Sunteți pe pagina 1din 14

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/232577676

Support Seeking and Support Giving Within


Couples in an Anxiety-Provoking Situation:
The Role of Attachment Styles

Article in Journal of Personality and Social Psychology · March 1992


DOI: 10.1037/0022-3514.62.3.434

CITATIONS READS

811 4,888

3 authors, including:

Jeffry A Simpson William S Rholes


University of Minnesota Twin Cities Texas A&M University
216 PUBLICATIONS 15,048 CITATIONS 82 PUBLICATIONS 6,709 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Current Opinion in Psychology View project

Early unpredictability, attachment, emotion regulation, and reproductive behavior View project

All content following this page was uploaded by Jeffry A Simpson on 12 February 2015.

The user has requested enhancement of the downloaded file.


Journal of Personality and Social Psychology Copyright 1992 by the American Psychological Association, Inc.
1992, Vol. 62. No. 3, 434-446 0022-3514/92/$3.00

Support Seeking and Support Giving Within Couples in an Anxiety-


Provoking Situation: The Role of Attachment Styles
Jeffry A. Simpson, William S. Rholes, and Julia S. Nelligan
Texas A&M University

This study examined how adult attachment styles moderate spontaneous behavior between dating
couples when 1 member of the dyad is confronted with an anxiety-provoking situation. Eighty-three
dating couples were unobtrusively videotaped for 5 min in a waiting room while the woman waited
to participate in an "activity" known to provoke anxiety in most people. Independent observers
then evaluated each partner's behavior on several dimensions. Results revealed that persons with
more secure attachment styles behaved differently than persons with more avoidant styles in terms
of physical contact, supportive comments, and efforts to seek and give emotional support. Findings
are discussed in the context of theory and research on attachment.

Recently, a growing number of researchers have begun to of attachment behavior. Children involved in secure relation-
explore how different attachment styles influence what tran- ships use their caregivers as a base of security to regulate feel-
spires within adult relationships (e.g., Collins & Read, 1990; ings of distress and anxiety that arise in the Strange Situation.
Feeney & Noller, 1990; Hazan & Shaver, 1987; Levy & Davis, For these children, the caregiver's presence provides a clear
1988; Simpson, 1990). Virtually all of this research has been source of comfort and support. Children involved in avoidant
guided by Bowlby's (1969,1973,1980) attachment theory. relationships neither actively seek support from the caregiver
Bowlby sought to understand why infants forge strong emo- nor use the caregiver to regulate and dissipate negative affect
tional bonds to their primary caregivers and why they often when it arises. Children involved in anxious/ambivalent rela-
exhibit pronounced anxiety and distress when they are sepa- tionships make inconsistent, conflicted, and ambivalent at-
rated from them. Adopting an evolutionary-ethological per- tempts to derive emotional support from their caregivers, ac-
spective, he argued that the specific sequence of behavioral and tions that seem to reflect an underlying sense of uncertainty
emotional reactions associated with separation—protest, de- about the caregiver's availability and supportiveness.
spair, and detachment—might reflect the operation of an in- According to attachment theory, the relationships an individ-
nate attachment system designed to promote close physical ual has during infancy, childhood, and adolescence give rise to
contact between vulnerable infants and their primary care- "mental models" of both self and others that influence patterns
givers. By maintaining close proximity with their caregivers, of support-proximity seeking and support giving in adult rela-
infants would be more likely to survive, to reproduce, and ulti- tionships. Although not impervious to change (Main, Kaplan,
mately to pass attachment and proximity-seeking propensities
& Cassidy, 1985), these mental models are thought to become
on to subsequent generations.
stable and traitlike over time (Bretherton, 1985). Early relation-
Although the tendency to seek proximity seems to be univer- ships are thus presumed to exert long-term impact on subse-
sal in infants, its development and elaboration over time is sen- quent relationships by affecting the nature and development of
sitive to specific environmental conditions, especially transac- these mental models.
tions between the infant and his or her primary caregiver. Past
empirical research has examined characteristic patterns of at- Research on the continuing impact of early relationships on
tachment behavior that develop between infants and their care- later ones is still preliminary. Hence, strong conclusions about
givers that signify different types of infant-caregiver relation- long-term effects are premature at this time. Nevertheless,
ships. Using the Strange Situation paradigm, Ainsworth, Ble- some encouragingfindingshave been reported recently. Elicker
har, Waters, and Wall (1978) identified three primary patterns and Sroufe (in press, cited in Sroufe, Egeland, & Kreutzer,
1990), for instance, have found that assessments of attachment
style conducted during infancy reliably predict social skills and
Portions of this research were presented at the 99th Annual Conven- self-confidence in children 10 years later, even when more con-
tion of the American Psychological Association, San Francisco, Cali- temporaneous measures of adjustment and home environment
fornia, August 1991. are controlled for. Infant attachment also reliably predicts the
Jeffry A. Simpson and William S. Rholes contributed equally to this incidence of behavior problems, difficulties in adjusting to
article. We thank Angela Floyd, Karen Galloway, Tiffany Keith, Traci school, the quality of relationships with peers and adults other
Littleton, Lucretia Marmor, Peggy Philpot, Lucille Reiter, Linda
Wagner, Katy Watts, Laura Webb, and Laura Woodruff for their assis- than parents, and the quality of family relationships during
tance with this research. toddler, preschool, and elementary school years (see Erickson,
Correspondence concerning this article should be addressed to Sroufe, & Egeland, 1985; Lewis, Feiring, McGuffog, & Jaskir,
Jeffry A. Simpson or William S. Rholes, Department of Psychology, 1984; Londerville& Main, 1981; Main etal.,1985;Sroufe, 1983;
Texas A&M University, College Station, Texas 77843-4235. Sroufe et al., 1990; Waters, Wippman, & Sroufe, 1979).
434
ATTACHMENT STYLES 435

