Vous êtes sur la page 1sur 30

ATTACHMENT STYLE AND MEMORY

FOR ATTACHMENT-RELATED EVENTS

Judi Beinstein Miller


Oberlin College

ABSTRACT
Previous research has demonstrated attachment-schematic
processing of attachment-related stories when attachment-
related memories are activated before exposure to a story.
This study investigated the possibility of obtaining attach-
ment-schematic processing without the activation of attach-
ment-related memories. Undergraduate participants provided
attachment style self-ratings and several weeks later read
either a positive or negative version of a friendship story.
After a distractor task, they took a recall test about the story.
Secure and fearful attachment ratings were positively associ-
ated with recall of focal story events, relative to peripheral
story details, when participants read the negative story.
However, secure attachment was associated with recall of
focal events involving the friends’ joint activities, whereas
fearful attachment was associated with recall of those involv-
ing their separate activities. These and other results are dis-
cussed in terms of attachment-related models of self and
other.

KEY WORDS • attachment style • memory • working models

Beliefs about attachment can influence interpersonal understanding by


guiding the processing of attachment-related events. Yet despite their
potentially important cognitive role, surprisingly little research has exam-
ined their information-processing effects. The purpose of the present study
was to investigate the information-processing effects of attachment beliefs
by means of a memory paradigm. The research was intended to extend
knowledge about the circumstances in which attachment beliefs can influ-
ence the processing of attachment-related events.

The author is grateful to Gavin Malcolm whose patience in the construction of stimulus
materials and coding of story recall helped bring this study to fruition. Thanks are also due to
Evelyn Leung for her independent coding of story recall and to the anonymous reviewers for
their constructive and helpful suggestions on an earlier version of this manuscript. Address
correspondence to Judi Beinstein Miller, Department of Psychology, Oberlin College,
Oberlin, OH, USA. [E-mail: judith.miller@oberlin.edu].

Journal of Social and Personal Relationships Copyright © 1999 SAGE Publications (London,
Thousand Oaks, CA and New Delhi), Vol. 16(6): 773–801. [0265–4075 (199912) 16:6; 010454]
774 Journal of Social and Personal Relationships 16(6)

According to attachment theory (Bowlby, 1973), stability in attachment


behavior is afforded by the generalized beliefs about close relationships
that individuals develop from consistencies in their memories about past
relationship experiences. These beliefs, along with their constitutive mem-
ories, are thought to be internalized as ‘working models’ of attachment and
to perpetuate themselves by directing information processing resources
(Shaver, Collins, & Clark, 1996). Working models of attachment are
believed to influence attention to and interpretation of attachment-related
events. Consistent with this idea, correlations have been found between
young adults’ attachment styles and their appraisals of romantic partners
(Young & Acitelli, 1998), expectations of treatment from partners
(Baldwin, Fehr, Keedian, Seidel, & Thomson, 1993; Baldwin, Keelan, Fehr,
Enns, & Koh-Rangarajoo, 1996), and attributions for partners’ ambiguous
behavior (Collins, 1996). Working models of attachment thus contribute
meaning to relationship experiences and prepare individuals for future
interaction.
Working models of attachment are also thought to facilitate the encod-
ing and retrieval of information that is consistent with attachment experi-
ences and beliefs and to result thereby in attachment-schematic
information processing (Collins & Read, 1994). Evidence for this idea is
provided by correlations between 3-year-olds’ recognition of positive and
negative events from a puppet show and their attachment styles (Belsky,
Spritz, & Crnic, 1996). Securely attached children recognized positive
events more accurately than negative ones, whereas insecurely attached
children recognized negative events more accurately than positive ones.
Similar correlations have been obtained between college students’ attach-
ment styles and their recall of positive and negative story events, but only
when their attachment beliefs were activated before they read the story
(Miller & Noirot, 1999).
In related research, Baldwin and his colleagues (1993) have found that
college students’ reaction times to recognize positive versus negative inter-
personal outcomes are consistent with their attachment style. Secure par-
ticipants were quicker in identifying words that represented positive
outcomes, whereas insecure participants were quicker in identifying words
that represented negative outcomes. Similarly, Mikulincer (1998) found
that latencies to retrieve personal experiences of trust were associated with
attachment style. Securely attached participants retrieved positive experi-
ences faster, whereas insecurely attached participants retrieved negative
experiences faster. Results such as these have begun to provide evidence
for attachment-schematic information processing of attachment-related
events. Secure attachment appears to be associated with recognition and
recall of positive relationship events, whereas insecure attachment appears
to be associated with recognition and recall of negative ones. These are
information processing biases that can reinforce attachment beliefs by pro-
viding evidence for their validity. Attachment-related differences in inter-
personal perception and relationship functioning can thus be perpetuated
by attachment-schematic processing of attachment-related events.
Beinstein Miller: Attachment and memory 775

Activation of attachment beliefs and information


processing

Of particular relevance to the present research are results from Miller and
Noirot’s study (1999) of attachment style correlates of story encoding. In
this study, memories of supportive and rejecting experiences in close
relationships were expected to activate attachment-relevant material in
memory and affect the encoding of subsequent attachment-related infor-
mation. The four attachment styles defined by Bartholomew (1990) —
secure, dismissing, preoccupied, and fearful — were used to reflect the
beliefs that secure and insecure individuals have about themselves versus
the significant others in their environments (Bowlby, 1973).
Undergraduates provided self-ratings on each of these attachment styles
and several weeks later were randomly assigned to write an account of
either a supportive or rejecting friendship experience to activate their
attachment beliefs. Half then read a story in which the main character
experienced an equal number of positive and negative events in a close
friendship. After reading the story they solved a set of figural analogies and
then responded to a cued recall test about the story. The other half read the
story after solving the analogies, but before writing their friendship
accounts.
Results from this experiment provided evidence that the activation of
attachment-related memories would influence story encoding and that the
nature of their influence would depend on participants’ attachment styles.
The more fearfully attached the participant, the better he/she recalled
negative focal events from the story when either supportive or rejecting
memories had been activated. The more securely attached the participant,
the better he/she recalled positive focal events, but only when rejecting
memories had been activated. These results occurred when attachment-
related memories were activated before rather than after the story and
therefore supported an attachment-schematic encoding hypothesis. Their
asymmetry could have been caused by differences in accessibility of attach-
ment beliefs between the securely and fearfully attached. As suggested by
Simpson and Rholes (1994), attachment beliefs might be easier to activate
among those who are chronically concerned about their relationships, and
fearfully attached individuals are by definition chronically more concerned.
Benign attachment memories might be sufficient to activate their attach-
ment beliefs, whereas more threatening, negative memories might be
required to activate the beliefs of the securely attached.
This experiment also provided evidence that the activation of attach-
ment-related memories could influence the processing of attachment-unre-
lated information and that the nature of their influence would depend on
participants’ attachment styles. Scores on the figural analogies test were
positively associated with dismissing attachment and negatively associated
with fearful attachment, when participants’ memories of rejection had been
activated before the story and test. These results are consistent with prior
research in which participants with dismissing attachment styles have been
776 Journal of Social and Personal Relationships 16(6)

better able to suppress negative relationship memories than have those with
more consciously anxious styles (Fraley & Shaver, 1997). They are consis-
tent with prior research in which anxious attachment and perceptions of dif-
ferences in conflict responses were negatively correlated among study
participants who had previously written accounts about rejecting relation-
ship experiences (Miller, 1996). Results from this study thus provided evi-
dence for attachment-related allocation of information processing resources
as well as for attachment-schematic encoding of attachment-related infor-
mation when attachment beliefs are activated. However, the results also
raise three questions, addressed by the present study, about the contexts of
information processing in which attachment style effects can be obtained.

Research questions

The first question concerns the possibility of obtaining attachment-


schematic information-processing effects when attachment-related mem-
ories are not activated. Belsky and his colleagues (1996) have obtained
attachment-schematic information-processing effects among 3-year-olds
without activating their attachment memories, but 3-year-olds’ working
models of attachment would be less complex than those of adults and more
reflective of a single attachment style. Although children often develop dif-
ferent attachment bonds with their parents (cf. Bridges, Connell, & Belsky,
1988 and Fox, Kimmerly, & Schafer, 1991), by adulthood they can report
multiple attachment styles because of additional attachment experiences in
peer relationships (Baldwin et al., 1996; Hazan & Zeifman, 1994; Trinke &
Bartholomew, 1997). A currently salient experience that is inconsistent
with an individual’s general attachment expectations might diminish the
information-processing effects of his/her more general attachment beliefs.
In the Miller and Noirot (1999) study, attachment style effects on story
recall were obtained only when participants’ attachment-related memories
were activated before they read the story. Those whose memories were
activated after the story did not manifest attachment-schematic recall
effects. Consequently, it appeared as though information-processing effects
depended on the activation of participants’ attachment-related memories.
However, the activation of participants’ memories after the story and
before the recall test could have created retrieval effects that masked
effects at encoding. For example, writing about a personal experience after
the story could have activated a variety of memories and/or anxieties that
used up information processing resources and interfered with retrieval of
the story, even though its initial encoding might have been good. In
addition, it is not clear whether the information biases obtained in this
study reflected differences in participants’ general attachment expectations
alone or differences in the ways that specific attachment memories primed
attachment-schematic encoding. A better test of the independent effects of
attachment style would be to have participants read and recall the story
without deliberately activating their attachment memories.
Beinstein Miller: Attachment and memory 777

