-report studies; neither
tholomew, 1994). People who hold positive fearful-avoidant individuals nor dismissing-
models of the self and others fit the secure avoidant individuals differed from any of the
prototype and report feeling comfortable with other groups in the predicted fashion.
closeness and intimacy. People who hold nega- Empirical evidence also generally supports
tive models of both the self and others fit the the prediction that secure people hold positive
fearful-avoidant prototype and report both a views of others, but the evidence that preoccu-
fear of and a desire for closeness. People who pied people hold positive views of others and
hold a negative model of the self and a positive that avoidant people (both dismissing and fear-
model of others fit the preoccupied prototype ful) hold negative views of others is inconsis-
and are characterized by a desire for a high level tent. Some studies report the expected findings.
of closeness and by their fear of abandonment. The most often cited study (Bartholomew &
Finally, people who report a positive model of Horowitz, 1991) used sociability as an indicator
the self but a negative model of others fit the of views of others and showed that secure and
dismissing-avoidant prototype and report being preoccupied individuals held more positive
uncomfortable with closeness and overly self- views of others than did fearful and dismissing-
reliant. The four attachment styles are thought avoidant individuals. Similarly, in a study in
to be prototypes that individuals may fit to a which participants' judgments about specific
greater or lesser degree, depending on where close others were explicitly compared with their
they fall on each working model dimension. judgments of themselves, preoccupied individ-
Evidence. In general, evidence from self- uals were more likely to believe that others
report studies is consistent with Bartholomew viewed them more positively than they viewed
and Horowitz's (1991) predictions that different themselves, whereas avoidants were more likely
attachment styles are associated with positive or to believe that others viewed them less posi-
negative esteem for the self. Studies using the tively than they viewed themselves (Mikulincer,
three-category attachment model generally have 1995, Study 6).
shown that secure individuals have higher self- In contrast, some studies report findings that
esteem than do anxious-ambivalent individuals are contrary to the theoretically expected pat-
(Collins & Read, 1990; Feeney & Noller, 1990; tern. Two studies tapping general views of oth-
Hazan & Shaver, 1987). Avoidant individuals ers (Collins & Read, 1990; Hazan & Shaver,
sometimes show lower self-esteem than do se- 1987) showed that preoccupied and avoidant
cure individuals (Feeney & Noller, 1990; Hazan people hold similarly negative views of others,
& Shaver, 1987, Study 2), but other times they in contrast with the prediction that these two
do not differ from secure individuals (Collins & groups should differ in their views of others. In
Read, 1990). Studies using the four-category addition, an experience-sampling study (Pi-
model have clarified these findings. In these etromonaco & Feldman Barrett, 1997b) did not
studies, secure and dismissing-avoidant indi- support the hypothesis that secure and preoccu-
viduals evidence higher self-esteem than do ei- pied individuals generally hold more positive
ther preoccupied or fearful-avoidant individu- views of others than fearful-avoidant or dis-
als (Bartholomew & Horowitz, 1991; Brennan missing-avoidant individuals. When individu-
& Morris, 1997; Bylsma, Cozzarelli, & Sumer, als rated their views of specific interaction part-
1997; Griffin & Bartholomew, 1994; Pietromo- ners immediately after specific interactions,
naco & Feldman Barrett, 1997b). Support for people from the four attachment groups did not

differ in their perceptions of others (Pietromo- examining elaborated knowledge about differ-
naco & Feldman Barrett, 1997b). Preoccupied ent aspects of the self as enacted in close rela-
people, however, showed more positive views tionships and in particular situational contexts.
of their partners during high conflict interac- Likewise, a social-cognitive perspective im-
tions than did either secure or dismissing- plies that views of others may not be unidimen-
avoidant individuals. Thus, findings from the sional but, rather, may vary with interaction
few studies examining views of others do not partner and context. Thus, investigations of
consistently support the predictions for insecure views of others, like those of views of the self,
individuals; rather, they suggest that insecure may benefit from the use of a broader range of
individuals' views of others are neither consis- measures as well as measures that are more
tently positive nor consistently negative. valid (e.g., measures other than sociability that
Evaluation. Taken together, the findings more directly assess views of others).
generally suggest that people who display dif- Second, investigations of working models of
ferent attachment styles differ in theoretically others have included a broad range of measures
predicted ways in their views of the self. The (e.g., general beliefs about human nature vs.
differences in views of the self are robust across views of specific partners), and their association
studies involving interviews (Bartholomew & with attachment style is variable. This variation
Horowitz, 1991) and retrospective self-reports may reflect the current tension in literature re-
(e.g., Collins & Read, 1990). Evidence, how- garding whether attachment styles reflect gen-
ever, is less consistent for views of others. eral interpersonal dispositions or are specifi-
Views of others appear to vary depending on cally manifest in close relationships, and it may
which aspects are assessed (e.g., whether the account for some of the inconsistencies in the
judgment is at the general or specific level). existing literature.
Some individuals (i.e., those with a preoccupied Third, the majority of research on the content
attachment style) appear to hold mixed, incon- of working models has thus far been conducted
sistent views of others (Pietromonaco & Feld- with the use of explicit measures (i.e., self-
man Barrett, 1997b), with the positivity or neg- report). When evaluating any target (the self or
ativity of the view depending on the particular other), people are prone to making biased or
situational context. self-protective judgments (Greenwald & Banaji,
Conclusions about attachment differences in 1995), and thus research on working models
working models of the self and others are tem- would benefit from the use of more implicit
pered by three observations. First, most research measures (e.g., Greenwald and Banaji's Implicit
on views of the self has measured working Association Test [IAT]). Measures that do not
models in terms of global positive and negative rely on conscious self-report are especially im-
feelings about the self and therefore provides portant for examining working models because
only limited information about the content of many aspects are hypothesized to operate with-
self models. Social-cognitive research on the out conscious awareness and in a self-protective
self (see Kihlstrom & Klein, 1994; Markus & fashion (Bowlby, 1980). Despite the wide-
Wurf, 1987) suggests, however, that (a) the self spread use of self-report measures, some studies
is a dynamic, multifaceted structure that varies, have used multiple methods, such as interview
to some extent, with the situational context and behavior, peer reports, and self-reports (Bar-
(b) the self includes not only positive and tholomew & Horowitz, 1991), to assess aspects
negative feelings but also a broader range of of working models. Others are also beginning to
content such as central self-conceptions or compare explicitly generated reports with im-
self-schemas (Markus, 1977); possible selves plicit responses derived from the IAT (e.g.,
(Markus & Nurius, 1986); ideal, actual, and Feldman Barrett, McCabe, Costa, Bevaqua, &
ought selves (Higgins, 1987); and the self as Bliss, 1999). These studies have yielded find-
enacted in particular situational and cultural ings that are generally consistent with those
contexts (Markus & Kitayama, 1991; Markus & using solely self-report measures (for detailed
Kunda, 1986; McGuire & McGuire, 1988). discussions of additional measurement issues,
From this perspective, views of the self in the see Bartholomew & Shaver, 1998; Brennan,
context of attachment relationships need to be Clark, & Shaver, 1998; Fraley & Waller, 1998;
investigated with greater specificity, including Klohnen & John, 1998).

Finally, working models of the self and oth- (Bowlby, 1980; Bretherton, 1990; Main, 1991).
ers have been approached empirically as if they Individuals can hold different working models
have independent effects on relationship-related for different significant others because each
thought, feeling, and action, yet they are clearly model can be interconnected with other models
interdependent. In Bowlby's original theory, within a complex hierarchical network (Collins
working models of others serve the purpose of & Read, 1994).
helping individuals know whether close others Evidence. Little direct evidence exists that
will be available and responsive, and thus they is relevant to the organization of working mod-
carry implications for the self. Furthermore, els. Several studies, however, bear indirectly on
working models of the self develop initially this issue and are consistent with the ideas that
through experiences with specific others and people hold both general and relationship-spe-
how they respond (see Markus & Cross, 1990). cific working models and that different working
Thus, working models of the self are best seen models might exist at the specific level. Gener-
as models of the self in relation to others. This alized and relationship-specific measures of
idea fits with theory and research suggesting working models of the self and others are pos-
that the self is inextricably connected to others itively associated (e.g., more positive general
(see Andersen, Reznik, &Chen, 1997; Baldwin, models of the self are associated with more
1992; Markus & Cross, 1990). For example, positive relationship-specific models of the
others may be incorporated into representations self), but the correlations are small to moderate
of the self via interactions in a given situation (less than .40), indicating that they are not iden-
(Schlenker & Weigold, 1989) or particular re- tical (Cozzarelli, Hoekstra, & Bylsma, in press).
lationship (Aron, Aron, Tudor, & Nelson, 1991; Despite these different attachment patterns in
Ogilvie & Ashmore, 1991), by providing stan- different relationships, people are able to report
dards for self-evaluation (Higgins, 1987), or by a general attachment style (Baldwin, Keelan,
describing the self in reference to others Fehr, Enns, & Koh-Rangarajoo, 1996). People
(Rosenberg, 1988) through statements such as report more relationships that are consistent
"I am a supportive spouse." Researchers might with their generalized (non-relationship-specif-
better incorporate this idea of a relational self by ic) attachment style, and they more easily gen-
taking into account the self as it unfolds in erate examples of relationships that match their
particular situational contexts (e.g., see Markus generalized style (Baldwin et al., 1996). In ad-
& Kitayama, 1991), relationships (Hinkley & dition, individuals are able to list multiple peo-
Andersen, 1996), or interaction sequences (see ple who may serve as attachment figures (e.g.,
Baldwin, 1992; Baldwin, Fehr, Keedian, Seidel, romantic partners, parents, best friends, and sib-
& Thomson, 1993). lings; Trinke & Bartholomew, 1997).
Evaluation. Theoretical accounts of the
Structure of Working Models structure of working models are well elabo-
rated, but they have not been tested directly in
Theory. Several theorists have proposed the adult attachment literature. This area may
that working models are organized in a hierar- have received less attention, in part, because of
chical fashion, ranging from general to specific the reliance on self-report methods noted ear-
models (Bowlby, 1980; Bretherton, 1985, 1990; lier. Questions of structure will probably require
Collins & Read, 1994; Main et al., 1985). From the use of response latencies or other cognitive
this perspective, people do not hold a single set methods. Thus, the question of structure is ripe
of working models of the self and others; rather, for empirical investigation. Specific questions
they hold a family of models that include, at to be addressed include the following: (a) Are
higher levels, abstract rules or assumptions working models organized in a hierarchical
about attachment relationships and, at lower fashion, or might they be organized within a
levels, information about specific relationships complex network of associations that are not
and events within relatior~ships. These ideas linked within a strict hierarchy (e.g., Andersen
also imply that working models are not a single & Klatzky, 1987; Kihlstrom & Klein, 1994)?
entity but are multifaceted representations in (b) When and how is attachment behavior
which information at one level need not be guided by working models at the most abstract,
consistent with information at another level general level versus those at the more specific