Several recent studies have suggested that three attachment (Berscheid, Snyder, & Omoto, 1989), should be most clearly
styles phenotypically similar to those discovered by Ainsworth evident. Indeed, the paradigm initially used to identify and
et al. (1978) also may characterize adults (see Ainsworth, 1985; study attachment styles in children—Ainsworth et al.'s (1978)
Collins & Read, 1990; Crowell & Feldman, 1988; Feeney & Strange Situation—was designed explicitly to induce anxiety in
Noller, 1990; Hazan & Shaver, 1987; Kobak & Sceery, 1988; infants so that (a) the attachment system would be strongly acti-
Main & Goldwyn, 1984; Main et al., 1985). People who possess vated and (b) behaviors associated with different styles would
a secure attachment style tend to develop mental models of be directly elicited. To date, the role that distress and anxiety
themselves as being valued and worthy of others' concern, sup- assume in eliciting proximity-seeking behaviors in adults has
port, and affection, and of significant others as being accessi- not been examined. Research on this issue is critical to the
ble, reliable, trustworthy, and well-intentioned. Secure individ- extension of attachment theory from childhood to adulthood.
uals report that they develop closeness with others easily, feel According to attachment theory (Ainsworth et al., 1978;
comfortable depending on others and having others depend on Bowlby, 1973, 1980), the behavior of securely and avoidantly
them, and rarely are concerned about being abandoned or attached adults should differ markedly in situations that foster
others becoming extremely close to them (Hazan & Shaver, anxiety. Secure attachment originates from a history of experi-
1987). Their romantic relationships, in turn, tend to be charac- ences in which attempts to establish physical and psychological
terized by more frequent positive affect (Simpson, 1990), by contact with attachment figures during times of distress have
higher levels of trust, commitment, satisfaction, and interde- been routinely successful (Crittenden & Ainsworth, 1989). As a
pendence (Collins & Read, 1990; Simpson, 1990), and by result, securely attached persons consider themselves to be
happy, positive, and trusting styles of love (Hazan & Shaver, worthy and valued and others to be reliable and trustworthy,
1987; Hendrick & Hendrick, 1989; Levy & Davis, 1988). and they tend to be involved in pleasant, emotionally support-
Those who manifest an anxious/ambivalent style tend to har- ive relationships that corroborate these mental models. More
bor mental models of themselves as being misunderstood, un- securely attached persons, therefore, ought to seek out more
confident, and underappreciated, and of significant others as emotional support from their partners when they feel dis-
being undependable and either unwilling or unable to pledge tressed or anxious and to offer more support when they sense
themselves to committed, long-term relationships. They report their partners require it.
that others are reluctant to get as close as they prefer, frequently Avoidant attachment, on the other hand, results from re-
worry that their partners do not truly love them or will aban- peated experiences in which efforts to establish contact with
don them, and often desire to become extremely close to their attachment figures have been rebuffed or thwarted (Crittenden
partners (Hazan & Shaver, 1987). Their relationships tend to & Ainsworth, 1989). Repeated failure to gain proximity eventu-
involve more frequent negative affect (Simpson, 1990), lower ally suppresses future behavioral attempts to achieve it. Accord-
levels of trust, commitment, satisfaction, and interdependence ing to attachment theory (Ainsworth et al, 1978), however,
(Collins & Read, 1990; Simpson, 1990), and obsessive, preoccu- these failures do not destroy the desire for proximity. Rather,
pied, and jealous forms of love (Hazan & Shaver, 1987; Hen- avoidant persons associate the proximity need with rejection,
drick & Hendrick, 1989; Levy & Davis, 1988). and this ultimately produces defensive efforts to keep the desire
Finally, those who have an avoidant style perceive themselves for proximity in check. Ainsworth et al. suggest that avoidant
as being aloof, emotionally distant, and skeptical, and signifi- persons want and need proximity yet simultaneously fear it.
cant others as being unreliable or overly eager to make long- Under conditions of stress, therefore, the attachment system
term commitments to relationships. Avoidant individuals indi- should be suppressed because of defensive information process-
cate that they are uncomfortable being close to others, find it ing designed to prevent awareness of attachment needs (Bowlby,
difficult to completely trust and depend on others, and become 1980; Main & Goldwyn, 1984). Accordingly, more avoidant per-
apprehensive whenever anyone gets too close (Hazan & Shaver, sons confronted with strong anxiety—an event that normally
1987). Similar to anxious people, avoidant individuals tend to ought to activate the attachment system—should be less likely
be involved in relationships characterized by more frequent neg- to seek support from their partners. To suppress the attachment
ative affect (Simpson, 1990) and by lower levels of trust, com- system, they also should physically and psychologically dis-
mitment, satisfaction, and interdependence (Collins & Read, tance themselves from their partners. More avoidant persons
1990; Simpson, 1990). The negative nature of their relation- also should display corresponding suppression tendencies
ships, however, appears to stem from acute fear of intimacy when their partners are distressed and need support. As their
rather than obsessive preoccupation with their romantic partner's need for support increases, they should not provide
partners (Hazan & Shaver, 1987; Hendrick & Hendrick, 1989; commensurate increases in support. In fact, they ought to de-
Levy & Davis, 1988). fensively retract from their partner.
The present research explores whether and how attachment Anxious attachment stems from experiences in which at-
styles moderate spontaneous support seeking and support giv- tempts to make contact with attachment figures have been as-
ing in romantic couples when one member of the dyad experi- sociated with inconsistent or unpredictable behavioral re-
ences anxiety. Bowlby (1973,1979) has argued that the attach- sponses from them (Crittenden & Ainsworth, 1989). Because
ment system should be most strongly activated under condi- the prospect of gaining proximity is unpredictable yet possible,
tions of distress (e.g., when an individual is ill, fatigued, or the attachment system in anxious persons is believed to be
afraid). Moreover, it is under these conditions that the behav- readily and strongly activated during times of distress (Bowlby,
ioral effects associated with attachment per se, relative to other 1980; Main & Goldwyn, 1984). Thus, according to one interpre-
facets of relationships such as love (Rubin, 1970) or closeness tation of attachment theory, persons who are more anxiously
436 J. SIMPSON, W RHOLES, AND J. NELLIGAN