The second question concerns the cognitive mechanism by which infor-


mation processing is biased. Do individuals recall attachment style-congru-
ent information better than attachment style-incongruent information
because they are more attentive to congruent information or because they
encode, or assimilate, congruent information better? Attachment schemas
are believed to develop from attachment-related experiences and to reflect
the subjective quality of those experiences (Bowlby, 1973). Individuals
whose attachment-related experiences have been primarily positive should
therefore have a larger number of richly interconnected positive exemplars,
relative to negative exemplars, for matching and assimilating ongoing
attachment-related events. Those with primarily negative experiences
should have a larger number of negative exemplars, relative to positive
ones, for the matching and assimilation of events. The attachment schemas
of secure individuals should thus provide a larger network of pathways for
the assimilation of positive attachment-related events than of negative
ones, whereas the attachment schemas of insecure individuals should pro-
vide a larger network of pathways for the assimilation of negative events
than of positive ones. Yet if individuals do not exhibit attachment-
schematic encoding of purely positive and purely negative events then such
assimilation effects would be questionable. In prior research, participants
have been asked to process attachment-related events that are mixed —
half are positive and half are negative. Consequently each type of event is
made distinctive in the context of the other and can draw attention from the
other. In a context of contrasting events, superior assimilation of one or the
other could be caused by participants’ allocation of attentional resources
rather than by their superior assimilation of events per se. If only positive
or negative events were presented, these would lose the distinctiveness
advantage in attracting attention. If attachment-congruent recall was
nevertheless obtained then this would implicate an explanation based on
assimilation processes. If attachment-congruent recall was not obtained
then this would implicate an explanation based on allocation of attention.
The third question concerns the nature of the stimulus that is subject to
attachment style-related information processing biases. Are attachment-
related biases caused by the positive or negative outcomes of events or by
their relational nature? In the friendship story used by Miller and Noirot
(1999), the consequences of the friends’ joint activities tended to be posi-
tive, whereas the consequences of their separate actions were more often
negative. Consequently, the tendency for securely attached persons to
favor the processing of positive events could have occurred because of the
joint activities that constituted the events rather than because of their
positive outcomes per se. Similarly, the tendency for fearfully attached
individuals to favor the processing of negative events could have been
caused by the separate activities that constituted the events rather than by
their negative outcomes per se. With purely positive or purely negative
stories it would be possible to compare memory for joint and separate
activities by the valence of outcomes and clarify the reason for bias.
The possibility of obtaining attentional differences to the story charac-
778 Journal of Social and Personal Relationships 16(6)

ters’ joint and separate activities is also suggested by Bartholomew’s (1990)


typology of attachment styles. This typology is based on Bowlby’s (1973)
idea that individuals use working models of self and others to perceive
events, predict the future, and plan accordingly. The model of self includes
beliefs about self-worth and worthiness for care. The more positive the
model the less the anxiety about being loved and the less the dependency
on others for approval. In contrast, the more negative the model the greater
the preoccupation with partners’ love and the greater the dependence on
others for self-esteem. The model of others includes beliefs about partners’
availability and supportiveness. The more positive the model the less the
avoidance of closeness and dependency. In contrast, the more negative the
model the greater the independence and self-reliance. Bartholomew pro-
posed that four attachment prototypes could be derived by dichotomizing
models of self and others as positive or negative. Secure attachment would
be characterized by positive models of self and others, whereas fearful
attachment would be characterized by negative models of both.
Preoccupied attachment would be characterized by negative models of self
and positive models of others, whereas dismissing attachment would be
characterized by positive models of self and negative models of others.
If it is assumed that positive models direct attention to sources of well-
being, whereas negative models direct attention to sources of threat, then
specific predictions can be made about attachment-related differences in
attention to partners and self. Secure and fearful individuals would be
expected to focus attention on both the partner and self, the former to
maintain proximity to sources of well-being and the latter to monitor
threat, because the models of the former are both positive, whereas the
models of the latter are both negative. Moreover, because feelings of threat
increase anxiety, the attention of fearful individuals would be heightened
relative to that of secure individuals. Dismissing and preoccupied individ-
uals would experience conflicting motives to monitor sources of threat and
maintain proximity to sources of well-being because one of their models is
positive and the other is negative. According to attachment theory and
research, both would resolve the conflict by avoiding threat and seeking
sources of potential well-being. Bowlby (1980) wrote that consistent rejec-
tion by attachment figures encouraged individuals to deactivate their
attachment systems and prevent future rejection. To maintain a deactivated
attachment system they avoided situations and information — including
their attachment figures — that stimulated attachment behavior and feel-
ings. Instead of seeking support from others they learned to rely on them-
selves, a consequence that has been demonstrated in studies of both
children and adults (Fraley, Davis, & Shaver, 1998). The lesson that others
are not responsive can thus direct attention to the self, as appears to
happen for dismissing individuals.
In contrast, individuals who have received inconsistent support from
attachment figures come to doubt their self-worth and depend excessively
on others for reassurance. They develop an attachment style that has been
labeled ambivalent or resistant in childhood and preoccupied in adulthood.
Beinstein Miller: Attachment and memory 779

Ambivalent infants express excessive anxiety when separated from their


mothers, monitor their mothers’ behavior at the expense of play, yet are
not comforted by their mothers following separation from them
(Ainsworth, Blehar, Waters, & Wall, 1978). Preoccupied adults report low
self-esteem, excessive jealousy, and obsession with love relationships that
can interfere with work and productivity (Hazan & Shaver, 1987; Shaver &
Hazan, 1990). The lesson of inconsistent care from others can thus direct
their attention toward partners.
If story readers take the main character’s perspective then their models
of self and others might influence their attention to and hence memory for
story events. Readers with dismissing and fearful attachment styles would
process information about the main character, relative to other infor-
mation, better than those with secure or preoccupied styles. Readers with
preoccupied and fearful attachment styles would process information about
the main character’s partner, relative to other information, better than
those with secure or dismissing attachment styles.
These predictions suggest that securely attached readers might use up
fewer resources than other readers in processing information about story
characters’ separate activities and have more available for processing other
information, such as about their relationship and interaction. Readers with
fearful attachment styles and heightened concerns from two sources of
threat — self and other — would use most of their attentional resources
processing information about the main character and his/her partner.
Readers with preoccupied attachment styles would have heightened con-
cerns for positive responses from other people and would focus their atten-
tion on activities of the partner. Similarly, readers with dismissing
attachment styles would have heightened concerns for self-reliance and
focus their attention on the main character. However, the ability of dismiss-
ing individuals to suppress attachment-related concerns might also result
simply in their paying less attention to the character’s partner, leaving more
resources for the processing of other information. Therefore, an additional
prediction is that securely attached readers, and possibly those with dis-
missing attachment styles, will have a memory advantage for information
about story characters’ joint activities, relative to their separate activities,
because they have more resources available to process this information.

The present study

The present study addressed these three questions: (i) eliciting attachment-
schematic information processing without priming, (ii) the cognitive mech-
anism that elicits attachment-schematic information processing, and (iii)
the nature of the stimulus that is subject to attachment-related information-
processing biases — by changing three aspects of Miller and Noirot’s
experimental design. To determine whether attachment-schematic encod-
ing could be obtained without activating attachment-related memories,
participants were asked to read and recall a friendship story, but were not
780 Journal of Social and Personal Relationships 16(6)

asked to write about a personal friendship experience. Evidence for


attachment-schematic information processing would be provided by
positive correlations between secure attachment and recall of positive focal
events, and between fearful attachment and recall of negative focal events.
To determine whether information-processing effects could be caused by
participants’ superior assimilation of attachment style-congruent infor-
mation, two versions of the story were written, one in which the outcomes
of events were purely positive and one in which the outcomes were purely
negative. Participants read and recalled either a positive or a negative story,
rather than one with mixed outcomes that could compete differentially for
their attention. Evidence for superior assimilation of attachment-congruent
information would be demonstrated by positive correlations between
secure attachment and recall of focal events from the positive story, and by
positive correlations between fearful attachment and recall of focal events
from the negative story.
No predictions were made about the effects of dismissing or preoccupied
attachment on recall of focal events for both theoretical and empirical
reasons. Theoretically, both dismissing and preoccupied attachment styles
should be associated with the superior processing of negative relationship
events relative to positive ones, because both are insecure attachment
styles. However, additional factors complicate the information processing
of individuals with these attachment styles. Dismissing attachment is associ-
ated with suppression of negative relationship events (Fraley & Shaver,
1997). Preoccupied attachment is associated with anxiety, which can inter-
fere with information processing (Miller, 1996). Unlike fearful attachment,
which is also associated with anxiety, preoccupied attachment is not tem-
pered by avoidance of attachment-related concerns. Attachment-related
concerns might therefore be more easily activated among the preoccupied
than among the fearful with the consequence of distracting the former
more than the latter. Consistent with these complications, correlations of
story recall with fearful attachment, but not with dismissing or preoccupied
attachment, have been statistically significant in previous research (Miller
& Noirot, 1999).
To determine whether biased attention to story events was caused by the
valence of story outcomes or their relational nature, the positive and nega-
tive story were each written to contain both joint and separate activities of
the friends. Participants read and recalled either positive or negative infor-
mation about the friends’ joint and separate activities. If information-pro-
cessing effects were caused by the valence of story outcomes then a positive
correlation would be obtained between fearful attachment and recall of
focal events from the negative story, regardless of their relational nature. A
positive correlation might also be obtained between secure attachment and
recall of focal events from the positive story, regardless of their relational
nature.
In contrast, if information-processing effects were caused by the rela-
tional nature of story events, then positive correlations might be obtained
between fearful attachment and recall of the friends’ separate activities,
Beinstein Miller: Attachment and memory 781