level? (c) How are working models at different themselves by not relying on an attachment
levels (or within a level) interconnected? and figure who is unlikely to provide comfort
(d) Does the organizational structure of working (Cassidy & Kobak, 1988; Main & Weston,
models vary across individuals, with some peo- 1982). Dismissing-avoidant adults similarly
ple possessing a more complex structure that may seek security by downplaying their need
incorporates multiple working models and other for close relationships and by emphasizing
people possessing a less complex structure? An- their self-reliance (Fraley et al., 1998).
swering such questions will help to clarify other Evidence. A variety of indirect evidence
aspects of the concept of attachment noted ear- has been cited in support of the notion that
lier, including whether attachment styles typi- working models guide the processes underlying
cally reflect general interpersonal dispositions attachment patterns. For example, working
or whether they function as more context-spe- models are the hypothesized mechanism direct-
cific representations of particular relationships. ing people's patterns of explanations for rela-
tionship events (Collins, 1996), their percep-
Processes Underlying Working Models tions of romantic relationships (Baldwin et al.,
Theory. Working models are assumed to 1993; Carnelley, Pietromonaco, & Jaffe, 1994,
guide attention, interpretation, and memory in a 1996; Collins & Read, 1990; Hazan & Shaver,
way that allows individuals to generate expec- 1987; Pietromonaco & Camelley, 1994), their
tations about future interpersonal situations perceptions of the self and others (Mikulincer,
and to develop plans for dealing with those 1995, 1998a, 1998b), their choice or liking of
situations. On the basis of theory and evi- particular kinds of partners (Chappell & Davis,
dence in the information-processing litera- 1998; Frazier, Byer, Fischer, Wright, & De-
ture, Bowlby (1980) proposed that working Bord, 1996; Kirkpatrick & Davis, 1994; Pi-
models, through repeated use, begin to func- etromonaco & Carnelley, 1994), their organiza-
tion automatically, without conscious aware- tion of relationship knowledge (Fishtein, Pi-
ness. This proposition is consistent with re- etromonaco, & Feldman Barrett, 1999), their
cent work demonstrating that much cognitive behavior in romantic relationships (Simpson,
activity occurs automatically and outside of 1990; Simpson, Rholes, & Nelligan, 1992;
conscious awareness (see Bargh, 1994, 1997). Simpson, Rholes, & Phillips, 1996), and their
Consistent with the psychoanalytic roots of emotional experience and coping styles (Fraley
attachment theory, Bowlby also hypothesized & Shaver, 1998; Mikulincer, Florian, & Weller,
that this less conscious side of working mod- 1993; Mikulincer & Orbach, 1995; Pietromo-
els may serve defensive, self-protective func- naco & Feldman Barrett, 1997b; Tidwell, Reis,
tions (Bowlby, 1980; Cassidy & Kobak, & Shaver, 1996).
1988; Crittenden, 1990; Main, 1991). For ex- Many of these studies, however, have relied
ample, individuals may hold multiple models on conscious self-reports that cannot adequately
in which a particular aspect of reality is seen assess the cognitive mechanisms underlying
in ways that are contradictory and incoherent working models. Some notable studies (e.g.,
(e.g., "My mother deeply loves and cares Baldwin et al., 1993; Collins, 1996; Fishtein et
about me" and "My mother criticizes and al., 1999; Fraley & Shaver, 1997; Mikulincer,
rejects me and doesn't care about me"). In 1995, 1998a, 1998b), however, have used im-
this case, one model may operate within con- plicit measures (e.g., reaction time, recall, cod-
scious awareness, whereas the other model ing of open-ended inferences, and physiological
may operate primarily outside of conscious measures) that provide a stronger basis from
awareness, defending the person from a threat which to infer cognitive processes. These stud-
to the self (Bowlby, 1973; Main, 1991). Al- ies have demonstrated attachment differences in
though such defensive processes are assumed the accessibility and recall of positive and neg-
to exist to some extent in all individuals, they ative content (Baldwin et al., 1993; Mikulincer,
are thought to be particularly evident among 1995), in recall and response time for informa-
individuals with a dismissing-avoidant style tion about the self and others under differ-
(Cassidy & Kobak, 1988; Fraley, Davis, & ent contextual conditions (Mikulincer, 1998a,
Shaver, 1998). In their search for security, 1998b), in open-ended explanations for rela-
dismissing-avoidant children may protect tionship-relevant events (Collins, 1996), and in

the organizational complexity of relationship the attachment system (Fraley & Shaver, 1997).
knowledge (Fishtein et al., 1999). In addition, a Other work (Pietromonaco & Feldman Barrett,
few studies (e.g., Simpson et al., 1992, 1996) 1999b) using an implicit measure of defensive-
have used behavioral measures, which also are ness also has shown that individuals who
more implicit measures than self-report. These were higher in dismissing-avoidance evidenced
studies have shown that people with different greater defensive verbal behavior in open-
attachment styles display different behavior in ended, written narratives about a specific con-
their interactions with a romantic partner under flict in their romantic relationship, whereas
anxiety-provoking conditions (e.g., avoidant in- those higher in preoccupation evidenced less
dividuals engage in less physical contact with defensive verbal behavior. In addition, negative
their partner than do secure individuals). The self-referent words interfered less with avoidant
degree to which these behavioral differences individuals' responses on the Stroop color-nam-
reflect underlying cognitive processes will need ing task than for secure or preoccupied individ-
to be clarified, however, by studies that also uals (Mikulincer, 1995, Study 2), suggesting
assess cognitive variables as mediators of this that negative self-referent content may be less
effect. accessible in memory for avoidant individuals.
Several studies have examined whether adult Evaluation. People who evidence different
attachment is linked to defensive processing. attachment styles differ in their perceptions and
Within the developmental literature, researchers interpretations of themselves, others, and their
have examined the organization and coherence relationships; in their reported experience of
of adults' descriptions of their childhood rela- emotion; and in the ease with which they access
tionship with their parents during the Adult and recall information about themselves. Espe-
Attachment Interview (Main et al., 1985) as an cially important are studies that have relied on
implicit indicator of processing. In this re- more implicit methods and have generally sup-
search, trained judges code respondents' de- ported the idea that attachment style is associ-
scriptions along a variety of dimensions, such as ated with different patterns of construal, as well
coherence and the ability to provide specific as accessibility and memory, for some kinds of
memories. This work has demonstrated that dis- information (e.g., Baldwin et al., 1993; Collins,
missing-avoidant individuals often make gen- 1996; Fishtein et al., 1999; Fraley & Shaver,
eral idealized statements about their parents, but 1997; Mikulincer, 1995, 1998a, 1998b; Pi-
their more specific memories often focus on etromonaco & Feldman Barrett, 1999b). These
negative experiences such as being neglected or studies have the methodological advantage
rejected by a parent. This discrepancy suggests of relying on outcome measures (e.g., reac-
that dismissing-avoidant individuals may be at- tion time, behavior, physiological arousal, and
tempting to suppress their painful experiences, coded patterns of explanation or description)
but these attempts may not be completely suc- that are less subject to self-report biases. In
cessful because dismissing-avoidant adults also addition, event-contingent diary studies (Pi-
have shown increased physiological responsive- etromonaco & Feldman Barrett, 1997b; Tidwell
ness (skin conductance level) at points in the et al., 1996) in which individuals report on their
Adult Attachment Interview where they have interactions immediately after they occur also
denied the negativity of their childhood experi- are thought to minimize memory biases in self-
ences (Dozier & Kobak, 1992). reports. Studies that have used retrospective
Within the literature on attachment in roman- self-reports as outcome measures are more
tic relationships, dismissing-avoidant individu- problematic because people often are not able to
als, relative to individuals with other attachment provide accurate reports of their cognitive pro-
styles, appear to more easily suppress their cesses (see Nisbett & Wilson, 1977).
thoughts about an attachment threat (Fraley & Questions remain about why the processing
Shaver, 1997). In contrast to the findings for differences found so far might occur. For exam-
adult attachment to a parent (e.g., Dozier & ple, the extent to which processing differences
Kobak, 1992), however, this suppression was are controlled versus automatic cannot be de-
associated with decreased physiological arousal termined in most studies because participants
(skin conductance level), suggesting that these are consciously aware of the stimuli and their
individuals may have successfully deactivated responses (e.g., Bargh & Tota, 1988; Williams,