attached and who are confronted with strong anxiety should be inclined to do so. As anxiety increases, support seeking may be
highly motivated to seek and perhaps give support during times most pronounced among more anxiously attached women.
of distress, maybe even more so than secure persons. These (See the qualifications discussed above concerning hypotheses
predictions, however, may need to be qualified. During adult- for anxious/ambivalent persons.)
hood, anxious individuals yearn to receive considerable emo-
tional support from their romantic partners. Yet they also tend Hypothesis 2
to perceive partners as being unpredictable in terms of their
emotional availability and supportiveness, and they tend to be As their partners display greater anxiety, more securely at-
involved in affectively unpleasant relationships in which strong tached men should provide greater reassurance and emotional
emotional support does not exist. Thus, although certain forces support, whereas more avoidant men should offer less support.
might compel anxious people to seek support, other forces may Under these conditions, support giving may be most pro-
inhibit these tendencies. Moreover, anxious/ambivalent adults nounced among more anxiously attached men.
tend to be preoccupied with their own needs and to display
anger toward others because of unresolved conflicts stemming Hypothesis 3
from past or present relationships (Main et al., 1985). This self-
absorption and resentment might make them unreliable as As women discuss their anxiety in greater detail with their
sources of support for others. As a result of these factors, rela- partners, more securely attached men should provide more sup-
tions between the anxious/ambivalent style and indicators of portive and reassuring comments compared with more avoid-
both support seeking and support giving could be attenuated. ant men. Supportive comments may be highly pronounced
among anxiously attached men.

Overview and Hypotheses Hypothesis 4


In the present study, female members of 83 dating couples As men make more supportive comments, their partners
were told that they would participate in an "activity" known to should be more calmed. This calming effect, however, should
elicit anxiety in most people. The nature of the activity was be more pronounced for more securely attached women than
unspecified. Before participating in it, they were escorted to a for more avoidant ones. Attachment theory makes no clear pre-
private "waiting room" where they waited for five min with dictions about the relation between anxious attachment and the
their male partner, who ostensibly was about to take part in a extent of calming.
different study. During this period, the couple's spontaneous
interaction was unobtrusively videotaped. Trained observers Hypothesis 5
then evaluated each couple's interaction during the five min
period in terms of (a) global adjective ratings of each woman's As their level of anxiety increases, more avoidantly attached
support-seeking and each man's support-giving behavior; (b) women should display greater resistance to physical contact
ratings of the content and consequences of the couple's discus- (e.g., touching) from their partners, whereas more securely and
sion about the anxiety-provoking event; and (c) ratings of the anxiously attached women should display less resistance. More
frequency with which various overt physical behaviors oc- anxiously attached women should be least likely to resist con-
curred (e.g., touching the partner, smiling, and eye contact). tact.
The attachment system was activated through different
channels for men and women in the present study. For women, Method
the attachment system was elicited by exposure to a stressful
situation designed to directly evoke strong distress and anxiety; Participants
for men, it was triggered by the vulnerability and distress dis- Eighty-three dating couples (83 men and 83 women), at least one
played by their partners. According to attachment theory member of whom was enrolled in introductory psychology at Texas
(Bowlby, 1973), distress that is experienced either directly or A&M University, participated in this investigation. Mean ages of men
vicariously through the emotional needs of one's partner can and women were 19.5 years and 18.9 years, respectively. To ensure that
activate the mental models of self and others harbored by se- participants were involved in established relationships, couples were
curely, anxiously, and avoidantly attached people. These mental required to have dated one another for at least 3 months prior to partici-
models, in turn, should guide responses to the partner. Securely pation. Mean length of dating relationship was 17.9 months. Members
attached persons should respond by encouraging proximity and
providing support, whereas avoidant and anxious/ambivalent
1
persons ought to react differently. These hypotheses are stated in terms of the behaviors of male and
In view of the theoretical considerations discussed above, we female partners. According to attachment theory, however, the same
tested the followingfivehypotheses:' predictions would be made if the roles of support seeking and support
giving enacted by men and women were reversed. Given the limited
number of couples available for research, we subjected only women to
Hypothesis 1 the anxiety-provoking procedures. We did so because we thought that
women, relative to men, might be more willing to mention and openly
As they experience greater anxiety, more securely and more discuss their feelings about the stressful situation. We intend to study
anxiously attached women should seek greater support from support seeking by men and support giving by women in a subsequent
their partners, whereas more avoidant women should be less study.
ATTACHMENT STYLES 437