regardless of whether the story was positive or negative. A positive corre-


lation might be obtained between secure attachment and recall of their
joint activities, regardless of story version. Positive correlations might also
be obtained between dismissing attachment and recall of the main charac-
ter’s separate activities, and between preoccupied attachment and recall of
the partner’s separate activities. Interaction effects also could occur if the
friends’ separate and joint activities became differentially salient in nega-
tive circumstances only, that is, when the friends’ activities posed threats to
each other and to their friendship. Positive correlations would then be
obtained only among participants who read the negative story.
Participants read the friendship story, solved a set of figural analogies,
and then took the recall test about the story. The figural analogies test was
intended primarily as a distractor task, but correlations between figural
analogies scores and attachment ratings were also possible if attachment-
related memories or concerns were stimulated by the story, lingered during
the test, and used up problem-solving resources. Because preoccupied and
fearful attachment are both characterized by anxiety over attachment
relations, correlations between both these attachment styles and test scores
were anticipated if the friendship story activated attachment-related mem-
ories or concerns. Miller and Noirot (1999) reported evidence of this effect.
Fearful attachment and figural analogies test scores were negatively associ-
ated among participants who had written about a rejecting friendship
experience. A similar but statistically non-significant association was
obtained for preoccupied attachment.
In summary, the major purpose of this study was to determine whether
attachment-schematic information processing could be obtained using uni-
formly positive or negative stimulus materials, without deliberately activat-
ing participants’ attachment-related memories. It would have been
additionally informative to estimate the effects of deliberately priming the
attachment system, by including additional groups of participants whose
attachment memories were activated before the story. However, the pur-
pose of this study was to determine whether information-processing effects
could be obtained without activating attachment-related memories rather
than to determine the added effects of a priming manipulation.
Attachment-schematic information processing would be demonstrated by
positive correlations between secure attachment and recall of focal details
from the positive story and between insecure forms of attachment and
recall of focal details from the negative story. In addition, it was possible to
determine whether the relational nature of story events would make a dif-
ference in patterns of recall.

Method
Participants
There were 182 undergraduates from an introductory psychology course who
participated in return for partial course credit. Of the 179 who provided com-
plete information about their attachment style, 78.2 percent were Caucasian,
782 Journal of Social and Personal Relationships 16(6)

8.4 percent were African-American, 6.1 percent were Asian, 3.4 percent were
Latino, and 3.9 percent were of other ethnic backgrounds. Of the participants
with complete attachment information, there were 35 males and 58 females
who read the positive story and 29 males and 57 females who read the negative
one. Most (82%) were first-year students. Their romantic relationship status
was not ascertained. Participants provided information about their attachment
styles early in the semester and 2–4 weeks later participated in the experimen-
tal session. They were tested in groups of 20 to 30.

Pretest materials
Early in the semester, participants rated themselves on Bartholomew and
Horowitz’s (1991) four attachment prototypes using 5-point Likert scales (1 
strongly disagree to 5  strongly agree). Participants rated themselves on the
prototypes instead of selecting the single prototype that best represented their
style because individuals commonly develop multiple attachment relationships
that can create different attachment expectations (Baldwin et al., 1996; Bowlby,
1969). Within this sample, only 45 percent had a single attachment style, as
defined by ratings of 4 on one prototype and ratings of 3 on the other pro-
totypes. Twenty-seven percent had multiple attachment styles, as defined by
equally high ratings above 3 on two or more prototypes. Another 13 percent
rated two or more prototypes above 3 but not equally high, thereby indicating
moderately strong secondary styles, and the remaining 14 percent rated each of
the prototypes at 3, thereby indicating moderately undifferentiated attach-
ment expectations.
Of those participants who rated one style highest, 43 percent designated
themselves as secure, 12 percent designated themselves as dismissing, 20 per-
cent designated themselves as preoccupied, and 25 percent designated them-
selves as fearful. This distribution is consistent with Brennan, Clark, and
Shaver’s (1998) figures for a sample of 1086 undergraduates, which were 46, 15,
16, and 23 percent, respectively.
Mean ratings for secure, dismissing, preoccupied, and fearful attachment
were 3.01 (SD  1.29), 2.80 (SD  1.18), 2.73 (SD  1.36), and 2.78 (SD 
1.48), respectively. As indicated in Table 1, there was a moderately strong
negative correlation between secure and fearful attachment ratings and a small
negative correlation between dismissing and preoccupied attachment ratings.
Secure attachment ratings also had small negative correlations with dismissing
and preoccupied attachment ratings.

Stimulus materials and experimental procedures


At the beginning of the experiment, participants received test booklets with a
cover page that summarized what they would be doing and concealed the story
version from the experimenter. The test booklets were counterbalanced for dis-
tribution by story version so that approximately half of the participants would
read the positive story and the other half would read the negative story. They
were counterbalanced separately for males and females because male charac-
ters were used in the story for male participants and female characters were
used in the story for female participants.
The story was about Emily (or Eric in the males’ version) and her (or his)
long-term friend, Sara (Scott in the males’ version). It is told from
Emily’s/Eric’s perspective while she/he waits at the airport for the arrival of
Sara’s/Scott’s plane. Emily/Eric and Sara/Scott had been friends since elemen-
Beinstein Miller: Attachment and memory 783

tary school. In third grade they worked on a science project together, and
Sara/Scott, because both of her/his parents worked full time, went to
Emily’s/Eric’s house every day after school. In fifth grade they helped organize
an activities carnival, but each worked on a separate activity. After elementary
school they became involved in separate curricular activities and Sara/Scott
developed another friendship with someone else. But when they were seniors
in high school they took a trip to Florida and spent a lot of time doing things
together. Their last meeting had been before Sara/Scott left for college, and at
that meeting they reminisced and thought about the future.
There were eight focal episodes in the story and a variety of peripheral details
that provided additional information but were not essential aspects of the story.
An example of a peripheral detail would be the name of the music group that
the friends listened to on their road trip or the name of a movie that the friends
saw together. Focal events involved their joint activities, such as their interac-
tion over the science project, and/or their separate activities, such as their devel-
opment of separate extracurricular interests after elementary school. Two
versions of the story were written, one in which the focal events had positive
outcomes (positive story condition) and the other in which the outcomes were
negative (negative story condition). For example, the positive version of the
grade school project depicts the friends joking about their project and having a
good time, but the negative version depicts them arguing about their project
and having a miserable time. Thus participants read about a main character,
Emily/Eric, who experienced either uniformly positive or uniformly negative
outcomes largely because of her/his friend, Sara/Scott. A copy of the negative
version, which elicited more attachment-schematic recall than did the positive
version, is appended. Participants were given 7 minutes to read the story.
For the figural analogies task, 20 problems were selected from Thorndike
and Hagen’s Cognitive Abilities Tests (1971, 1974). Participants were given 9
minutes to solve the analogies and told not to worry if they could not complete
them all.
After the figural analogies task, the test packets were collected and a cued
recall test about the story was administered. Participants were given 15 minutes
to answer the questions as precisely as they could. Twenty-five questions were
about the focal events and 13 were about the peripheral details. Of the ques-
tions about the focal events, 9 were about the friends’ joint activities or feelings
(e.g., How did things in Sara’s/Scott’s and Emily’s/Eric’s relationship change
after elementary school?) and 16 were about their separate activities or feelings
(e.g., How did Emily/Eric feel about Sara’s/Scott’s relationship with
Amy/Alex?), 9 about Emily’s/Eric’s and 7 about Sara’s/Scott’s.

Analytic procedures
Scoring of the recall test was conducted by giving 2 points for each correct
answer, 1 point for partially correct answers, and 0 points for incorrect answers.
An independent coder scored one-fourth of the tests. Coding agreement was
94.7 percent, Cohen’s  .91.
Four recall scores were computed for use in subsequent analyses, the pro-
portion of total recall (0–76 points) as a result of memory for focal events (0–50
points), and proportions for the three types of events that constituted the focal
events, the friends’ joint activities (0–18 points), Emily’s/Eric’s separate activi-
ties (0–18 points), and Sara’s/Scott’s separate activities (0–14 points). A fifth
recall score, the proportion of total recall because of memory for both friends’
784
TABLE 1
Intercorrelations of study variables

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.