Mathews, & MacLeod, 1996). One way to ad- tions that can be updated, elaborated, or re-
dress this problem would be to present stimuli placed as life circumstances change (Bowlby,
subliminally, thereby reducing the possibility 1973; Hazan & Shaver, 1987). Such changes,
that conscious awareness could account for the when they occur, happen gradually and with
effects (see Williams et al., 1996). Another some difficulty.
question concerns the degree to which some of Although working models may show some
the processing differences, such as those found continuity in content over time, their structure is
for memory (e.g., Mikulincer, 1995), reflect dif- likely to evolve substantially from infancy to
ferences in mood (e.g., dysphoria) rather than in childhood and adulthood. The initial models
cognitive representations of the self and others. formed in infancy and early childhood are likely
For example, in one study (Mikulincer, 1995, to become more complex and sophisticated as
Study 1), preoccupied people showed less pos- children develop more abstract cognitive abili-
itivity in their recall of self-referent words, a ties (Bowlby, 1969). Young children's working
finding that could be explained in terms of the models are likely to include simple information
content and structure of their mental represen- about caregivers' availability and responsive-
tations or in terms of underlying differences ness, whereas those of older children and adults
in mood. In other words, the computations are apt to include more detailed, elaborated in-
underlying the psychological processes remain formation; to incorporate more advanced cog-
unclear. nitive processes such as imagining the partner's
No studies to date have examined the auto- responses; and to be organized within a com-
matic activation of working models (see Bargh, plex network of hierarchies (Bretherton, 1990).
1997), despite the theoretical importance of this Evidence. Working models are thought to
concept. The few studies addressing the opera- have some continuity from childhood to adult-
tion of working models outside of conscious hood because of their propensity for stability
awareness have focused on the role of defen- over time (Bowlby, 1979; Hazan & Shaver,
siveness. This work suggests that dismissing- 1987; Shaver, Collins, & Clark, 1996). Evi-
avoidant individuals are more likely to show dence based on individuals' memories of their
defensive processing, but many questions re- childhood experiences supports the notion that
main about when, why, and how these defensive some continuity does indeed exist (Camelley et
processes occur. In addition, the two studies al., 1994; Hazan & Shaver, 1987). For example,
using physiological measures (Dozier & Kobak, Hazan and Shaver (1987) found that secure adults
1992; Fraley & Shaver, 1997) yielded opposing reported warmer relationships with their parents
patterns, for reasons that will need to be clad- than did anxious-ambivalent or avoidant adults;
fled by future research. Finally, further investi- avoidant adults reported colder and more reject-
gations are needed to explore other processing ing relationships with their mothers than did
components (e.g., attentional mechanisms) that anxious-ambivalent adults; and anxious-am-
have yet to be investigated. bivalent adults reported that their parents were
more unfair and intrusive. In addition, one lon-
Stability and Continuity gitudinal study (Klohnen & Bera, 1998) has
documented that women's reports of attach-
Theory. Working models are usually con- ment-related characteristics (e.g., interpersonal
sidered to be fairly stable within a relationship closeness, social confidence, and emotional dis-
over time (Bowlby, 1973; see also Cassidy, tance) show continuity when assessed at the
2000, and Fraley & Shaver, 2000). This stability ages of 27, 43, and 52 years.
occurs, in part, because the quality of interac- Although some evidence exists for the stabil-
tions between two individuals remains stable ity of working models, it is also the case that
within the relationship, but also because work- working models can be modified as life circum-
ing models function to direct attention to repre- stances change. Studies of attachment between
sentation-consistent information and to produce mothers and children (Thompson, Lamb, & Es-
interpretations of interpersonal events that are tes, 1983; Vaughn, Egeland, Sroufe, & Waters,
consistent with those representations (Ains- 1979) suggest that major life changes alter
worth, 1989). Despite their stability, working working models, but little empirical evidence
models also are viewed as dynamic representa- exists about how life events might lead to

change or stability in working models in adult- that are important for further clarifying the un-
hood (Rothbard & Shaver, 1994). No studies derlying content, structure, and process compo-
have examined changes that may occur over nents of working models.
time in the structure of working models.
Evaluation. Similar to the research reported The Role of Affect
for other aspects of the working models con-
cept, most of the research suggesting that work- Bowlby came from the object relations tradi-
ing models are stable over time has relied on tion, and, like other object relation theorists, he
self-report, in this case respondents' memories viewed working models and their associated
of previous relationship-relevant details. Ample goals as inherently tied to affect. Much of the
evidence from other areas of psychology sug- empirical work in the adult literature has asked
gests, however, that people may not accurately the following question: How do working mod-
recall their past experiences. Current memories els influence emotional responses? The empha-
may derive primarily from working model rep- sis has been on how cognitive representations
resentations independent of relationship-rele- trigger or influence affect. Differences clearly
vant events, or they may be biased by current exist in the emotional responses of people with
relationship experiences or goals (for a discus- different attachment styles. Evidence suggests
sion of bias in retrospective judgments, see that people who report different attachment
Ross, 1989; Schacter, 1996). Thus, the degree to styles, and who presumably differ in their un-
which continuity exists from childhood through derlying working models, differ in their
adulthood remains an open question. The one emotional reactivity and in what they do in re-
longitudinal study (Klohnen & Bera, 1998) of sponse to those emotions (e.g., Bartholomew &
adults suggests that attachment-related charac- Horowitz, 1991; Carnelley et al., 1994; Hazan
teristics may show continuity from early to mid- & Shaver, 1987; Pietromonaco & Feldman Bar-
dle adulthood. rett, 1997b; Simpson et al., 1992). For example,
people who more closely fit the preoccupied
Current Status o f the Working prototype report intense emotions (Collins &
Models Concept Read, 1990; Pietromonaco & Carnelley, 1994;
Pietromonaco & Feldman Barrett, 1997b), fre-
It should be evident that the working models quent emotional ups and downs (Hazan &
concept is theoretically rich and has served as Shaver, 1987), high emotional expressiveness
the foundation for a large body of research in (Bartholomew & Horowitz, 1991), and high
both developmental and social psychology. Yet, anxiety and impulsiveness (Shaver & Brennan,
detailed descriptions of working models and 1992). In contrast, people who more closely fit
their characteristics tend to be general or impre- the dismissing-avoidant prototype report damp-
cise, and many of their core features have yet to ened emotionality (Pietromonaco & Feldman
be empirically documented. A sound theory of Barrett, 1997b), interviewers rate them as less
adult attachment requires that the working mod- emotionally expressive than other individuals
els concept be specified more clearly and em- (Bartholomew & Horowitz, 1991, Study 1), and
pirically validated. they are more able to suppress their feelings
We have identified four critical questions that (Fraley & Shaver, 1997).
are relevant for further delineating the proper- The view that working models produce emo-
ties of working models in adults: (a) What is the tion reflects the roots of many adult attachment
role of affect in working models? (Is affect an researchers in the social-cognitive tradition.
outcome of working models or an organizing Definitions of working models often appear
force?) (b) Are working models generalizations similar to definitions of schemas, but the work-
across relationships (an individual difference ing models concept reflects more motivated,
variable), or are they specific to particular rela- dynamic, affectively charged processes. From
tionships (a relational variable)? (c) What is the an object relations standpoint, we propose that
role of the relationship context in the activation relationship cognitions are inextricably tied to
of working models? and (d) What is the role of one another by their emotional content. In this
attachment goals in how working models func- view, emotions are not merely an outcome of
tion? Each of these questions touches on issues working models but are fundamental to the way

Emotional Reactivity
Low I High I
Reliance on Willin8 ..... ..... i-- r_eoc_ up_ ed.. _}
Others Unwilling Dismissive i Fearful [

Figure 1. Emotional reactivityand reliance on others for regulation.