of 15 dyads reported that they were dating people in addition to the my partners) don't really love me"; (1) "I rarely worry about my
current partner, 63 dyads indicated that they were dating each other partner(s) leaving me"; and (m) "I often want to merge completely with
exclusively (but were not engaged), and 5 dyads reported that they were others, and this desire sometimes scares them away." Items a through e
engaged. were taken from Hazan and Shaver's "secure" vignette description.
Items f through i and j through m were taken from the avoidant and
anxious/ambivalent vignettes, respectively.
Procedure For purposes of providing discriminant validity, participants also
completed Rubin's Love scale (Rubin, 1970) and the Relationship
Overview. The investigation consisted of two phases. In Phase 1, Closeness Inventory (RCI; Berscheid et al, 1989).
couples responded to a series of questionnaires that inquired about Phase 2: Stress induction and behavioral observation. Once the fe-
their attachment style and their current relationship. In Phase 2, cou- male member of each dyad had completed the survey, she was escorted
ples were unobtrusively videotaped after the woman had been told she to a waiting room, where the stress induction procedure took place.
was going to be exposed to an anxiety-provoking activity. Independent The experimenterfirstpretended to take the woman's pulse to accentu-
observers then rated the behavior of both members of each dyad from ate the stressful nature of the "impending" experimental situation. The
these videotapes. experimenter then said the following:
Phase 1: Questionnaire survey. When couples arrived for the study
(in individual pairs), they were told that they would be participating in In the next few minutes, you are going to be exposed to a situa-
two unrelated projects. The first one involved completing a series of tion and set of experimental procedures that arouse considerable
questionnaires that enquired about themselves and their current rela- anxiety and distress in most people. Due to the nature of these
procedures, I cannot tell you any more at the moment. Of course,
tionship. To guarantee that couples did not communicate during Phase 111 answer any questions or concerns you have after the experi-
1, they completed the questionnaires in different rooms. ment is over.
As a prelude to the stress manipulation administered to women in
Phase 2, female members of each dyad were informed (when reading The purpose of this statement was to arouse anxiety without disclosing
the informed consent form) that some of the procedures in the second the nature of the stressful procedures.
study provoked stress and anxiety in many people. However, they also The experimenter then led the woman to a room normally used for
were told that they could discontinue participation in the study at any psychophysiological research. She opened a heavy metal door, expos-
time and for any reason without loss of promised compensation. ing a darkened, windowless room that looked like an isolation
Both members of each dyad then completed the questionnaire sur- chamber and contained psychophysiological equipment. The experi-
vey. Embedded in the survey was a Likert-type version of Hazan and menter made sure each participant peered into the room and saw the
Shaver's (1987) measure of the three attachment styles (see Simpson, equipment. She then stated that the equipment was not fully set up,
1990). Even though Hazan and Shaver's attachment measure was devel- after which she led the participant back to the waiting room. The
oped according to descriptions provided by Ainsworth et al. (1978) of experimenter gave no further details about the nature ofthe experimen-
avoidantly, securely, and anxiously attached infants, the measure has tal procedures.
some problematic psychometric properties (see Hazan & Shaver, 1987; The experimenter then escorted the man to the waiting room and
Hendrick & Hendrick, 1989). First, people classify themselves as be- left the couple alone for 5 min. Nothing was said to the man about the
longing to one of three mutually exclusive attachment categories with- woman's impending stressful situation. During this time, the couple's
out specifying the extent to which the chosen category characterizes spontaneous interaction was unobtrusively videotaped by a small
them. As a result, meaningful individual difference variability that video camera encased in an opaque, domed housing unit mounted
exists within each category cannot be assessed. Second, this categori- from the ceiling in one corner of the room. Five min later, the experi-
cal classification method assesses an individual's standing on only one menter reentered the room and explained that, due to "malfunction-
attachment style, despite the fact that some adults may be best charac- ing" equipment, the experiment could not be conducted. Participants
terized as a blend of two or more styles. Third, this method of classifica- were informed, however, that they wouldreceivethe promised experi-
tion places severe constraints on the types of statistical analyses that mental credit.2
can be performed. Fourth, measurement of individual differences in Following this, participants were debriefed and informed that they
adults traditionally has focused on linear relations between multiple- had been videotaped. The reason for the deception was explained, and
item measures designed to assess specific, continuously distributed they were given the opportunity to erase their videotapes if they
constructs (Jackson, 1971). Cases in which discrete classes underlie wanted to do so. No couple requested such action. Participants then
adult individual differences are rare. signed areleaseform allowing their videotapes to be used for research
Hence, the three attachment vignettes originally created by Hazan purposes. All participants agreed to this request.
and Shaver (1987) were decomposed into 13 individual sentences, each
of which was answered on a 7-point Likert-type scale (1 = strongly
disagree and 7 = strongly agree). To control for acquiescence response Coding of Videotapes
biases, 3 sentences were worded in a negative direction. Moreover, very
slight alterations in wording designed to reduce item response diffi- Three trained raters, blind to both the experimental hypotheses and
culty were made for 2 sentences (Items d and m below). Participants participants' attachment styles, then viewed the videotapes and evalu-
rated the following items according to how they typically felt toward
romantic partners in general: (a) "I find itrelativelyeasy to get close to
2
others"; (b) "I'm not very comfortable having to depend on other peo- As a prelude to debriefing, each partner evaluated the woman's
ple"; (c) "I'm comfortable having others depend on me"; (d) "I rarely general emotional reaction to the experimental situation. Because
worry about being abandoned by others"; (e) "I don't like people get- these ratings were made after participants had been told the anxiety-
ting too close to me"; (f) "I'm somewhat uncomfortable being too close provoking event would not occur and because they were developed to
to others"; (g) "I find it difficult to trust others completely"; (h) "I'm assess how participants were generally feeling prior to debriefing, they
nervous whenever anyone gets too close to me"; (i) "Others often want are less valid markers of anxiety compared with ratings made by
me to be more intimate than I feel comfortable being"; (j) "Others trained observers during the 5 min observation period. Thus, observer
often are reluctant to get as close as I would like"; (k) "I often worry that ratings were used to index women's anxiety.
438 J. SIMPSON, W RHOLES, AND J. NELLIGAN