1. Condition (Positive)
2. Sex (Male) .04
Attachment rating

Journal of Social and Personal Relationships 16(6)


3. Secure .06 .08
4. Dismissing .02 .23 .15
5. Preoccupied .11 .11 .29 .22
6. Fearful .10 .05 .53 .13 .08
Recall scores
7. Focal details .06 .08 .02 .02 .04 .12
8. Joint activities .02 .06 .20 .12 .15 .05 .36
9. Separate activities .04 .04 .17 .10 .06 .15 .70 .40
10. Emily’s activities .28 .04 .04 .03 .00 .14 .63 .11 .67
11. Sara’s activities .25 .02 .19 .11 .09 .03 .29 .40 .61 .16
12. Total recall .13 .23 .10 .06 .05 .07 .18 .27 .04 .17 .25
13. Figural analogies .14 .09 .09 .03 .10 .03 .01 .17 .12 .01 .18 .20
Note: Condition was dummy coded for the positive condition and sex was dummy coded for male; rs  .15 have ps  .05, rs  .19 have ps  .01, and rs 
.24 have ps  .001.
Beinstein Miller: Attachment and memory 785

separate activities, was computed by adding together the proportions for


Emily’s/Eric’s and Sara’s/Scott’s separate activities. The mean proportions
caused by memory for focal events, the friends’ joint activities, Emily’s/Eric’s’
separate activities, and Sara’s/Scott’s separate activities were .66 (SD  .05), .24
(SD  .04), .23 (SD  .04), and .18 (SD  .04), respectively. Proportions
caused by memory for focal events, Sara’s/Scott’s separate activities, and the
sum of Emily’s/Eric’s and Sara’s/Scott’s separate activities were each negatively
skewed and corrected by squaring the values. The proportions of recall caused
by memory for the friends’ joint activities were positively skewed and corrected
by taking the natural logarithm of values. As indicated in Table 1, recall of the
friends’ joint activities was negatively correlated with recall of their separate
activities, especially Sara’s/Scott’s. There was also a small, negative correlation
between recall of Sara’s/Scott’s activities and recall of Emily’s/Eric’s.
Scoring of the figural analogies test was conducted by giving 1 point for each
correct solution (0–20 points). The mean score was 14.86 (SD  2.71). Figural
analogies scores were negatively skewed and corrected by squaring the values.
As indicated in Table 1, there were small positive correlations between figural
analogies scores, total recall, and recall of Sara’s/Scott’s activities and a nega-
tive correlation between figural analogies scores and recall of the friends’ joint
activities.

Results

Attachment style and story recall


Multiple regression analyses were used to address the three research questions.
The first two questions — about eliciting attachment-schematic information
processing without priming and about the cognitive mechanism by which
attachment-schematic processing is elicited — were addressed by an analysis of
recall caused by memory for focal story events. The third question — about the
nature of the stimulus that is subject to attachment-related information pro-
cessing biases — was addressed by analyses of the types of focal events recalled.
Hierarchical regression procedures were used in all analyses. In each analysis,
the appropriate recall proportion was regressed on the four attachment ratings,
participants’ reading condition (dummy coded for the positive condition), and
the attachment style condition interactions. The main effects of attachment
style and condition were entered first and their interactions second. In a
third step, all statistically non-significant interactions were removed from the
analysis.
These analyses were conducted for the total sample and then repeated for
the subsample of 130 participants who did not have multiple attachment styles.
The 49 eliminated from the secondary analyses were those who had equally
high ratings above 3 on two or more attachment prototypes. Those with sec-
ondary attachment styles, defined by ratings of 5 on one prototype and 4 on one
or more other prototypes, were retained in the analysis. Also retained were
those who rated no style higher than 3. The secondary analyses were conduc-
ted to determine whether a sample that was purer than the total with regard to
a single attachment style would yield stronger results. Ideally only those par-
ticipants who rated a single attachment style 3 should have been included in
the analysis, but the sample size of participants with a single, highly rated style
was too small to accommodate all of the comparisons to be made in the analy-
786 Journal of Social and Personal Relationships 16(6)

ses, especially in the supplementary analyses that included the effects of parti-
cipant sex. Consequently, the subsample was purer than the total with regard
to a single attachment style but not as pure as it could have been.

Memory for focal events. The analysis of proportions of recall caused by


memory for focal events yielded two attachment style condition interaction
effects for the total sample (See Table 2). A fearful attachment condition
interaction effect indicated a positive association between fearful attachment
ratings and recall of the negative, but not of the positive, story, as can be dem-
onstrated with partial correlations. Correlations between fearful attachment
ratings and recall, while controlling for the effects of participants’ other attach-
ment ratings, were .37, p  .001, in the negative condition and .03, ns, in the
positive condition. A secure attachment condition interaction effect indi-
cated that secure attachment ratings were also associated with better recall of
the negative than of the positive story, as can be shown with partial correla-
tions. Correlations between secure attachment ratings and recall, while
controlling for the effects of participants’ other attachment ratings, were .28,
p  .01, in the negative condition and .14, ns, in the positive condition.
The analysis of the subsample yielded the same results as the total sample.
Correlations between fearful attachment ratings and recall, while controlling for
the effects of participants’ other attachment ratings, were .49, p  .001, in the
negative condition and .03, ns, in the positive condition. Correlations between
secure attachment ratings and recall, while controlling for the effects of partici-
pants’ other attachment ratings, were .35, p  .01, in the negative condition and
.16, ns, in the positive condition. Results for the subsample were thus stronger
than those for the total sample. The total predicted (adjusted) variance
increased from 5 percent in the total sample to 12 percent in the subsample.
These results are partially consistent with predictions based on attachment

TABLE 2
Regression of the proportion of recall caused by memory for focal events

Predictor Total sample Subsample


Standardized Standardized

t p
t p
Condition (positive) 1.09 3.16 .01 1.27 3.40 .001
Attachment style
Secure .38 2.73 .01 .43 3.00 .01
Dismissing .01 0.18 .08 0.92
Preoccupied .05 0.60 .09 1.00
Fearful .52 3.56 .001 .64 4.35 .001
Attachment style condition
Secure condition .74 3.06 .01 .82 3.06 .01
Fearful condition .69 3.03 .01 .76 3.35 .01
Note. The distributions of recall proportions for focal events were negatively skewed and
corrected by squaring the values; total sample positive condition n  93 and negative
condition n  86; subsample positive condition n  64 and negative condition n  66; total
R2 (adjusted) for total sample  .05, F(7, 171)  2.37, p  .05 and for subsample  .12,
F(7,122)  3.44, p  .01.
Beinstein Miller: Attachment and memory 787

style-schematic encoding. Fearful attachment appeared to enhance recall of


focal story events, relative to other story details, in the negative condition.
However, secure attachment also appeared to further the processing of nega-
tive rather than positive focal events, and this result was not predicted by
attachment-schematic encoding. As will be indicated in the analyses that
follow, although secure and fearful attachment styles appeared to have similar
effects on the proportion of recall caused by memory for focal events, the
nature of the events recalled differed by attachment style.

Memory for the friends’ joint activities. Regression of recall proportions


for the friends’ joint activities yielded one interaction effect involving secure
attachment ratings and a main effect of dismissing attachment ratings in the
total sample (See Table 3). The secure attachment condition interaction indi-
cated a positive effect of secure attachment on recall in the negative but not in
the positive story condition, as can be demonstrated by partial correlations.
Correlations between secure attachment ratings and recall, while controlling
for the effects of participants’ other attachment ratings, were .26, p  .05, in the
negative condition and .09, ns, in the positive condition. Dismissing attachment
ratings had a weak but positive main effect on recall. Thus secure attachment
appeared to enhance the processing of the friends’ joint activities, relative to all
other story details, in the negative story condition and dismissing attachment
appeared to have a similar effect, regardless of condition.
Results for the subsample were somewhat different from those for the total
sample. Only one main effect was obtained from the analysis of the subsample,
a positive effect of secure attachment ratings. The secure attachment con-
dition interaction term (standardized
 .36) was consistent with the effect
obtained from the total sample, but it was somewhat smaller and statistically
non-significant. Correlations between secure attachment ratings and recall,

TABLE 3
Regressions of the proportions of recall caused by memory for the friends’
joint activities

Predictor Total sample Subsample


Standardized Standardized

t p
t p
Condition (positive) .39 2.09 .05 .04 0.50
Attachment style
Secure .43 3.55 .001 .33 3.19 .01
Dismissing .16 2.11 .05 .03 0.33
Preoccupied .06 0.76 .03 0.28
Fearful .08 0.95 .10 1.04
Attachment style condition
Secure condition .47 2.35 .05 .— .— —
Note. The distributions of recall proportions for the friends’ joint activities were positively
skewed and corrected by taking their natural logarithms; total sample positive condition n 
93 and negative condition n  86; subsample positive condition n  64 and negative
condition n  66; total R2 (adjusted) for total sample  .07, F(6, 172)  3.12, p  .01 and for
subsample  .06, F(5,124)  2.54, p  .05.
788 Journal of Social and Personal Relationships 16(6)

while controlling for the effects of participants’ other attachment styles, were
.29 (p  .05) in the negative condition and .20 (ns) in the positive condition.
The effect of dismissing attachment, however, was close to zero. The total
(adjusted) variance predicted from this analysis was comparable to that for the
total sample. Results from both analyses are thus consistent with predictions
based on attachment-related attention to different types of activities in the
story, in this case, that secure attachment would be positively correlated with
recall of friends’ joint activities, relative to other story details.