in which people organize knowledge about their nizes these affect-related processes as integral
relationships. to the processes associated with how working
In our view, working models may be best models function.
characterized in terms of underlying affective Recent work (Niedenthal, Halberstadt, &
processes. We (Pietromonaco & Feldman Bar- Innes-Ker, 1999) offers some interesting meth-
rett, 1999a) have proposed that two similar af- ods for examining how emotion influences the
fect-related processes are implicated in the op- way in which people organize their perceptions.
eration of the attachment system for adults (see Studies by Niedenthal and her colleagues (e.g.,
Figure 1): (a) emotional reactivity, defined as Halberstadt & Niedenthal, 1997; Niedenthal,
the frequency with which the need for felt se- Halberstadt, & Setterlund, 1997) suggest that
curity is activated, and (b) emotional regulation emotion serves to promote conceptual coher-
strategies, defined as the patterns of relationship ence, leading individuals to categorize together
behavior that individuals enact in an attempt to experiences that elicit similar emotional re-
maintain or restore felt security, that is, the sponses. As a result, affect may function as the
frequency with which individuals use others in "glue" that binds information within mental
the service of affect regulation (for a comple-
representations. Nowhere should this be more
mentary model, see Fraley and Shaver, 2000).
true than for working models. Events that occur
These two affective processes capture the affec-
across different domains of life might be cate-
tive aspects of working models that have been
gorized more in terms of the emotional re-
postulated by Bowlby and other object relations
theorists and are consistent with developmental sponses they elicit than in terms of their specific
work suggesting that attachment style is broadly semantic features (e.g., what was actually said
connected to temperamental differences mani- or accomplished in an interaction). For exam-
fested as emotional reactivity and to the strate- ple, a woman who felt angry in response to an
gies (e.g., approach or avoidance behaviors) interpersonal event that occurred at work (e.g.,
that people use to modulate emotional experi- an employee's failure to arrive at meetings on
ences (see Bridges & Grolnick, 1995; Eisenberg time) and to another event that occurred at home
& Fabes, 1992; Rothbart & Derryberry, 1981). 1 (e.g., a dispute over handling family finances)
Furthermore, the idea that adults who hold neg- might more closely associate these two events,
ative self-views tend to be more emotionally even though their semantic content is quite dif-
reactive has received some indirect support ferent. Furthermore, she would associate the
(Gross, Sutton, & Ketelaar, 1998); people who two interactions to a greater degree than if the
show greater affective reactivity are more likely two interactions had elicited different emotional
to evidence neuroticism, which has been asso- reactions (e.g., anger and fear).
ciated with lower self-esteem. Although this The affective, dynamic aspects of working
perspective will need to be tested directly, it models may be best captured by more implicit
may be advantageous because it focuses re- measures (see Greenwald & Banaji, 1995;
searchers on the process by which working Greenwald, McGhee, & Schwartz, 1998) rather
models direct attachment-related behaviors. In
contrast to related theoretical perspectives (e.g.,
Mikulincer & Florian, 1998; Shaver et al., Although the degree to which emotional reactivityre-
flects innate biologicalpredispositionsor learned responses
1996) in which working models are seen as remains a point of debate, many theorists agree that it
influencing emotion regulation, our view recog- reflects both internal and external influences.

than traditional self-report measures for reasons relationship functioning (Carnelley et al., 1996;
that we have already outlined. For example, the Collins & Read, 1990; Feeney, NoUer, & Cal-
belief that people should be influenced by emo- lan, 1994; Kirkpatrick & Davis, 1994; Simpson
tional experience is inconsistent with the West- et al., 1992). Diary studies that have examined
ern cultural assumption that people should act to the same person's reactions across different
limit the influence of emotion on their thoughts kinds of relationships (e.g., best friends and
and actions (Damasio, 1994; Lutz, 1990). As romantic partners) also have revealed some gen-
a result, respondents may not accurately de- eral effects of attachment style across different
pict the role of emotion in their conscious relationships (Pietromonaco & Feldman Barrett,
self-reports. 1997b; Tidwell et al., 1996). Together, these
In summary, working models are "hot" struc- findings suggest that attachment style, when
tures that are dynamic and affectively charged. measured at a general level, operates to some
Attachment researchers have not yet capitalized extent like a broad interpersonal style or per-
on this feature of Bowlby's theory, or more sonality characteristic that affects individuals'
generally of object relations theory, but we responses to all kinds of relationships.
think that an important new direction will be to If working models are more of a relational
focus on the affect-related processes (i.e., emo- variable, then they should be relationship spe-
tional reactivity and regulation) that underlie cific and show some variability across different
working models and to examine emotion as an attachment partners. Indeed, developmental re-
organizing force in how people think about and search suggests that children do not always
behave in their relationships. show the same attachment patterns with their
mothers as with their fathers (see Fox, Kim-
Individual Difference Versus Relational merly, & Schafer, 1991). Adult attachment re-
Variable searchers typically have not measured attach-
ment style as a relational variable, but there are
A primary question in the attachment litera- some recent exceptions (Baldwin et al., 1996;
ture concerns whether attachment styles, and Cozzarelli et al., in press; Trinke & Bar-
the working models that underlie them, are an tholomew, 1997). These studies indicate that
individual difference variable or a relational people (a) can identify more than one person
variable (e.g., Kobak, 1994). As an individual who may serve as an attachment figure and are
difference variable, working models would be able to rank order in a hierarchical fashion their
associated with a consistent pattern of attach- use of these people as attachment figures
ment-related behaviors across relationships (Trinke & Bartholomew, 1997); (b) report a
with different partners. Consistent with this range of attachment styles across relationships,
view, adult attachment researchers typically evidencing feelings of security in some relation-
measure attachment style at a general level, ships but anxious-ambivalence or avoidance in
asking people to describe their general style others (Baldwin et al., 1996); and (c) are able to
across relationships (e.g., by choosing a proto- report both generalized and specific working
type or making dimensional ratings to describe models of the self and others (Baldwin et al.,
their general feelings about relationships). 1996; Cozzarelli et al., in press).
Using this approach, researchers have dem- Thus, evidence exists in support of working
onstrated that people who report different gen- models as a general personality variable and as
eralized attachment styles differ in their beliefs a relationship-specific variable. Generalized and
about themselves, others, and relationships specific models are probably related, but it is
(Baldwin et al., 1993; Bartholomew & Horo- likely that, in many cases, they are not identical
witz, 1991; Camelley & Janoff-Bulman, 1992; (see Cozzarelli et al., in press). It will be im-
Collins & Read, 1990; Hazan & Shaver, 1987; portant for future work to assess attachment at
Mikulincer, 1995; Pietromonaco & Feldman both general and specific levels to determine the
Barrett, 1997b); their emotional responses (Col- relative ability of each type of measure to pre-
lins & Read, 1990; Hazan & Shaver, 1987; dict different behavioral outcomes. When at-
Mikulincer & Orbach, 1995; Pietromonaco & tachment is measured at the most general level,
Carnelley, 1994; Pietromonaco & Feldman Bar- thoughts, feelings, and behaviors that are de-
rett, 1997b; Shaver & Brennan, 1992); and their rived from generalized expectations of others

are apt to operate much like a broad personality Second, how do people aggregate across all
variable. This relatively nonspecific working of their relationship experiences to be able to
model may be most evident in contexts in which respond to general questions about attachment
little is known about the relationship context style? Do they use the style that characterizes
(e.g., at the beginning of a new relationship) or their most frequent type of attachment relation-
when individuals are unable or unwilling to ship, their most intense attachment relationship,
attend to relationship-distinguishing details. As or their most recent relationship? Or do they
a result, few differences across relationships summarize across relationships to estimate how
will be evident. In contrast, when attachment is they feel on average? It is unclear to what extent
measured at a more specific relationship level, individuals respond in any of these ways, or
greater variation may occur in behaviors across whether different individuals rely on different
relationships. Indeed, initial evidence (Cozza- strategies to estimate their general attachment
relli et al., in press) suggests that relationship- style. Furthermore, not all relationship contexts
specific measures better predict relationship will necessarily evoke attachment processes to
outcomes than more general measures. the same extent, even for relationships that
A number of interesting theoretical issues clearly serve attachment functions (see Simpson
arise from the notion that working models vary & Rholes, 1994). How are such base rates fac-
in their degree of specificity. First, what does tored into aggregate estimates?
the concept of an attachment "style" mean when Third, is there a single working model of the
working models vary from relationship to rela- self or multiple working models of the self?
tionship? Perhaps "style" should reflect the Because working models of the self include
working models that are modally activated (i.e., how acceptable and worthwhile one is in the
eyes of an attachment figure, it is possible that
most frequent model used with others), or per-
there are multiple models of the self, each as-
haps it should reflect the working models asso-
sociated with a different attachment relation-
ciated with the most formative or most affec-
ship. This view fits closely with social-cogni-
tively important relationship (see Weiss, 1982).
tive conceptions of the self (Markus & Cross,
Some theorists (Ainsworth, 1989; Hazan &
1990; Markus & Wurf, 1987) that depict self-
Shaver, 1994) have suggested that the effects of
representations as multifaceted and varying
working models should be strongest in a pri-
with the situational or relational context. If re-
mary attachment relationship and that people
lational context is important, as this view sug-
are likely to have a single primary attachment
gests, then it may be more useful to view work-
relationship. In this case, the definition of at- ing models of attachment as representations of
tachment style would need to be more con- the self in relation to others (Andersen et al.,
strained than it has been in the literature, with 1997; Hinkley & Andersen, 1996), wherein the
greater emphasis on examining those relation- other might be a generalized other or a specifc
ships that are most likely to serve attachment relationship parmer. Furthermore, for a given
needs. Given that individuals possess multiple relationship, people may have multiple working
working models at the more relationship-spe- models for the self in relation to that particular
cific level, perhaps the concept of attachment partner (Ogilvie & Ashmore, 1991). Thus,
"style" should come to include the flexibility working models of the self in relation to any
with which individuals shift from one set of particular attachment figure may vary with the
models to another as they change relationship situational context.
contexts. Perhaps it might be useful to abandon Finally, are the processes associated with
the attachment "style" concept all together in working models that underlie attachment pat-
favor of attachment "trajectory." An attachment terns stable individual differences, or are they
trajectory can occur over the life span, as work- relationship specific? Take, for example, our
ing models change with important relationship suggestion that working models are associated
experiences. An attachment trajectory also can with differences in emotional reactivity and re-
occur within a relationship over time, as indi- liance on others for emotion regulation pur-
viduals move from using their most general poses. Emotional reactivity, defined as the fre-
working model of others to fashioning a model quency with which an individual is threatened
of a specific other. and security needs arise, may be more of an