ated the behavior of both members of each dyad during the 5 min ranged from .70 to .75). On occasions when raters' evaluations were not
waiting period. unanimous for the three dichotomous items, the majority response
Global adjective ratings. Raters first evaluated men and women on was treated as the rated value.
global adjectives. Given our predictions, women were evaluated on a Physical behavior ratings. Finally, raters coded various physical be-
series of adjectives designed to broadly assess two major dimensions: haviors displayed by men and women during the 5 min waiting period.
their level of anxiety-fear (indexed by adjectives such as anxious, fear- Many of the behaviors that transpire in dyadic interactions can be
ful, etc.) and the extent to which they sought comfort and support from divided into two categories: those that indicate approach tendencies
their partner (indexed by adjectives such as clingy, promoting contact, and those that reflect resistance or avoidance (see Knapp, 1978). Behav-
etc.). Relevant adjectives were chosen in part on the basis of descrip- iors indicative of approach can be further subdivided into those that
tions of each dimension found in previous theory (Bowlby, 1973) and can occur at a distance (e.g., smiling or eye contact) and those that
research (Ainsworth et al, 1978) on attachment. involve direct physical contact (e.g., touching). Because tendencies to
Ratings for women were made on 7-point Likert-type scales (1 = not approach one's partner under stress ought to reflect comfort seeking,
at all and 7 = a great deal/extremely). To canvas each dimension and tendencies to avoid him or her ought to connote contact resistance
broadly (see Loevinger, 1957), ratings were made on 21 adjectives: (see Bowlby, 1973), we had raters record the number of times each of 15
warm, self-confident, friendly, relaxed, anxious, nice, self-disclosing, behaviors occurred: touching partner on the body, touching partner on
humorous, hostile, cold/distant, submissive, dull/boring, critical/nega-the face, placing arm or hand on partner's shoulder, hugs, kisses, hand-
tive, fearful, independent, clingy/dependent, comfortable with physicalholding, resisting contact, leaning toward partner, moving toward
contact, promoting physical contact, paying attention to partner, quarrel-partner, turning head or body toward partner, leaning away from
some, and emotionally avoidant. Interrater reliabilities were low for 4 partner, moving away from partner, turning head or body away from
items (hostile, cold/distant, critical/negative, and quarrelsome), and, partner, eye contacts, and smiles. For each behavioral category, in-
therefore, these items were dropped from further analyses. Interrater terrater reliabilities were good (for men, M= .69 across all behaviors,
reliabilities for each of the remaining 17 adjectives were reasonable range = .39-90; for women, M = .75, range = .46-97). Moreover,
given only three raters (M = .60 across all adjectives; range = .41 -.86). interrater agreement also was good (mean Tau statistic was .73 for
Raters' ratings were aggregated for each adjective, resulting in 17 women and .75 for men across all 15 behaviors; see Tinsley & Weiss,
summed indexes for women. 1975). Hence, raters' behavioral counts were aggregated for each cate-
gory, resulting in 15 summed indexes for both members of each dyad.
Given our predictions, men were evaluated on several adjectives de-
signed to broadly tap one major dimension: the extent to which they
offered reassurance and emotional support (e.g., indexed by adjectives Results
such as supportive, reassuring, sympathetic, etc.). Adjectives were se-
lected according to descriptions from prior theory and research on Construction of Attachment Indexes
attachment of what this dimension should entail. Following Loe-
vinger's (1957) suggestions, ratings for men were made on the same Before constructing continuous measures of attachment
7-point Likert-type scales for the following 27 adjectives: warm, self- styles, we first performed two principal-axis factor analyses
confident, emotionally supportive, friendly, competent at calming the (SPSSX PA2) followed by varimax rotation on the 13 attach-
partner, relaxed, reassuring, nice, open/accepting, humorous, socially ment items, one for women and one for men.3 Factors were
skilled, hostile, cold/distant, dependable, dull/boring, critical/negative, extracted on the basis of eigenvalue scree (Cattell, 1966) and
condescending, comfortable with physical contact, promoting physicalfactor interpretability. In line with previous research (Hendrick
contact, paying attention to partner, nurturant, quarrelsome, emotion-& Hendrick, 1989; Simpson, 1990), these analyses revealed two
ally avoidant, helpful, sympathetic, attentive to partner's needs, and re-factors within each sex: Secure versus Avoidant and Anxious
sponsive to partner's needs. Interrater reliabilities were low for 5 items
(self-confident, relaxed, hostile, critical/negative, and condescending),
versus Nonanxious. Once all items were keyed in the proper
and, thus, these items were excluded from further analyses. Interrater direction, the avoidant/secure attachment index was con-
reliabilities for each of the 22 remaining adjectives were reasonable structed by aggregating Items a, b, c, e, f, g, h, and i (Cronbach's
(M= .64 across all adjectives; range = .41-83). Thus, raters' ratings alpha = .81 for both men and women). Higher scores reflected
were aggregated on each adjective, resulting in 22 summed indexes greater avoidance, whereas lower scores reflected greater secu-
for men.
Conversation ratings. Raters then coded the specific verbal content
of partners' conversations on the following dimensions: (a) "Does the 3
All of the factor analyses reported in this article used principal axis
female mention (to her partner) she is to undergo a stressful proce- factoring followed by varimax rotation. Principal axis factoring was
dure?" (coded yes or no); (b) "If yes, to what extent does she discuss her used instead of principal components analysis for three reasons. First,
feelings about being anxious, worried, or upset?" (evaluated on a 9- principal components analysis assumes that all variables are measured
point Likert-type scale, with 1 = little and 9 = a great deal); (c) "If the without error, whereas principal axis factoring does not (West & Finch,
female does mention she is to undergo a stressful procedure, does the in press). Because our measures were not perfectly reliable, principal
male make supportive comments?" (coded yes or no); (d) "If yes, to axis factoring was deemed to be a more appropriate technique. Second,
what extent was he supportive?" (evaluated on a 9-point scale, with 1 = principal axis factoring provides more stable and replicable results,
little and 9 = extremely); (e) "If the female does mention she is to un- particularly when factors are large or when fewer than 35 variables are
dergo a stressful procedure, does the male try to avoid or downplay her being analyzed, as was the case in our study (see Loehlin, 1990; Ve-
comments?" (coded yes or no); (f) "If yes, to what extent did he avoid or licher & Jackson, 1990; West & Finch, in press). Third, principal axis
downplay them?" (evaluated on a 9-point scale, with 1 = little and 9 = factoring extracts the common item variance underlying a covariance
extremely); and (g) "To what extent does she seem reassured or calmed matrix, whereas principal components analysis extracts all the vari-
by the (conversational) behavior of the male?" (evaluated on a 9-point ance (West & Finch, in press). We, of course, were interested in extract-
scale, with 1 = not at all and 9 = extremely). Raters' responses to each ing the common variance underlying our measures. Varimax rotations
question were aggregated. For the four questions requiring Likert-type were used because all dimensions were presumed to be relatively or-
responses, raters' evaluations were reliable (interrater reliabilities thogonal.
ATTACHMENT STYLES 439