Memory for the friends’ separate activities. In contrast to recall propor-


tions for the friends’ joint activities, those for both their separate activities
yielded an interaction effect involving fearful attachment ratings (See Table 4).
The fearful attachment condition interaction indicated a positive effect of
fearful attachment on recall in the negative story condition but not in the posi-
tive story condition, as can be demonstrated with partial correlations.
Correlations between fearful attachment ratings and recall, while controlling
for the effects of participants’ other attachment ratings, were .34, p  .01, in the
negative condition and .10, ns, in the positive condition. There was also a
trend for dismissing attachment ratings to be negatively associated with recall.
Results for the subsample were partly consistent with those for the total
sample and stronger than those for the total. Correlations between fearful
attachment ratings and recall, while controlling for the effects of participants’
other attachment ratings, were .48, p  .001, in the negative condition and .17,
ns, in the positive condition. An effect for dismissing attachment was not
obtained in this analysis, but an additional interaction effect indicated a negative
association between secure attachment and recall in the positive condition,
which was eliminated in the negative condition. Correlations between secure

TABLE 4
Regressions of the proportions of recall caused by memory for both the
friends’ separate activities

Predictor Total sample Subsample


Standardized Standardized

t p
t p
Condition (positive) .30 1.94 .10 1.21 3.27 .01
Attachment style
Secure .12 1.26 .15 1.02
Dismissing .15 1.91 .10 .08 0.94
Preoccupied .00 0.05 .09 0.98
Fearful .34 2.62 .01 .60 4.07 .001
Attachment style condition
Secure condition .— .— .— .63 2.34 .05
Fearful condition .49 2.61 .01 .88 3.88 .001
Note. The distributions of recall proportions for both friends’ separate activities were
negatively skewed and corrected by squaring their values; total sample positive condition n
 93 and negative condition n  86; subsample positive condition n  64 and negative
condition n  66; total R2 (adjusted) for total sample  .06, F(6,172)  2.92, p  .01 and for
subsample  .13, F(7,122)  3.68, p  .01.
Beinstein Miller: Attachment and memory 789

TABLE 5
Regressions of the proportions of recall caused by memory for Sara’s/Scott’s
separate activities

Predictor Total sample Subsample


Standardized Standardized

t p
t p
Condition (positive) .55 3.62 .001 .67 3.72 .001
Attachment style
Secure .23 2.58 .01 .21 2.10 .05
Dismissing .15 2.03 .05 .11 1.30
Preoccupied .03 0.37 .07 0.75
Fearful .10 0.77 .15 1.18
Attachment style condition
Fearful condition .41 2.26 .05 .51 2.58 .05
Note. The distributions of recall proportions for Sara’s/Scott’s separate activities were
negatively skewed and corrected by squaring the values; total sample positive condition n 
93 and negative condition n  86; subsample positive condition n  64 and negative
condition n  66; total R2 (adjusted) for total sample  .11, F(6,172)  4.85, p  .001 and
for subsample  .12, F(6,123)  4.05, p  .001.

attachment ratings and recall, while controlling for the effects of participants’
other attachment ratings, were .29, p  .05, in the positive condition and .16,
ns, in the negative condition. The total predicted (adjusted) variance increased
from 6 percent in the total sample to 13 percent in the subsample. These effects
are also consistent with predictions based on attachment-related attention to
different types of activities in the story, in this case that fearful attachment would
be positively correlated with recall of the friends’ separate activities, relative to
other story details. The fact that the effect was obtained primarily in the nega-
tive condition provides evidence for attachment-schematic encoding as well.
The regression of recall proportions for Sara’s/Scott’s separate activities
yielded two attachment main effects and one interaction effect in the total
sample. As indicated in Table 5, secure and dismissing attachment ratings were
each negatively associated with recall. A fearful attachment condition inter-
action indicated a negative effect of fearful attachment ratings in the positive
condition that was eliminated in the negative condition. Correlations between
fearful attachment ratings and recall, while controlling for the effects of par-
ticipants’ other attachment ratings, were .25, p  .05, in the positive con-
dition, but .12, ns, in the negative condition.
In the analysis of the subsample, the main effect of secure attachment was
obtained again, but the main effect of dismissing attachment was smaller than
for the total sample and statistically non-significant. This analysis, like that for
the total sample, also yielded a fearful attachment condition interaction.
Correlations between fearful attachment ratings and recall, while controlling
for the effects of participants’ other attachment ratings, were .31, p  .05, in
the positive condition and .22, p  .10, in the negative condition. However, the
two analyses predicted comparable variance in recall proportions for
Sara’s/Scott’s activities.
In the analysis of recall proportions for Emily’s separate activities, there was
790 Journal of Social and Personal Relationships 16(6)

TABLE 6
Regressions of the proportions of recall caused by memory for Emily’s/Eric’s
separate activities

Predictor Total sample Subsample


Standardized Standardized

t p
t p
Condition (positive) .30 4.08 .001 .91 2.00 .05
Attachment style
Secure .05 0.57 .28 1.89 .10
Dismissing .04 0.59 .02 0.29
Preoccupied .02 0.26 .14 1.06
Fearful .20 2.32 .05 .48 3.24 .01
Attachment style condition
Secure condition .— .— .— .55 1.97 .10
Preoccupied x condition .— .— .— .39 1.77 .10
Fearful condition .— .— .— .48 2.10 .05
Note. The distributions of recall proportions for Emily’s/Eric’s separate activities were
negatively skewed and corrected by squaring the values; total sample positive condition n 
93 and negative condition n  86; subsample positive condition n  64 and negative
condition n  66; total R2 (adjusted) for total sample  .08, F(5,173)  4.24, p  .01 and for
subsample  .12, F(8,121)  3.30, p  .01.

only one attachment main effect obtained for the total sample. Fearful attach-
ment was positively associated with recall (See Table 6).
The analysis of recall proportions for Emily’s/Eric’s separate activities
yielded a different result in the subsample. The addition of attachment con-
dition interaction effects for secure, preoccupied, and fearful attachment
together indicated that fearful attachment ratings were positively associated
with recall in the negative condition, but not in the positive condition, and that
there was a similar trend for secure attachment ratings. Correlations between
fearful attachment ratings and recall, while controlling for the effects of par-
ticipants’ other attachment ratings were .39, p  .05, in the negative condition
but .07, ns, in the positive condition. Correlations between secure attachment
ratings and recall, while controlling for the effects of participants’ other attach-
ment ratings were .24, p  .10, in the negative condition, but .11, ns, in the
positive condition. Comparable correlations for preoccupied attachment
ratings were only .14, ns, in the negative condition and .19, ns, in the positive
condition. The total (adjusted) predicted variance was 12 percent, in compari-
son to 8 percent for the full sample. Results for the subsample thus suggest that
fearful attachment was associated with attention to Emily/Eric primarily when
he/she experienced negative outcomes, a result consistent with attachment
schematic encoding as well as with predictions about attachment-related atten-
tion to different types of activities in the story. The similar trend for secure
attachment was not predicted.

Condition effects on story recall. As indicated in Table 1, there were two


main effects of condition on recall. Positive condition participants recalled
Sara’s/Scott’s activities, relative to other details, better than did negative con-
Beinstein Miller: Attachment and memory 791

dition participants. Negative condition participants recalled Emily’s/Eric’s


activities, relative to other details, better than did positive condition partici-
pants. Similar correlations were obtained in the subsample, a correlation of .25,
p  .01, for recall of Sara’s/Scott’s activities and a correlation of .27, p  .01,
for Emily’s/Eric’s activities. No other condition effects were obtained. These
correlations were not changed when the effects of participants’ attachment
ratings were statistically controlled. Participants thus appeared to pay more
attention to Sara/Scott when her/his behavior was a source of well-being for
Emily/Eric than when it posed a threat, and to pay more attention to
Emily/Eric when her/his well-being was threatened than when it was enhanced
by her/his friend.

Attachment style and figural analogies scores


A main effect of preoccupied attachment ratings was obtained in the regression
of figural analogies scores for the total sample (See Table 7). The more preoc-
cupied the participant, the lower his/her score. A similar trend was obtained for
secure attachment. However, the final regression equation did not predict stat-
istically significant variance in test scores.
A different effect was obtained for the subsample. The addition of attach-
ment condition interaction effects for dismissing, preoccupied, and fearful
attachment together increased the predictive power of the overall analysis. The
total (adjusted) predicted variance was 8 percent, in comparison to 2 percent
for the full sample. Together, the interaction effects indicated that test scores
were negatively correlated with dismissing attachment ratings in the negative
condition and with fearful attachment ratings in the positive condition. Test
scores were also negatively correlated with preoccupied attachment ratings in
the positive condition but the condition difference in effect was of marginal sta-
tistical significance. Correlations between dismissing attachment ratings and
test scores, while controlling for the effects of participants’ other attachment

TABLE 7
Regression of figural analogies scores

Predictor Total sample Subsample


Standardized Standardized

t p
t p
Condition (positive) .15 2.02 .05 .28 0.84
Attachment style
Secure .16 1.75 .10 .11 1.09
Dismissing .02 0.28 .23 1.98 .05
Preoccupied .17 2.04 .05 .01 0.06
Fearful .05 0.59 .11 0.82
Attachment style condition
Dismissing condition .— .— .— .62 2.46 .05
Preoccupied condition .— .— .— .37 1.69 .10
Fearful condition .— .— .— .50 2.45 .05
Note. The distribution of analogies scores was negatively skewed and corrected by squaring
the values; total sample positive condition n  93 and negative condition n  86; subsample
positive condition n  64 and negative condition n  66; total R2 (adjusted) for total sample
 .02, F(5,173)  1.83, ns and for subsample  .08, F(8,121)  2.46, p  .05.
792 Journal of Social and Personal Relationships 16(6)

ratings, were .26, p  .05, in the negative condition and .20, ns, in the positive
condition. Correlations between fearful attachment ratings and test scores,
while controlling for the effects of participants’ other attachment ratings, were
.29, p  .05, in the positive condition and .05, ns, in the negative condition.
Comparable correlations for preoccupied attachment ratings were .27, p 
.05, in the positive condition and .01, ns, in the negative condition. Results for
the subsample thus indicated that the positive story was distracting for partici-
pants with preoccupied and fearful attachment styles because it appeared to
interfere with their subsequent problem-solving. In contrast, it was the negative
story that appeared to distract participants with dismissing attachment styles
because their subsequent problem-solving suffered in the negative condition.