individual difference variable. People who are selves) have been less clear about establishing
more preoccupied or fearful-avoidant are likely the boundaries of what is and what is not an
to be more reactive and to feel threatened more attachment relationship in adulthood, yet this
frequently, whereas those who are secure will distinction is important for assessing the impli-
be less reactive and less frequently threatened. cations of working models for relationship
As a result, preoccupied and fearful-avoidant processes.
individuals will experience more intense nega- Research following Hazan and Shaver's
tive affect more frequently and therefore have a (1987) article has examined "adult attachment"
greater need to regulate their affect than will in a variety of relationships: those with romantic
secure individuals. Dismissing-avoidant indi- partners (e.g., Carnelley et al., 1994), parents
viduals, because they are apt to use more de- (e.g., Collins & Read, 1990), same-sex and op-
fensive strategies (Cassidy & Kobak, 1988; Fra- posite-sex friends (e.g., Tidwell et al., 1996),
ley et al., 1998), are not likely to consciously strangers (e.g., Mikulincer & Nachshon, 1991),
experience threat and therefore will be less coworkers (Hazan & Shaver, 1990), and even
likely to seek security from others. How indi- God (Kirkpatrick & Shaver, 1992). Thus, re-
viduals go about regulating their affect--or, searchers have focused on a variety of relation-
more specifically, the extent to which they rely ships that may or may not serve attachment-
on close others for regulatory support--may related needs. As we have noted, some research
vary from relationship to relationship. Thus, the on attachment has been conducted without ref-
attachment patterns observed may stem from both erence to a specific relationship partner. This
stable and context-dependent characteristics. lack of specificity may have occurred because
In summary, assessing attachment at the gen- we have not yet examined what constitutes an
eral level will provide only a limited view of the attachment relationship in adulthood (see Fraley
working models that underlie attachment be- and Shaver, 2000).
haviors. Attachment measured at a more general Felt security. Ainsworth (1989) suggested
level is likely to reflect a more dispositional that the identifying feature of an attachment
characteristic that will predict responses across relationship is that it serves the function of
different kinds of relationships, and it may be providing felt security. To know what an attach-
closely connected to characteristics such as tem- ment relationship is in adulthood, one needs to
perament or frequency of threat activation. Re- know what felt security is for an adult. On the
lationship-specific variation may also exist, but basis of both the psychodynamic literature from
to assess it properly, attachment phenomena which attachment theory originally derives and
must be measured at the level of the specific literature on the self, we propose that adults
relationship in a way that takes into account the experience felt security when their attachment
characteristics of that particular relationship. figure confirms that (a) they are loved and lov-
We also need to consider that, when attachment able people, and (b) they are competent or have
is measured at one point in time, the findings mastery over their environment. A feeling of
may not generalize across time or context for a threat occurs when a press arises for which the
given individual. individual feels ill equipped to cope (Tomaka &
Blascovich, 1994). The press can be generated
The Role of the Relationship Context either externally, from the environment, or in-
ternally, from negative affect. When individuals
Internal working models may be most likely experience a self-relevant threat, they may need
to be activated within the context of attachment to engage in behaviors that will help to reestab-
relationships, but what is an attachment rela- lish or promote feelings of security. Thus, indi-
tionship in adulthood? Within the developmen- viduals experience threat (and therefore a need
tal literature (e.g., Ainsworth, 1989; Sroufe & for security) when their self-esteem is in ques-
Waters, 1977), an attachment relationship is tion, either because of negative information or
defined as a close emotional bond between a when they feel unable to deal with a perceived
child and his or her primary caregiver that danger on their own. From this perspective,
serves the important function of providing the then, attachment relationships are those that
child with physical and psychological security. have the potential to provide felt security in the
Adult attachment researchers (including our- face of threat and in which working models of

the self are modified or reinforced in some views of themselves (e.g., preoccupied people)
significant way by the actions of another. may see many situations as potentially threat-
Although attachment relationships may pro- ening to their sense of self. As a result, they may
vide a source of felt security throughout the life attempt to treat many other people--even inap-
course, felt security may not be established and propriate ones, such as strangers--as if they
maintained via precisely the same behaviors in were attachment figures in an effort to achieve
adulthood as it is in childhood. Attachment re- felt security. Furthermore, they may engage fre-
lationships between two adults differ in several quently in attachment-related behaviors, and of-
respects from those between a parent and child ten these behaviors will appear to be inconsis-
(Ainsworth, 1989; Hazan & Shaver, 1994; tent with the current context. Dismissing-
Shaver, Hazan, & Bradshaw, 1988). Adults in avoidant individuals, who prefer not to depend
romantic relationships are motivated not only on others, are less likely to use others as a way
by the attachment system but also by the care- of regulating felt security, even when they
giving and reproductive systems. The caregiv- ought to do so. This unwillingness to rely on
ing system produces behaviors designed to calm others, combined with the notion that they may
or help another and typically is activated when be threatened infrequently, may account for
individuals see that their romantic partner is why dismissing-avoidant individuals rarely en-
distressed. The reproductive system is associ- gage in attachment-related behaviors. Recent
ated with sexual behaviors, which may be acti- evidence (Fraley & Davis, 1997) suggests that
vated by cues such as physical or social attrac- dismissing-avoidant adults are less likely to
tiveness. In adults, it is likely that the three establish an attachment relationship with a ro-
systems (attachment, caregiving, and sexuality) mantic partner and may even try to handle their
work together to produce felt security because attachment needs on their own. Thus, their neg-
individuals may feel worthwhile and effective ative views of others may lead them to try to
in the context of sexual intimacy or when pro- serve as their own attachment figure in an effort
viding care to their partner. Thus, the aspects to achieve felt security.
that define an attachment system in children For relationships that serve attachment func-
may not have a complete parallel in adulthood. tions, other aspects of the situational context
As with all life span developmental questions, will determine whether individuals actually dis-
the challenge for researchers is to determine play attachment-related behaviors. For exam-
which behaviors are functionally equivalent at ple, in the face of distressing events, such as
different points in the life cycle. separation from an attachment figure or a phys-
Anchoring attachment to the concept of felt ical threat, children are more likely to engage in
security in adults may have implications for attachment behaviors such as proximity seeking
how the concept of attachment style and the (Bowlby, 1980). Similarly, separation (Fraley
underlying working models are understood. It & Shaver, 1998), distress (Mikulincer, 1998a,
may be that global attachment style is, in part, 1998b; Simpson et al., 1992, 1996), or interper-
determined by the frequency with which indi- sonal conflict (Pietromonaco & Feldman Bar-
viduals treat partners as attachment figures as a rett, 1997b) in adulthood may trigger the oper-
way of regulating their emotions. Although ation of attachment processes.
some contexts might require that individuals Summary. Defining the degree to which a
temporarily rely on an interaction partner for particular relationship serves attachment func-
felt security when normally they would not, tions will be important for understanding when
attachment can be viewed as variation in the and how working models guide relationship
tendency of individuals to rely on others for felt processes. Working models should be most
security. Securely attached adults, who gener- likely to be activated when individuals perceive
ally feel competent and worthy, may seek out an a threat to their sense of self and turn to another
attachment figure only when they experience a to reestablish feelings of security. Some indi-
specific, external threat to the self. Because they viduals, such as those who are more likely to
infrequently feel threatened, they will engage experience threat frequently, may use a larger
in attachment-related behaviors (e.g., support number of relationship partners to meet their
seeking) only when necessary. In contrast, peo- need for felt security, whereas other individuals
ple who have less certain and less positive may treat only a few people as attachment fig-

High Low
Intimacy I High ..... S~ure ......... P~_~up_i~__.