rity. The anxious attachment index was created by aggregating on the Face and included kisses (84) and leaning toward
Item d and Items j through m. Higher scores reflected greater partner (55). Factor 4 assessed the frequency of attempts to
anxiousness (Cronbach's alphas = .58 and .61 for men and Resist Contact and included resisting contact (75) and moving
women, respectively). away from partner (64). We then constructed factor scores
for men.
The second factor analysis consisted of raters' counts of be-
Construction of Rating Indexes haviors displayed by women. Three factors accounting for 50%
Adjective factors. We performed two principal-axis factor of the variance emerged. Factor 1 reflected the amount of
analyses (SPSSX PA2) followed by varimax rotation on raters' Touching on the Face. Items loading above .50 included touch-
aggregated adjective ratings. The first analysis involved raters' ing on face (82), hugs (78), and kisses (75). Factor 2 indexed the
evaluations of the 22 aggregated adjective indexes on which extent of Smiling/Eye Contact and included eye contacts (93),
men were evaluated. On the basis of eigenvalue scree (Cattell, turning toward partner (74), and smiles (67). Factor 3 tapped
1966) and factor interpretability, two factors accounting for 72% efforts to Resist Contact and included resisting contact (92)
of the variance were extracted. The first factor reflected the and leaning away from partner (88). We then constructed fac-
amount of Reassurance/Emotional Support displayed by men tor scores for women.
toward their partners. Items loading at or above .50 included
the following: reassuring (86), sympathetic (84), emotionally Analyses of Couples' Interaction
supportive (82), competent at calming partner (75), responsiveto
partner's needs (73), helpful (.73), nurturant (69), attentive to A complete set of analyses identical to the ones reported
partners needs (.65), dependable (56), not emotionally avoidant below did not reveal reliable findings for the anxious attach-
(.54), and warm (.53). The second factor assessed the extent of ment index.4 Hence, this index is not discussed in the Results
general Warmth/Friendliness displayed by men. Items loading section, although we return to these nullfindingsin the Discus-
highly (50 or greater) included the following: promoting physi- sion. The avoidant/secure (A/S) index, however, was associated
cal contact (78), not dull/boringP4), paying attention to partner with how partners responded to one another.
(70), comfortable with physical contact (.68), not cold/distant Analyses ofadjective factors. Thefirstanalysis examined the
(60), humorous (60), warm (58), friendly (56), not emotionally relationship between anxiety and support seeking in women.
avoidant (55), and attentive to partner's needs (50). Given our Hypothesis 1 predicted that support seeking would increase
predictions, we then calculated factor scores for men on the with anxiety in more secure women, whereas it would decrease
first factor (i^e., Reassurance/Emotional Support). in more avoidant women. To test this hypothesis, we conducted
The second factor analysis involved raters' aggregated evalua- a hierarchical regression analysis. Observer-rated support seek-
tions of the 17 adjectives on which women were rated. Accord- ing (assessed by the Comfort/Support Seeking factor) served as
ing to the eigenvalue scree (Cattell, 1966) and factor interpret- the criterion variable. The predictor variables included
ability, three factors accounting for 79% of the variance women's self-reported A/S index (Step 1), observer-rated anxi-
emerged. The first factor reflected the amount of Warmth/ ety in women (the Anxiety/Fear factor; Step 2), and the interac-
Friendliness displayed by women. Items loading highly (50 or tion of these two variables (Step 3). Only the interaction term
greater) included the following: warm (91), not dull/boring(88), was significant, F(l, 79) = 3.97, p < .05. Specifically, among
friendly {$l),notcold/distant{$b\nice{$$), paying attentionto more secure women, increases in observer-rated anxiety (ije,
partner's needs (79), not emotionally avoidant (78), humorous increases across subjects) were positively related to increased
(75), self-disclosing (74), self-confident (61), promoting contact levels of observer-rated support seeking. More avoidant
(.53), and comfortable with contact (53). The second factor in- women, however, showed the opposite tendency; increases in
dexed Anxiety/Fear and included anxious (94), not relaxed anxiety were associated with less support seeking. As depicted
(92), and fearful (86). The third factor tapped Comfort/Sup- in Figure 1, the unstandardized beta relating anxiety to support
port Seeking and included clingy (84), not independent (78), seeking for women with higher avoidance scores (1 SD above
comfortable with contact (67), and promoting contact (67). In the A/S mean) was —0.21; for highly secure women (1 SD below
view of our predictions, we then calculated factor scores on the the A/S mean), the beta was 0.23.
second and third factors for women (La, Anxiety/Fear and The two predictor variables used in this analysis—observer-
Comfort/Support Seeking). rated anxiety in women and women's self-reported A/S scores
Physical behavior factors. We next performed two principal- —were not correlated (r = —.05). This finding is noteworthy
axis factor analyses (SPSSX PA2) followed by varimax rotation
on raters' physical behavior counts. The first analysis involved
4
raters' counts of behaviors displayed by men. On the basis of Internal analyses revealed that the null effects for the anxious at-
eigenvalue scree (Cattell, 1966) and factor interpretability, four tachment index were not attributable to its lowerreliabilitycompared
factors accounting for 48% of the variance emerged. Factor 1 to the avoidant/secure index. Specifically, we conducted the same set
of analyses on the anxious attachment index as are reported for the
reflected the amount of Smiling/Eye Contact directed at the avoidant/secure one. Once all main effects and interactions were calcu-
women. Items loading highly (50 or greater) included eye con- lated, we estimated how large each effect would have been if the
tacts (96), turning toward the partner (81), and smiles (70). anxious attachment index possessed perfectreliability(using a formula
Factor 2 tapped the extent of Hugs/Touching on the Body and found in Cohen & Cohen, 1983). Even after correcting for attenuation
included touching on shoulders (89), touching on body (74), because of the lower reliability of the anxious attachment index, no
and hugs (.68). Factor 3 indexed the extent of Kisses/Touching significant effects emerged.
440 J. SIMPSON, W RHOLES, AND J. NELLIGAN