Supplementary analyses
Because women often report greater intimacy in friendships than men do
(Canary, Emmers-Sommer, & Faulkner, 1997), and might therefore assimilate
related information better than men, the effects of participants’ sex were also
examined. Although females had slightly better total recall for story events
than males did in the total sample, they did not differ from males on any of the
recall proportions (See Table 1). In the subsample, females also had better total
recall than males did (r  .23, p  .01), but but did not differ on any of the
recall proportions.
Participants’ sex, condition sex, attachment style sex, and attachment
style condition sex interactions were also added to the main analyses of
recall proportions. Participant sex, dummy coded for male, was entered in the
first step with all other main effects, all two-way interactions were entered in
the second step, and all three-way interactions were entered in the third step.
In subsequent steps, all statistically non-significant three-way interactions were
removed, followed by all statistically non-significant two-way interactions.
These analyses did not change the results reported earlier and yielded only one
statistically significant attachment style sex interaction for the total sample.
In the analysis of memory for Sara’s/Scott’s activities, a negative regression
coefficient was obtained for the preoccupied attachment sex interaction
term, which provided the added effect of preoccupied ratings among males
(standardized
 .40, p  .05). A positive, but statistically non-significant,
coefficient was obtained for the effect of preoccupied ratings among females
(standardized
 .11). Correlations between preoccupied attachment ratings
and recall of Sara’s/Scott’s activities, while controlling for the effects of partici-
pants’ other attachment styles, condition, and the fearful attachment con-
dition interaction, confirmed the difference in effect. The correlation was .27,
p  .05, for males and .11, ns, for females. In the subsample, however, this inter-
action effect was not statistically significant (standardized
 .13).
Correlations between preoccupied attachment ratings and recall, while con-
trolling for the effects of participants’ other attachment ratings and condition,
were .15, ns, for males and .02, ns, for females.
To determine whether attachment-related information processing biases
extended to the entire story, total recall scores were regressed on condition,
attachment ratings, participant sex, and their interactions as described earlier.
Only two effects were obtained in the analysis of the total sample. As indicated
previously, males’ total recall scores were significantly lower than females’
(standardized
 .22, p  .01). Also, a preoccupied attachment condition
interaction effect (standardized
 .40, p  .05) indicated a negative effect
Beinstein Miller: Attachment and memory 793

of preoccupied attachment ratings in the positive, but not in the negative, con-
dition, as can be demonstrated with partial correlations. The correlations
between preoccupied attachment ratings and total recall, while controlling for
the effects of participants’ other attachment ratings, were .25, p  .05, in the
positive condition but .06, ns, in the negative condition.
In the analysis of the subsample, the preoccupied attachment condition
interaction effect was not statistically significant (standardized
 .26, ns).
Nevertheless, correlations between preoccupied attachment ratings and total
recall scores, while controlling for the effects of participants’ other attachment
ratings, continued to show a negative trend in the positive condition.
Correlations were .23, p  .10, in the positive condition and .00, ns, in the
negative condition. In addition, the analysis for the subsample yielded a main
effect of secure attachment. The more secure the participant, the worse was
his/her recall.

Discussion

Results from this study provide intriguing answers to the three research
questions that guided the investigation. The first question concerned the
possibility of obtaining attachment-schematic processing of a friendship
story when attachment-related memories were not deliberately activated.
Consistent with an attachment-schematic encoding hypothesis, fearful
attachment appeared to enhance the processing of focal events, relative to
all other story details, when the story was negative. This information pro-
cessing advantage occurred primarily because of the processing of infor-
mation about the friends’ separate activities, especially those of the main
character. As will be indicated below, the apparent tendency for fearful
attachment to facilitate the processing of information about the friends’
separate activities can be interpreted as one form of attachment-schematic
information processing.
Secure attachment also appeared to enhance the processing of focal
events, relative to all other story details, when the story was negative, but
this result is inconsistent with an attachment-schematic encoding hypothe-
sis. According to an attachment-schematic encoding hypothesis, secure
attachment should have been positively related to recall of focal events
from the positive rather than from the negative story. One reason for this
inconsistency could be the failure of the positive story to activate partici-
pants’ attachment beliefs uniformly. According to attachment theory, the
attachment system is activated primarily by threatening circumstances, and
the attachment beliefs of individuals with fewer attachment concerns are
thought to be more difficult to activate than those of individuals with more
numerous and frequently troublesome concerns (Simpson & Rholes, 1994).
Securely attached individuals are by definition those who have fewer unre-
solved attachment concerns. If the negative rather than the positive story
activated their attachment beliefs then their attention could have been
attracted to one or another types of negative events only. The finding that
the positive association between secure attachment and recall of focal
events from the negative story occurred because of superior recall of the
794 Journal of Social and Personal Relationships 16(6)

friends’ joint activities rather than their separate activities might also be
interpreted as one form of attachment-schematic information, as will be
discussed shortly.
Consistent with Miller and Noirot’s (1999) research, dismissing and pre-
occupied attachment styles were not associated with recall of focal details
from either the positive or negative story. Dismissing attachment appeared
to enhance recall of the friends’ joint activities and to limit recall of the
partners’ separate activities in the total sample, but this effect was not
obtained in the subsample of participants who did not have multiple attach-
ment styles and could therefore have resulted from competing attachment
expectations. Preoccupied attachment was not associated with enhanced
recall of any focal events but, instead, appeared to interfere generally with
the processing of the positive story. One reason for this result might be the
desires of preoccupied individuals for excessive closeness. These desires
could have been activated by the positive story and competed for infor-
mation-processing resources. The absence of a negative correlation in the
negative story condition could reflect greater attention to the negative
story, which compensated for distraction. The fact that preoccupied attach-
ment had a negative effect on figural analogies scores, and the fact that this
effect occurred primarily in the positive condition among participants
without multiple attachment styles, is consistent with the possibility that
attachment concerns could have used up information-processing resources
of preoccupied participants who read the positive story. This explanation is
also consistent with previous findings that unsuppressed anxious attach-
ment, when activated by prior relationship concerns, can interfere with the
processing of both attachment-related and attachment-unrelated infor-
mation (Miller, 1996; Miller & Noirot, 1999). It is consistent as well with the
negative effect of fearful attachment on figural analogies scores in the posi-
tive condition among participants without multiple attachment styles.
Dismissing attachment also had a negative effect on figural analogies
scores among participants without multiple attachment styles, but this
effect occurred in the negative rather than in the positive condition and is
inconsistent with previous research results. Miller and Noirot (1999)
reported a positive association between dismissing attachment and figural
analogies scores among participants whose negative attachment-related
memories had been activated. They attributed this effect to dismissing indi-
viduals’ superior ability to suppress negative relationship experiences, pro-
viding them with more resources available for problem-solving. Results
from this study are inconsistent with that idea unless the ability of dismiss-
ing individuals to suppress negative relationship information applies only
to their own experiences. It is also possible that the negative story did not
activate their attachment-related beliefs. If individuals with dismissing
attachment styles have superior ability to suppress attachment-related
experiences, then it would be particularly difficult to activate their attach-
ment systems. The negative story might have instead been distasteful and
reduced their interest in subsequent activities associated with the study
(i.e., the figural analogies test). Results for preoccupied and dismissing
Beinstein Miller: Attachment and memory 795

attachment might therefore be better explained by participants’ allocation


of information-processing resources than by their attachment-schematic
processing of the story.
The second question concerned the cognitive mechanism by which
attachment style influences information processing. Would individuals
process attachment style-congruent information better than attachment
style-incongruent information because they paid greater attention to it or
because they assimilated it better? It was argued that if information-pro-
cessing biases could be demonstrated with purely positive or negative
events, then this result would implicate an explanation based on assimila-
tion processes. Results from this study provide partial support for the role
of assimilation processes. Fearful attachment appeared to facilitate the
assimilation of negative focal events, but secure attachment did not facili-
tate the assimilation of positive ones. As previously noted, the positive
story might not have activated attachment concerns uniformly.
The third question concerned the nature of the stimulus event that was
subject to attachment style-related information-processing effects. Would
information-processing effects be caused by the relational nature of story
events or by the valence of their outcomes? It was argued that attachment-
schematic information processing might entail more than assimilation of
positive or negative outcomes that are consistent with attachment expecta-
tions. If attachment-related models of self and others have directive influ-
ences on attention, then attachment-schematic information processing
might also entail characteristic differences in attention to self and partners,
and these differences might be reproduced in reading. Specifically, it was
suggested that positive models might direct attention to sources of well-
being and negative models to sources of threat. Threat would heighten anx-
iety, and therefore attention, unless it could be suppressed. Consequently,
fearful attachment, which is characterized by negative models of both self
and other, was predicted to enhance attention to both the main character
of the story and his/her partner. Secure attachment, which is characterized
by positive models of both, was predicted to involve lesser attention to
both. The conflicting motives entailed in dismissing and preoccupied
attachment styles — to seek sources of well-being and monitor threat —
were predicted to be resolved by the seeking of well-being and avoidance
of threat. Dismissing attachment was predicted to enhance attention to the
main character or, because of anxiety suppression, to simply reduce atten-
tion to the partner. Preoccupied attachment was predicted to enhance
attention to the partner. Finally, securely attached readers, and possibly
dismissing readers, were predicted to have more resources available than
other readers for processing information about the characters’ joint activi-
ties. Their recall of these activities was therefore predicted to be better than
that of other readers.
Results from this study provide support for a majority of these predic-
tions. In both the total sample and subsample, fearful attachment appeared
to enhance recall of both the friends’ separate activities, relative to other
story details, although this effect occurred only in the negative condition
796 Journal of Social and Personal Relationships 16(6)