Low Dismissive Fearful

Figure 2. Chronicsubgoals associatedwith attachmentstyles.

ures. Thus, rather than assume that particular Figure 2). In seeking a sense of felt security,
kinds of relationships (e.g., romantic relation- secure people achieve a balance between estab-
ships) are attachment relationships, researchers lishing intimacy and maintaining independence
will need to assess the degree to which different and probably do so with a good deal of flexi-
individuals use particular relationship partners bility in terms of when they apply each goal.
in the service of attachment needs. In addition, People who are preoccupied with attachment
it will be important to specify further other appear to hold an overriding chronic goal to
aspects of the situation that are most likely to achieve intimacy as a way of attaining felt se-
activate working models within a given rela- curity, and part of that goal involves obtaining
tionship. Although broad classes of threatening responsiveness from others. In contrast, people
situations can be identified, it is likely that the who are dismissing of attachment seem to hold
variety of potential threats to the self, and the an overriding chronic goal to maintain their
degree to which individuals use particular rela- independence from others as a way of achieving
tionship partners to handle those threats, will felt security, which may be further linked to a
vary idiographically. goal to protect the self (Fraley, Davis, &
Shaver, 1998). And, finally, people who are
Goals fearful of attachment seem to hold conflicting
chronic goals to both achieve intimacy and
Just as attachment-related working models maintain independence from others in their at-
may be organized in a hierarchical fashion, so tempts to achieve felt security, but with less
too may attachment-related goals be organized flexibility than secure individuals; fearful indi-
hierarchically. As the definition of an attach- viduals may have both goals activated at the
ment relationship implies, working models are same time, leading to approach-avoidance
organized around a single, overarching goal to conflicts.
achieve felt security. How people go about try- These different subgoals also may lead to
ing to achieve this goal may depend, in part, on quite different interpretations and reactions to
their subgoals. Some of the subgoals that have interpersonal events. Two studies (Fishtein et
been discussed in the literature include midlevel al., 1999; Pietromonaco & Feldman Barrett,
goals used to direct behavior in the service of 1997b) using different methods and measures
the broader goal of obtaining felt security. have shown that preoccupied people, who de-
These subgoals include, for example, seeking sire a high level of intimacy and responsiveness
intimacy or closeness, a desire to maintain one's from others, view high conflict interactions or
independence, and protection of the self. Few high conflict relationships much more favorably
studies have directly examined attachment than people with other attachment styles who do
goals, although some theorists (Collins & Read, not hold this subgoal to the same extent. In a
1994; Shaver et al., 1996) have discussed their diary study (Pietromonaco & Feldman Barrett,
significance in working models. 1997b), we found that preoccupied people, in
From our perspective (Pietromonaco & comparison with people holding other attach-
Feldman Barrett, 1997a), chronic subgoals of ment styles, reported greater intimacy, more
achieving intimacy and maintaining indepen- positive emotions, and more positive views of
dence from others work in the service of main- their partners as their interactions increased in
taining felt security and fit within the four- conflict. In a laboratory study (Fishtein et al.,
category prototype model of attachment (see 1999), we found that preoccupied people who

were involved in higher conflict romantic rela- ing models concept have generated a wealth of
tionships held more complex views of positive information about issues central for understand-
aspects of their relationship than did people in ing attachment in close relationships, but many
other attachment groups, even though all people fundamental questions about the content, struc-
in higher conflict relationships reported more ture, and function of working models remain.
complex negative views. This finding suggests First, studies have yet to determine whether
that preoccupied people attend not only to the working models and their content are best char-
negative aspects of conflictual relationships but acterized in terms of distinct models of the self
also to the positive, intimacy-promoting as- and others or in more relational terms that re-
pects. High conflict interactions, even though flect the self in relation to others. Second, few
they may be unpleasant in some respects, may studies have attempted to examine the structure
offer preoccupied people the chance to elicit of working models, although theorizing about
responses from their partner (e,g., personal dis- structure abounds. Third, whether working
closures and expressions of emotion) that may models actually guide processes such as atten-
make them believe that they are achieving inti- tion, interpretation, and memory remains an
macy. Along similar lines, recent work (Miku- open question. The strongest evidence so far
lincer, 1998c) suggests that a sense of trust is comes from studies that have minimized self-
closely tied to a goal to achieve security for report biases by using more implicit measures.
preoccupied individuals and to a goal to achieve Fourth, more longitudinal evidence is required
control for avoidant individuals. Future work to determine the extent to which working mod-
applying social-cognitive methods (see Bargh, els show stability and change in adulthood.
1997) for creating temporary interpersonal The task for future research will be to specify
goals (e.g., to achieve intimacy or maintain in- the relational and situational conditions under
dependence) in the laboratory will be important which working models are most influential and
for determining the causal role of goals in rela- to develop a more precise knowledge base about
tionship perceptions and behavior. the structure and operation of working models.
In summary, working models are likely to If understanding of adult attachment processes
incorporate a variety of attachment subgoals is to move forward, future researchers will need
that are used in the service of felt security. to take steps to (a) take into account the more
Individuals may meet the overarching goal of dynamic, "hot" aspects of working models by
achieving felt security in different ways depend- further examining the role of goals and affect in
ing on their subgoals, such as achieving inti- attachment processes; (b) assess working mod-
macy. These goals, embodied in working mod- els at multiple levels, ranging from general to
els, may lead individuals to construe and re- relationship-specific models, as well as identify
spond to similar situations in very different the features of the situational context that may
ways. Attachment goals are likely to be an activate working models; (c) specify whether a
important aspect of working models, but the given relationship meets the criteria for an at-
challenge for future investigations will be to tachment relationship (i.e., providing felt secu-
better specify what they are and how they are rity and thereby influencing models of the self);
organized and to look more carefully at how and (d) determine the more proximal goals via
they might shape relationship perceptions and which felt security is maintained.

Conclusion References

Over the past 12 years, attachment theory has Ainsworth, M. D. (1989). Attachments beyond in-
become a driving force in the study of close fancy. American Psychologist, 44, 709-716.
relationships because it provides a comprehen- Ainsworth, M., Blehar, M., Waters, E., & Wall, S.
(1978). Patterns of attachment. Hillsdale, NJ: Erl-
sive perspective for understanding relationship baum.
processes. The working models concept is a Andersen, S. M., & Klatzky, R. L. (1987). Traits and
foundational component of this theory as ap- social stereotypes: Levels of categorization in per-
plied to both children and adults. Theory and son perception. Journal of Personality and Social
research implicitly based on the internal work- Psychology, 53, 235-246.

Andersen, S. M., Reznik, I., &Chen, S. (1997). The Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998).
self in relation to others: Motivational and cogni- Self-report measurement of adult attachment: An
tive underpinnings. Annals of the New York Acad- integrative overview. In J. A. Simpson & W. S.
emy of Sciences, 818, 233-275. Rholes (Eds.), Attachment theory and close rela-
Aron, A., Aron, E. N., Tudor, M., & Nelson, G. tionships (pp. 46-76). New York: Guilford Press.
(1991). Close relationships as including other in Brennan, K., & Morris, K. A. (1997). Attachment
the self. Journal of Personality and Social Psy- styles, self-esteem, and patterns of seeking feed-
chology, 60, 241-253. back from romantic partners. Personality and So-
Baldwin, M. W. (1992). Relational schemas and the cial Psychology Bulletin, 23, 23-31.
processing of social information. Psychological Bretherton, I. (1985). Attachment theory: Retrospect
Bulletin, 112, 461-484. and prospect. Monographs of the Society for Re-
Baldwin, M., Fehr, B., Keedian, E., Seidel, M., & search in Child Development, 50(1-2, Serial No.
Thomson, D. W. (1993). An exploration of the 209).
relational schemata underlying attachment styles: Bretherton, I. (1990). Open communication and in-
Self-report and lexical decision approaches. Per- ternal working models: Their role in the develop-
sonality and Social Psychology Bulletin, 19, 746- ment of attachment relationships. In R. A. Thomp-
754. son (Ed.), Nebraska Symposium on Motivation
Baldwin, M. W., Keelan, J. P. R., Fehr, B., Enns, V., (Vol. 36, pp. 57-113). Lincoln: University of Ne-
& Koh-Rangarajoo, E. (1996). Social-cognitive braska Press.
conceptualization of attachment working models: Bridges, L. J., & Grolnick, W. S. (1995). The devel-
Availability and accessibility effects. Journal of opment of emotional self-regulation in infancy and
Personality and Social Psychology, 71, 94-109. early childhood. In N. Eisenberg (Ed.), Social de-
Bargh, J. A. (1994). The four horsemen of automa- velopment: Review of personality and social psy-
ticity: Awareness, intention, efficiency, and con- chology (Vol. 15, pp. 185-211). Thousand Oaks,
trol in social cognition. In R. S. Wyer, Jr., & T. K. CA: Sage.
Snail (Eds.), Handbook of social cognition (2nd Bylsma, W. H., Cozzarelli, C., & Sumer, N. (1997).
ed., pp. 1-40). Hillsdale, NJ: Erlbaum. Relation between adult attachment styles and
global self-esteem. Basic and Applied Social Psy-
Bargh, J. A. (1997). The automaticity of everyday
chology, 19, 1-16.
life. In R. S. Wyer, Jr. (Ed.), Advances in social
Carnelley, K. B., & Janoff-Bulman, R. (1992). Opti-
cognition (Vol. 10, pp. 1-61). Mahwah, NJ: Erl-
mism about love relationships: General vs. specific
lessons from one's personal experiences. Journal
Bargh, J. A., & Tota, M. E. (1988). Context-depen-
of Social and Personal Relationships, 9, 5-20.
dent automatic processing in depression: Accessi- Camelley, K. B., Pietromonaco, P. R., & Jaffe, K.
bility of negative constructs with regard to self but (1994). Depression, working models of others, and
not others. Journal of Personality and Social Psy- relationship functioning. Journal of Personality
chology, 54, 925-939. and Social Psychology, 66, 127-140.
Bartholomew, K. (1990). Avoidance of intimacy: An Camelley, K. B., Pietromonaco, P. R., & Jaffe, K.
attachment perspective. Journal of Social and Per- (1996). Attachment, caregiving, and relationship
sonal Relationships, 7, 147-178. functioning in couples: Effects of self and partner.
Bartholomew, K., & Horowitz, L. M. (1991). Attach- Personal Relationships, 3, 257-278.
ment styles among young adults: A test of a four- Cassidy, J. (2000). Adult romantic attachments: A
category model. Journal of Personality and Social developmental perspective on individual differ-
Psychology, 61, 226-244. ences. Review of General Psychology, 4, 111-131.
Bartholomew, K., & Shaver, P. R. (1998). Methods Cassidy, J., & Kobak, R. (1988). Avoidance and its
of assessing adult attachment: Do they converge? relation to other defensive processes. In J. Belsky
In J. A. Simpson & W. S. Rholes (Eds.), Attach- & T. Nezworski (Eds.), Clinical implications of
ment theory and close relationships (pp. 25-45). attachment theory (pp. 300-323). Hillsdale, NJ:
New York: Guilford Press. Erlbaum.
Bowlby, J. (1969). Attachment and loss: VoL 1. At- Chappell, K. D., & Davis, K. E. (1998). Attachment,
tachment. New York: Basic Books. partner choice, and perceptions of romantic part-
Bowlby, J. (1973). Attachment and loss: Vol. 2. Sep- ners: An experimental test of the attachment-secu-
aration: Anxiety and anger. New York: Basic rity hypothesis. Personal Relationships, 5, 327-
Books. 342.
Bowlby, J. (1979). The making and breaking of af- Collins, N. L. (1996). Working models of attach-
fectional bonds. London: Tavistock. ment: Implications for explanation, emotion, and
Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss: behavior. Journal of Personality and Social Psy-
Sadness and depression. New York: Basic Books. chology, 71, 810-832.