because it indicates that attachment style is not confounded +0.5 r


with the degree of anxiety experienced.
Hypothesis 2 focused on the relationship between support E
offered by men and their partners' level of anxiety. We expected M
higher levels of partner anxiety to be associated with increased ,O •0.25 V
levels of support giving among more secure men, but not
among more avoidant ones. A hierarchical regression analysis \O
involving the men's self-reported A/S index (Step 1), observer-
rated anxiety in women (the Anxiety/Fear factor; Step 2), and
the interaction of these two variables (Step 3) revealed that the
interaction of self-reported A/S and observer-rated anxiety ac-
counted for a significant amount of variance in observer-rated
supportiveness in men (indexed by the Reassurance/Emotional -0.25
Support factor), F(l, 79) = 4.85, p < .05. Among more secure
men, higher levels of partner anxiety were associated with
greater support. Conversely, among men scoring higher in avoid-
ance, higher partner anxiety was associated with lower levels of -0.5
support. As presented in Figure 2, the unstandardized beta -1 -.5 •.5 +1
relating anxiety to support among more secure men (1 SD be- Low High
low the A/S mean) was 0.28. The beta for more avoidant men (1
SD above the A/S mean) was -0.25. 5 ANXIETY-FEAR
Sixteen women did not discuss the anxiety-provoking activ- Figure 2. The relation between women's standing on the Anxiety/
ity with their partners. Thus, we conducted a second regression Fear factor and men's standing on the Reassurance/Emotional Support
analysis of supportiveness by men, partialing out the dichoto- factor for highly secure and highly avoidant men. (Regression lines are
mous conversational variable (discuss vs. no discuss) in the first plotted for men 1 SD above and below the mean on the avoidant/secure
step of the analysis. Partialing did not affect the results re- attachment index. Values for the Anxiety/Fear and Reassurance/Emo-
ported above. We also dropped these 16 individuals and reana- tional Support factors are presented on the axes in SD units for each
lyzed the data, once again obtaining the same outcome. factor.)
In examining supportiveness by men, we found as expected

that women's anxiety level interacted with their male partners'