and was primarily caused by recall of the main character, Emily/Eric. In


both analyses, fearful attachment appeared to reduce recall of the partner,
Sara/Scott, when she/he posed no threat (i.e., in the positive condition).
In the total sample, fearful attachment appeared to have a general, posi-
tive effect on recall of Emily’s/Eric’s activities, relative to other details, but
in the subsample this effect occurred in the negative condition only. Also,
in the subsample, there was a trend for fearful attachment to enhance recall
of Sara’s/Scott’s activities in the negative condition. Results for fearful
attachment, particularly from the subsample, therefore provide evidence
that both the relational nature of events and their outcomes made a differ-
ence in information processing.
Results for secure and dismissing attachment also provided evidence for
attentional differences that were predicted to stem attachment-related
models of self and other. Secure and dismissing attachment appeared to
facilitate recall of the friends’ joint activities, relative to other story details,
although the effect for dismissing attachment was eliminated in the sub-
sample. Secure and dismissing attachment appeared to reduce recall of
Sara’s/Scott’s activities, although the effect for dismissing attachment was
reduced in the subsample. Secure attachment appeared more generally to
reduce attention to both the friends’ separate activities in the positive con-
dition, but this result was obtained only in the subsample. These results are
consistent with predictions that secure attachment would reduce attention
to both partners and self and free resources for processing other infor-
mation. However, secure attachment appeared to enhance recall of
Emily’s/Eric’s activities in the negative condition among subsample partici-
pants. Although the correlation was only marginally significant, it is incon-
sistent with predictions that secure attachment would encourage only
moderate attention to partners and self. One explanation for the result is
that threats to Emily/Eric in the negative condition heightened concerns
for her/his well-being and therefore attention to her/his activities. This
explanation is consistent with the general condition differences in recall of
the story characters’ separate activities. Whereas Sara’s/Scott’s activities
were recalled better by participants in the positive than in the negative con-
dition, Emily’s/Eric’s activities were recalled better by participants in the
negative than in the positive condition.
Other predictions about attachment-related patterns of attention were
not validated by differences in recall. Preoccupied attachment was pre-
dicted to correlate positively with recall of Sara’s/Scott’s activities, relative
to other details, but in the total sample the correlation for females was stat-
istically non-significant and the correlation for males was negative. The
result for males could reflect the contradictory nature of playing a male role
and having a preoccupied attachment style. If independence and strength
are internalized as part of the male role, then preoccupied attachment
could challenge that role. One solution would be to reduce attention to the
partner. However, the effect was not obtained in the subsample and so
could reflect conflicting attachment expectations that are associated with
having multiple attachment styles.
Beinstein Miller: Attachment and memory 797

There was also no evidence that dismissing attachment would enhance


recall of Emily’s/Eric’s activities, relative to other details. However, pre-
dictions about dismissing attachment were qualified by the possibility that
dismissing readers, adept at anxiety suppression, might simply pay less
attention to the partner. The negative correlation between dismissing
attachment and recall of Sara’s/Scott’s activities, relative to other details,
provides evidence for this possibility, but the correlation was obtained in
the total sample only. If partner avoidance is an attentional strategy associ-
ated with dismissing attachment, then the correlation should have been
stronger, not weaker, in the subsample than in the total sample. The corre-
lation in the total sample could thus reflect conflicts in attachment expecta-
tions that are associated with having multiple attachment styles. As Fraley,
Davis, and Shaver (1998) have noted, very little research has been pub-
lished about the psychological dynamics of dismissing adults. Although
there is some evidence for their defensive attention to and organization of
attachment-related information in memory, there is little known about the
conditions under which their defenses do and do not work. This is a ques-
tion for future research.
Although not all predicted correlations were obtained in this study, those
that were obtained, especially for secure and fearful attachment, provide
evidence that attachment-schematic information processing might entail
more complex considerations than the valence of stimulus information
alone. Attachment beliefs also appear to structure attention differentially
to representations of the self, other, and relationships, particularly in
threatening situations. For example, attachment style correlates of
responses to conflict — an event with potentially negative outcomes —
could occur partially because of attachment-related concerns for self, other,
and relationships. The positive correlations reported between secure
attachment and a constructive approach to conflict (Pistole, 1989; Scharfe
& Bartholomew, 1995) could occur because securely attached individuals
are not overwhelmed by concerns for self and partner and are able to direct
their information processing resources towards joint concerns in the
relationship. Future research could examine the attachment style correlates
of memory for conflict episodes to investigate this possibility.
Results from this study also indicate that attachment-schematic infor-
mation-processing effects can be obtained without deliberately activating
attachment-related memories but that these effects are small when acti-
vated by a friendship story, even one with negative outcomes. Although the
variance predicted by attachment ratings tended to be greater in the sub-
sample than in the total sample, in neither was more than 13 percent of the
variance in recall predicted. These results are a reminder of the many other
factors that can influence information processing, among them the reader’s
verbal abilities, prior related experiences, and current involvements.
Especially noteworthy are potential conflicts in attachment expectations
that can accrue to individuals who have multiple and secondary attachment
styles. Conflicting attachment expectations can dilute attachment effects, as
indicated by results from this study. Future research could investigate the
798 Journal of Social and Personal Relationships 16(6)

dynamics of multiple attachment styles to determine what it means to have


conflicting expectations and how these affect attachment-related behavior.
The fact that attachment style effects in this study were relatively small is
not surprising, but neither is it trivial. Even small biasing effects, when
repeatedly experienced, have the potential to reinforce attachment beliefs.
Repeated exposure to relationship events can thereby perpetuate attach-
ment beliefs and their influence on behavior in close relationships.

Appendix

The friendship story: Negative version


Emily stood in the airport terminal waiting for Sara. It was a cold and
snowy December morning. Sara asked Emily to pick her up because both
of her parents were working. It had been nearly four months since Emily
had last seen Sara. During that time, Emily had found a full-time job. She
was unhappy with her work and jealous that Sara was enjoying college. She
and Sara hadn’t spoken to each other much and had only e-mailed a few
times. This made Emily even more anxious to see Sara again. She thought
about their very long friendship and wondered if it would change now that
they were older.
Since it was snowing hard outside, some flights were delayed. Emily was
not surprised to see that Sara’s was one of them but did not like the idea of
waiting. The idea of seeing Sara made her think of the time that she and
Sara did a science-fair project together. They were both in Mrs Miller’s
class in the third grade. A fifth grade teacher arranged for an all school sci-
ence fair. Sara and Emily spent a lot of time together, arguing over what
they should do for the fair. Emily’s mom suggested they should design a
project about Morse code. They worked at Emily’s house every night for
two weeks and had a miserable time. While they worked on the project
they fought over who would present it at the fair because Sara wanted to
do the presentation. They created their project from a piece of plywood
with a light hooked up to it. Emily’s dad helped them do the wiring. There
was also a report and chart about Morse code. Emily was happy that her
parents helped them out so much. At the fair, Emily sent obscene messages
to Sara while she was talking. No one else understood what Emily was
saying but Sara was not amused. The girls hated doing their project
together. Sara’s parents both started working full-time that year, so Sara
went home with Emily every day after school. As a result, the two spent a
lot of time together and became irritatingly close.
Emily’s thoughts drifted to the time in fifth grade when she and Sara
helped to organize the End of the Year Activities Carnival. Sara’s mom
thought that it would be a good way for the girls to get involved with their
school but Emily wondered how they would share responsibilities. Sara
usually took charge in these situations. As it turned out, the girls did
not work together. Sara was put in charge of the Physical Education
activities and Emily was put in charge of the Language Arts activities.
Beinstein Miller: Attachment and memory 799

Emily was relieved to have her own project because the third grade
science fair project had been a lot of stress. Both of the girls worked very
hard on their activities and looked forward to the carnival. Everyone who
attended the carnival enjoyed the activities that the girls organized. Emily
had set up activities like storytelling, poetry reading, and a writing work-
shop. Sara had set up a basketball and a softball game. At the carnival,
Sara was disappointed with her own project because Emily’s had been so
creative.
After elementary school, the relationship began to falter in trust because
Sara was spending so much time with the soccer team. Emily was not on the
team because she had joined the Academic Challenge team. They only saw
each other at Mr Chicken, where they both worked, and argued whenever
they were together. In seventh grade, Sara became close with Amy, a girl
from the team. It wasn’t long before Sara introduced the two. Emily
remembers that weekend in particular. She and Sara were planning to see
‘Back to the Future’ at the Cinema 8. Sara called Emily in the afternoon to
ask if Amy could join them. Emily was disappointed, but did not tell Sara.
Emily said that she would meet them on the corner, outside the theater.
When she arrived at the movies, there were Amy and Sara, already in line.
Emily was late but was hurt that they had not waited. She ran up to join
them in line. She met Amy and found her to be outgoing and very smart.
She was very jealous of Sara.
Emily looked down at her watch and realized that the plane was already
fifteen minutes late. She tried to remain calm and collected but couldn’t.
She was really nervous at the thought of seeing Sara again. She looked
around at the other gates to distract herself. She noticed a group of pas-
sengers arriving from Orlando. There was a sunburned woman with two
little girls, a business man carrying a brown briefcase, and a family with five
kids coming off the plane. This made her think of the road trip that she and
Sara took on spring break of their senior year of high school. Sara and
Emily both wanted to take a trip to Florida, but could not afford the air-
fare. Emily’s parents were not very pleased with the thought of Sara and
Emily ‘experiencing life’ in Daytona for a week, but they agreed the girls
could go if Emily could earn the money herself. Emily and Sara spent a lot
more time working at Mr Chicken and less time with their extra-curricular
activities in order to raise the money. They had a terrible time because of
the fact that it was work. They raised the money and Sara’s parents lent
them their minivan to drive to Florida. They packed their clothes and beach
stuff, CDs and tapes, and junk food and sodas into the van and departed
early on a Saturday morning. It was an 18-hour drive, so they split it up into
two days. They stopped every couple of hours to snack and to take a break
from each other; the girls never realized how stressful it would be to drive
so far together. Sara made Emily listen to The Beatles over and over until
they both could sing every lyric to every song. On the road Sara and Emily
talked a lot. They argued about topics like abortion, religion, and the
future. It made them feel distant to understand each other’s different ideas
about these topics.
800 Journal of Social and Personal Relationships 16(6)