Collins, N., & Read, S. J. (1990). Adult attachment, friendships and romantic relationships. Personal
working models, and relationship quality in dating Relationships, 4, 131-144.
couples. Journal of Personality and Social Psy- Fraley, R. C., Davis, K. E., & Shaver, P. R. (1998).
chology, 58, 644-663. Dismissing-avoidance and the defensive organiza-
Collins, N., & Read, S. J. (1994). Cognitive repre- tion of emotion, cognition, and behavior. In J. A.
sentations of attachment: The structure and func- Simpson & W. S. Rholes (Eds.), Attachment the-
tion of working models. In D. Perlman & K. Bar- ory and close relationships (pp. 249-279). New
tholomew (Eds.), Advances in personal relation- York: Guilford Press.
ships (Vol. 5, pp. 53-90). London: Jessica Fraley, R. C., & Shaver, P. R. (1997). Adult attach-
Kingsley. ment and the suppression of unwanted thoughts.
Cooley, C. H. (1902). Human nature and the social Journal of Personality and Social Psychology, 73,
order. New York: Schocken Books. 1080-1091.
Cozzarelli, C., Hoekstra, S., & Bylsma, W. (in press). Fraley, R. C., & Shaver, P. R. (1998). Airport sepa-
General versus specific mental models of rations: A naturalistic study of adult attachment
attachment: Are they associated with different dynamics in separating couples. Journal of Per-
outcomes? Personality and Social Psychology sonality and Social Psychology, 75, 1198-1212.
Bulletin. Fraley, R. C., & Shaver, P. R. (2000). Adult romantic
Crittenden, P. M. (1990). Internal representational attachment: Theoretical developments, emerging
models of attachment relationships. Infant Mental controversies, and unanswered questions. Review
Health Journal, 11, 259-277. of General Psychology, 4, 132-154.
Damasio, A. R. (1994). Descartes' error: Emotion, Fraley, R. C., & Waller, N. G. (1998). Adult attach-
reason, and the human brain. New York: Grossett/ ment patterns: A test of the typological model. In
Putnam. J. A. Simpson & W. S. Rholes (Eds.), Attachment
Dozier, M., & Kobak, R. R. (1992). Psychophysiol- theory and close relationships (pp. 77-114). New
ogy in attachment interviews: Converging evi- York: Guilford Press.
dence for deactivating strategies. Child Develop- Frazier, P. A., Byer, A. L., Fischer, A. R., Wright,
ment, 63, 1473-1480. D. M., & DeBord, K. A. (1996). Adult attachment
Eisenberg, N., & Fabes, R. A. (1992). Emotion, reg- style and partner choice: Correlational and exper-
ulation, and the development of social compe- imental findings. Personal Relationships, 3, 117-
tence. In M. S. Clark (Ed.), Emotion and social 136.
behavior: Review of personality and social psy- Greenwald, A. G., & Banaji, M. R. (1995). Implicit
chology (Vol. 14, pp. 119-150). Newbury Park, social cognition: Attitudes, self-esteem, and ste-
CA: Sage. reotypes. Psychological Review, 102, 1-27.
Feeney, J. A., & Noller, P. (1990). Attachment style Greenwald, A. G., McGhee, D. E., & Schwartz, J. L.
as a predictor of adult romantic relationships. Jour- (1998). Measuring individual differences in im-
nal of Personality and Social Psychology, 58, 281- plicit cognition: The Implicit Association Test.
291. Journal of Personality and Social Psychology, 74,
Feeney, J. A., Noller, P., & Callan, V. J. (1994). 1464-1480.
Attachment style, communication and satisfaction Griffin, D., & Bartholomew, K. (1994). Models of
in the early years of marriage. In D. Perlman & K. self and other: Fundamental dimensions underly-
Bartholomew (Eds.), Advances in personal rela- ing models of adult attachment. Journal of Person-
tionships (Vol. 5, pp. 269-308). London: Jessica ality and Social Psychology, 67, 430-445.
Kingsley. Gross, J. J., Sutton, S. K., & Ketelaar, T. (1998).
Feldman Barrett, L., McCabe, J., Costa, J., Bevaqua, Relations between affect and personality: Support
L., & Bliss, M. (1999, August). Explicit versus for the affect-level and affective reactivity views.
implicit measures of attachment. Paper presented Personality and Social Psychology Bulletin, 24,
at the 107th Annual Convention of the American 279 -288.
Psychological Association, Boston. Halberstadt, J. B., & Niedenthal, P. M. (1997). Emo-
Fishtein, J., Pietromonaco, P. R., & Feldman Barrett, tional state and the use of stimulus dimensions in
L. (1999). The contribution of attachment style and judgment. Journal of Personality and Social Psy-
relationship conflict to the complexity of relation- chology, 72, 1017-1033.
ship knowledge. Social Cognition, 17, 228-244. Hazan, C., & Shaver, P. R. (1987). Romantic love
Fox, N. A., Kimmerly, N. L., & Schafer, W. D. conceptualized as an attachment process. Journal
(1991). Attachment to mother/attachment to fa- of Personality and Social Psychology, 52, 511-
ther: A meta-analysis. Child Development, 62, 524.
210-225. Hazan, C., & Shaver, P. R. (1990). Love and work:
Fraley, R. C., & Davis, K. E. (1997). Attachment An attachment-theoretical perspective. Journal of
formation and transfer in young adults' close Personality and Social Psychology, 59, 270-280.