Emily told Sara she wasn’t sure exactly what she wanted to do with her
life yet, but was thinking about something in music or writing. She told Sara
that she planned to stay home for a year to earn some money and to decide
what college she should go to. Sara said that she thought Emily’s plan was
pragmatic because college is so expensive and those majors could be very
useful someday. Emily felt like Sara was talking down to her. Sara then told
Emily that she really wanted to be a doctor and that the University of
Michigan was definitely the place for her next year. Emily felt inadequate
to know that Sara had such a good idea about her future.
On September 2, Sara and Emily spent their last time together before
Sara left for college. They hung out at Sara’s house and argued about old
times. Emily was sad, because she knew that she and Sara would not see
each other for a while. She was envious because she knew Sara would meet
a lot of new interesting people at college. Emily was also unhappy to know
that Sara would have very different experiences at college than Emily
would have at home. This made her regret her decision to stay home
because they would not have a lot to talk about. Sara did not look forward
to all the new experiences she would have. Before Emily went home, she
and Sara agreed to stay in touch by e-mail. Emily didn’t think that they
would.
Emily heard a flight being announced over the intercom. It was Sara’s.
By now, her flight was forty minutes late. Emily knew that any minute she
would be seeing Sara again and was very apprehensive. Emily looked into
the crowd of arriving passengers. So many were smiling and looking around
eagerly, probably for friends or family who were supposed to meet them,
but there was Sara, with a frown, walking towards her lugging two big,
green bags. Emily thought to herself, ‘not much has changed’.

REFERENCES

Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment: A
psychological study of the strange situation. Hillsdale, NJ: Erlbaum.
Baldwin, M. W., Fehr, B., Keedian, E., Seidel, M., & Thomson, D. W. (1993). An exploration
of the relational schemata underlying attachment styles: Self-report and lexical decision
approaches. Personality and Social Psychology Bulletin, 19, 746–754.
Baldwin, M. W., Keelan, J. P. R., Fehr, B., Enns, V., & Koh-Rangarajoo, E. (1996). Social-cog-
nitive conceptualization of attachment working models: Availability and accessibility
effects. Journal of Personality and Social Psychology, 71, 94–109.
Bartholomew, K. (1990). Avoidance of intimacy: An attachment perspective. Journal of Social
and Personal Relationships, 7, 147–178.
Bartholomew K., & Horowitz, L. M. (1991). Attachment styles among young adults: A test of
a four-category model. Journal of Personality and Social Psychology, 61, 226–244.
Belsky, J., Spritz, B., & Crnic, K. (1996). Infant attachment security and affective-cognitive
information processing at age 3. Psychological Science, 7, 111–114.
Bowlby, J. (1969). Attachment and loss, Vol. 1: Attachment. New York: Basic Books.
Bowlby, J. (1973). Attachment and loss, Vol. 2: Separation: Anxiety and anger. New York:
Basic Books.
Bowlby, J. (1980). Attachment and loss, Vol. 3: Loss: Sadness and depression. New York: Basic
Books.
Beinstein Miller: Attachment and memory 801

Brennan, K. A., Clark, C. L., & Shaver, P. R. (1998). Self-report measurement of adult attach-
ment: An integrative overview. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory
and close relationships (pp. 46–76). New York: Guilford Press.
Bridges, L. J., Connell, J. P., & Belsky, J. (1988). Similarities and differences in infant–mother
and infant–father interaction in the strange situation: A component process analysis.
Developmental Psychology, 24, 92–100.
Canary, D. J., Emmers-Sommer, T. M., & Faulkner, S. (1997). Sex and gender differences in
personal relationships. New York: Guilford Press.
Collins, N. L. (1996). Working models of attachment: Implications for explanation, emotion,
and behavior. Journal of Personality and Social Psychology, 71, 810–832.
Collins, N. L., & Read, S. J. (1994). Cognitive representations of attachment: The structure
and function of working models. In K. Bartholomew & D. Perlman (Eds.), Advances in per-
sonal relationships (Vol. 5, pp. 53–90). London: Jessica Kingsley.
Fox, N. W., Kimmerly, N. L., & Schafer, W. D. (1991). Attachment to mother/attachment to
father: A meta-analysis. Child Development, 62, 210–225.
Fraley, R. C., Davis, K. E., & Shaver, P. R. (1998). Dismissing-avoidance and the defensive
organization of emotion, cognition, and behavior. In J. A. Simpson & W.S. Rholes (Eds.),
Attachment theory and close relationships (pp. 249–279). New York: Guilford Press.
Fraley, R. C., & Shaver, P. R. (1997). Adult attachment and the suppression of unwanted
thoughts. Journal of Personality and Social Psychology, 73, 1080–1091.
Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an attachment process.
Journal of Personality and Social Psychology, 52, 511–524.
Hazan, C., & Zeifman, D. (1994). Sex and the psychological tether. In K. Bartholomew & D.
Perlman (Eds.), Advances in personal relationships (Vol. 5, pp. 151–178). London: Jessica
Kingsley.
Mikulincer, M. (1998). Attachment working models and the sense of trust: An exploration of
interaction goals and affect regulation. Journal of Personality and Social Psychology, 74,
1209–1224.
Miller, J. B. (1996). Social flexibility and anxious attachment. Personal Relationships, 3,
241–256.
Miller, J. B., & Noirot, M. (1999). Attachment memories, models, and information processing.
Journal of Social and Personal Relationships, 16, 147–173.
Pistole, M. C. (1989). Attachment in adult romantic relationships: Styles of conflict resolution
and relationship satisfaction. Journal of Social and Personal Relationships, 6, 505–510.
Scharfe, E., & Bartholomew, K. (1995). Accommodation and attachment representations in
young couples. Journal of Social and Personal Relationships, 12, 389–401.
Shaver, P. R., Collins, N., & Clark, C. L. (1996). Attachment styles and internal working
models of self and relationship partners. In G. J. O. Fletcher & J. Fitness (Eds.), Knowledge
structures in close relationships: A social psychological approach (pp. 25–61). Mahwah, NJ:
Erlbaum.
Shaver, P., & Hazan, C. (1990). Love and work: An attachment-theoretical perspective.
Journal of Personality and Social Psychology, 59, 270–280.
Simpson, J. A., & Rholes, W. S. (1994). Stress and secure base relationships in adulthood. In
K. Bartholomew & D. Perlman (Eds.), Advances in personal relationships (Vol. 5, pp.
181–204). London: Jessica Kingsley.
Thorndike, R. L., & Hagen, E. (1971). Cognitive abilities test: Form I. Boston: Houghton
Mifflin.
Thorndike, R. L., & Hagen, E. (1974). Cognitive abilities test: Form III. Boston: Houghton
Mifflin.
Trinke, S. J., & Bartholomew, K. (1997). Hierarchies of attachment relationships in young
adulthood. Journal of Social and Personal Relationships, 14, 603–625.
Young, A. M., & Acitelli, L. K. (1998). The role of attachment style and relationship status of
the perceiver in perceptions of romantic partner. Journal of Social and Personal
Relationships, 15, 161–173.
CALL FOR PAPERS
on Narratives In and About Relationships
A special issue of the Journal of Social and Personal Relationships
will be dedicated to research using narrative methodologies to study
interpersonal relationships. The special issue will highlight recent
methodological advances in the use of narratives to understand how
individuals find relevance in dyadic and group relationships (includ-
ing family relationships), which in turn may be related to relational
well-being. Authors must clearly address the theoretical rationale for
using narratives to study relationships as well as provide substantial
documentation of their methods. Preference will be given to em-
pirical studies, although theoretical papers may be considered.
Manuscripts will be peer-reviewed, consistent with policies of the
journal. Questions regarding the issue may be directed to
Barbara Fiese, bhfiese@psych.syr.edu or Harold Grotevant,
hgrotevant@che2.che.umn.edu. Submissions are due no later than
February 1, 2000. Six copies, prepared in APA style, should be sent
to Barbara H. Fiese, PhD, 430 Huntington Hall, Department of
Psychology, Syracuse University, Syracuse, NY 13244–2340, USA.

Vous aimerez peut-être aussi