Hazan, C., & Shaver, P. R. (1994). Attachment as an Markus, H. (1977). Self-schemata and processing
organizational framework for research on close information about the self. Journal of Personality
relationships. Psychological Inquiry, 5, 1-22. and Social Psychology, 35, 63-78.
Higgins, E. T. (1987). Self-discrepancy: A theory Markus, H., & Cross, S. (1990). The interpersonal
relating self and affect, Psychological Review, 94, self. In L. A. Pervin (Ed.), Handbook of personal-
319-340. ity: Theory and research (pp. 576-608). New
Hinkley, K., & Andersen, S. M. (1996). The working York: Guilford Press.
self-concept in transference: Significant other ac- Markus, H. R., & Kitayama, S. (1991). Culture and
tivation and self change. Journal of Personality the self: Implications for cognition, emotion, and
and Social Psychology, 71, 1279-1295. motivation. Psychological Review, 98, 224-253.
Kihlstrom, J. F., & Klein, S. B. (1994). The self as a Markus, H., & Kunda, Z. (t986). Stability and mal-
knowledge structure. In R. S. Wyer, Jr., & T. K. leability of the self-concept. Journal of Personality
Srull (Eds.), Handbook of social cognition (2nd and Social Psychology, 51, 858-866.
ed., pp. 153-208). Hillsdale, NJ: Erlbaum. Markus, H., & Nurius, P. (1986). Possible selves.
Kirkpatrick, L. A., & Davis, K. E. (1994). Attach- American Psychologist, 41, 954-969.
ment style, gender, and relationship stability: A Markus, H., & Wurf, E. (1987). The dynamic self-
longitudinal analysis. Journal of Personality and concept: A social psychological perspective. An-
Social Psychology, 66, 502-512. nual Review of Psychology, 38, 299-337.
Kirkpatrick, L. A., & Shaver, P. R. (1992). An at- McGuire, W. J., & McGuire, C. V. (1988). Content
tachment-theoretical approach to romantic love and process in the experience of self. In L.
and religious belief. Personality and Social Psy- Berkowitz (Ed.), Advances in experimental social
chology Bulletin, 18, 266-275. psychology (Vol. 21, pp. 97-144). San Diego, CA:
Klohnen, E. C., & Bera, S. (1998). Behavioral and Academic Press.
experiential patterns of avoidantly and securely Mead, G. H. (1934). Mind, self, andsoc&ty. Chicago:
attached women across adulthood: A 31-year lon- University of Chicago Press.
gitudinal perspective. Journal of Personality and Mikulincer, M. (1995). Attachment style and the
Social Psychology, 74, 211-223. mental representation of the self. Journal of Per-
Klohnen, E. C., & John, O. P. (1998). Working sonality and Social Psychology, 69, 1203-1215.
models of attachment: A theory-based prototype Mikulincer, M. (1998a). Adult attachment style and
approach. In J. A. Simpson & W. S. Rholes (Eds.),
affect regulation: Strategic variations in self-ap-
Attachment theory and close relationships (pp. praisals. Journal of Personality and Social Psy-
115-140). New York: Guilford Press.
chology, 75, 420-435.
Kobak, R. R. (1994). Adult attachment: A personality
Mikulincer, M. (1998b). Adult attachment style and
or relationship construct? Psychological In-
affect regulation: Strategic variations in subjective
quiry, 5, 42-44.
self-other similarity. Journal of Personality and
Lutz, C. A. (1990). Engendered emotion: Gender,
power and the rhetoric of emotional control in Social Psychology, 75, 436-448.
Mikulincer, M. (1998c). Attachment working models
American discourse. In C. A. Lutz & L. Abu-
Lughod (Eds.), Language and the politics of emo- and the sense of trust: An exploration of interac-
tion (pp. 69-91). Cambridge, England: Cambridge tion goals and affect regulation. Journal of Per-
University Press. sonality and Social Psychology, 74, 1209-1224.
Main, M. (1991). Metacognitive knowledge, meta- Mikulincer, M., & Florian, V. (1998). The relation-
cognitive monitoring, and singular (coherent) vs. ship between adult attachment styles and emo-
multiple (incoherent) model of attachment: Find- tional and cognitive reactions to stressful events. In
ings and directions for future research. In C. M. J. A. Simpson & W. A. Rholes (Eds.), Attachment
Parkes, J. Stevenson-Hinde, & P. Marris (Eds.), theory and close relationships (pp. 143-165). New
Attachment across the life cycle (pp. 127-159). York: Guilford Press.
London: Tavistock/Routledge. Mikulincer, M., Florian, V., & Weller, A. (1993).
Main, M., Kaplan, K., & Cassidy, J. (1985). Security Attachment styles, coping strategies, and post-
in infancy, childhood, and adulthood: A move to traumatic psychological distress: The impact of the
the level of representation. Monographs of the Gulf War in Israel. Journal of Personality and
Society for Research in Child Development, Social Psychology, 64, 817-826.
50(1-2, Serial No. 209). Mikulincer, M., & Nachshon, O. (1991). Attachment
Main, M., & Weston, D. R. (1982). Avoidance of the styles and patterns of self-disclosure. Journal of
attachment figure in infancy: Descriptions and in- Personality and Social Psychology, 61, 321-331.
terpretations. In C. M. Parkes & J. Stevenson- Mikulincer, M., & Orbach, I. (1995). Attachment
Hinde (Eds.), The place of attachment in human styles and repressive defensiveness: The accessi-
behavior (pp. 31-59). New York: Basic Books. bility and architecture of affective memories. Jour-

nal of Personality and Social Psychology, 68, 917- Schacter, D. L. (1996). Searching for memory: The
925. brain, the mind, and the past. New York: Basic
Niedenthal, P. M., Halberstadt, J. B., & Innes-Ker, Books.
A. H. (1999). Emotional response categorization. Schlenker, B. R., & Weigold, M. F. (1989). Goals
Psychological Review, 106, 337-361. and the self-identification process: Constructing
Niedenthal, P. M., Halberstadt, J. B., & Setterlund, desired identities. In L. A. Pervin (Ed.), Goals
M. B. (1997). Being happy and seeing "happy": concepts in personality and social psychology (pp.
Emotional state mediates visual word recognition. 243-290). Hillsdale, NJ: Erlbaum.
Cognition and Emotion, 11, 403-432. Shaver, P. R., & Brennan, K. (1992). Attachment
Nisbett, R. E., & Wilson, T. D. (1977). Telling more styles and the "Big Five" personality traits: Their
than we can know: Verbal reports on mental pro- connections with each other and with romantic
cesses. Psychological Review, 84, 231-259. relationship outcomes. Personality and Social Psy-
Ogilvie, D. M., & Ashmore, R. D. (1991). Self-with- chology Bulletin, 18, 536-545.
other representations as a unit of analysis in self- Shaver, P. R., Collins, N., & Clark, C. (1996). At-
concept research. In R. C. Curtis (Ed.), The rela- tachment styles and internal working models of
tional self." Theoretical convergences in psycho- self and relationship partners. In G. J. O. Fletcher
analysis and social psychology (pp. 282-314). & J. Fitness (Eds.), Knowledge structures in close
New York: Guilford Press. relationships: A social psychological approach
Pietromonaco, P. R., & Camelley, K. B. (1994). (pp. 25-61). Hillsdale, NJ: Erlbaum.
Gender and working models of attachment: Con- Shaver, P. R., Hazan, C., & Bradshaw, D. (1988).
sequences for perceptions of self and romantic Love as attachment: The integration of three be-
relationships. Personal Relationships, 1, 63-82.
havioral systems. In R. J. Sternberg & M. Barnes
Pietromonaco, P. R., & Feldman Barrett, L. (1997a,
(Eds.), The psychology of love (pp. 68-99). New
October). Internal working models: What do we
Haven, CT: Yale University Press.
know about knowing about the self in relation to
Simpson, J. A. (1990). The influence of attach-
others? Paper presented at the annual meeting of
ment styles on romantic relationships. Journal of
the Society of Experimental Social Psychology,
Personality and Social Psychology, 59, 971-
Toronto, Ontario, Canada.
Pietromonaco, P. R., & Feldman Barrett, L. (1997b).
Working models of attachment and dally social Simpson, J. A., & Rholes, W. S. (1994). Stress and
secure base relationships in adulthood. In K. Bar-
interactions. Journal of Personality and Social
Psychology, 73, 1409-1423. tholomew & D. Perlman (Eds.), Attachment pro-
Pietromonaco, P. R., & Feldman Barrett, L. (1999a). cesses in adulthood (pp. 181-204). London: Jes-
Emotional reactivity and emotional regulation: sica Kingsley.
Two affect-related processes underlying attach- Simpson, J. A., Rholes, W. S., & Nelligan, J. S.
ment dynamics in adulthood. Manuscript in prep- (1992). Support seeking and support giving within
aration. couples in an anxiety-provoking situation: The role
Pietromonaco, P. R., & Feldman Barrett, L. (1999b). of attachment styles. Journal of Personality and
Working models of attachment and defensive pro- Social Psychology, 62, 434-446.
cesses. Manuscript in preparation. Simpson, J. A., Rholes, W. S., & Phillips, D. (1996).
Rosenberg, M. (1988). Self and others: Studies in Conflict in close relationships: An attachment per-
social personality and autobiography. In L. spective. Journal of Personality and Social Psy-
Berkowitz (Ed.), Advances in experimental social chology, 71, 899-914.
psychology (Vol. 21, pp. 57-95). San Diego, CA: Sroufe, L. A., & Waters, E. (1977). Attachment as an
Academic Press. organizational construct. Child Development, 48,
Ross, M. (1989). Relation of implicit theories to the 1184-1199.
construction of personal histories. Psychological Thompson, R. A., Lamb, M. E., & Estes, D. (1983).
Review, 96, 341-357. Harmonizing discordant notes: A reply to Waters.
Rothbard, J. C., & Shaver, P. R. (1994). Continuity of Child Development, 54, 521-524.
attachment across the life span. In M. B. Sperling Tidwell, M. O., Reis, H. T., & Shaver, P. R. (1996).
& W. H. Berman (Eds.), Attachment in adults: Attachment, attractiveness, and social interaction:
Clinical and developmental perspectives (pp. 31- A diary study. Journal of Personality and Social
71). New York: Guilford Press. Psychology, 71, 729-745.
Rothbart, M. K., & Derryberry, D. (1981). Develop- Tomaka, J., & Blascovich, J. (1994). Effects of jus-
ment of individual differences in temperament. In tice beliefs on cognitive appraisal of and subjective
M. E. Lamb & A. L. Brown (Eds.), Advances in physiological and behavioral response to potential
developmental psychology (Vol. 1, pp. 37-86). stress. Journal of Personality and Social Psychol-
Hillsdale, NJ: Erlbaum. ogy, 67, 732-740.

Trinke, S. J., & Bartholomew, K. (1997). Hierarchies place of attachment in human behavior (pp. 171-
of attachment relationships in young adulthood. 184). New York: Basic Books.
Journal of Social and Personal Relationships, 14, Williams, J. M. G., Mathews, A., & MacLeod, C.
603-625. (1996). The emotional Stroop task and psychopa-
Vaughn, B. E., Egeland, B. R., Sroufe, L. A., & thology. Psychological Bulletin, 120, 3-24.
Waters, E. (1979). Individual differences in infant-
mother attachment at 12 and 18 months: Stability
and change in families under stress. Child Devel-
opment, 50, 971-975. Received March 12, 1999
Weiss, R. S. (1982). Attachment in adult life. In Revision received September 22, 1999
C. M. Parkes & J. Stevenson-Hinde (Eds.), The Accepted September 24, 1999 •