attachment style. In other words, men seemed to have been
c responding to their partner's affective state. It is possible, how-
o
M ever, that men also reacted to the support-seeking efforts of
F their partner, independent of her emotional state. We con-
O ducted a regression analysis similar to the ones described above
R +0.25 - to determine whether the level of support seeking displayed by
women also interacted with their partners' attachment style.
S Results revealed that partner support seeking was not related to
U
P Secure supportiveness by men, either alone or in interaction with men's
P 0_ \ x attachment style. These findings corroborate our original hy-
O Avoidant
R \
pothesis that the behavior of men is largely contingent on their
T partner's emotional state rather than on overt support-seeking
efforts.
| -0.25 \ Analyses of conversation items. We next focused on ob-
E servers' ratings of the extent to which women openly discussed
K their anxiety and how men verbally responded to these disclo-
1
N sures. Sixteen of the 83 women did not mention the anxiety-
G -0.5 1 i I i i
provoking situation to their partners. A t test revealed that
-1 -.5 0 +.5 +1 these women had more avoidant scores on the A/S index than
Low High women who did discuss the situation, ?(81) = 1.97, p < .05,
one-tailed. This is consistent with thefindingsconcerning sup-
ANXIETY-FEAR port seeking among more avoidant women discussed above.
Figure 1. The relation between the Anxiety/Fear factor and the Com- Among the 67 women who did mention the stressful proce-
fort/Support Seeking factor for highly secure and highly avoidant dure, the correlation between self-reported A/S scores and the
women. (Regression lines are plotted for women 1 SD above and below
the mean on the avoidant/secure attachment index. Values for the Anxi-
ety/Fear and Comfort/Support Seeking factors are presented on the 5
No predictions were made for the Warmth/Friendliness adjective
axes in SD units for each factor.) factors. Hence, they were not analyzed.
ATTACHMENT STYLES 441

observer-rated extent of discussion measure (Item b: ". . . to (Step 3). This analysis differed from the previous one in that it
what extent does she discuss her feelings about being anxious, examined calming as a function of the extent to which feelings
worried, or upset?") was not significant (r =. 13). This is consis- were discussed, independent of the extent to which supportive
tent with the previous finding that A/S scores were not strongly remarks were made by men. This analysis revealed a significant
associated with observers' ratings of women's anxiety. interaction between extent of discussion and women's A/S
We next performed a regression analysis on the 67 couples scores, F(l, 63) = 5.76, p < .05. Specifically, extent of discussion
who discussed the stressor. Ratings of the extent to which men was positively related to ratings of being calmed for more avoi-
made supportive comments in direct response to women's dis- dant women (0 = 0.34), but not for more secure women (/3 =
cussion of their feelings were treated as the criterion measure —0.11). That is, more extensive discussion alone had a more
(Item d: ". . . to what extent was he supportive?"), and the fol- positive effect on highly avoidant women.
lowing measures served as predictors: men's self-reported A/S To determine whether this interaction was independent of
scores (Step 1), observers' ratings of the extent to which the the one between women's A/S scores and their partners' level of
women discussed feelings of anxiety (Step 2), and the interac- supportive comments, an additional analysis was conducted in
tion of these two variables (Step 3). This analysis revealed a which the interaction between men's supportiveness and
significant main effect for the extent of discussion variable, F(l, women's A/S scores was partialed out prior to entering the ex-
64) = 29.92, p < .001, and a significant interaction between this tent of Discussion X Women's A/S Index interaction term. This
variable and the men's A/S index, F(l, 63) = 7.45, p < .01. The analysis revealed that this interaction accounted for a margin-
main effect indicates that men made more supportive com- airy reliable amount of variance over and above the men's Sup-
ments as women discussed their feelings more extensively. The portive Comments X Women's A/S Index interaction, F(\, 61) =
interaction qualifies this trend in that increased levels of sup- 3.69, p < .06.
portive comments were more pronounced for more secure men
Analyses of physical behavior factors. We next focused on
than for more avoidant ones. The beta relating men's supportive
physical behaviors that transpired during the observation pe-
comments to the extent of discussion by women variable was
0.89 among more secure men. Among more avoidant men, the riod. The first set of analyses focused on women's behaviors as
corresponding beta was 0.29. In other words, the tendency to predicted by women's A/S scores (Step 1), the frequency with
make extensive supportive comments was more pronounced which men touched their partners on the body (the Hugs/
among men scoring in the more secure range of the A/S index, Touching on the Body factor; Step 2), and the interaction of
but it was not totally absent among more avoidant men. these two variables (Step 3). Separate analyses were conducted
in which each of the women's three physical behavior factors
Next, we conducted a regression analysis in which the degree
(Touching on the Face, Smiling/Eye Contact, and Resisting
to which men attempted to avoid or downplay women's con-
Contact) were treated as dependent measures. These analyses
cerns was treated as the criterion variable (Item e:". . . does the
showed that men's touching on the body had an impact on
male try to avoid or downplay her comments?"), and men's
physical behavior by women that varied according to women's
self-reported A/S, observer-rated extent of discussion by
women, and their interaction served as predictors. This analysis attachment style. The first analysis revealed a main effect such
yielded no reliable effects. that women exhibited more resistance as men's touching on the
body increased, F(l, 80) = 48.29, p< .001. However, in accord
We then performed two analyses in which observers' ratings
of the degree to which women were calmed (Item g: "To what with Hypothesis 5, this tendency was qualified by a significant
extent does she seem reassured or calmed by the [conversa- interaction between men's touching and women's A/S scores,
tional] behavior of the male?") served as the criterion variable. F(l, 79) = 6.65, p < .05. Increased levels of touching by men
Predictors included (a) women's A/S index (Step 1 of each analy- were more strongly related to increased resistance among more
sis); (b) observer-rated supportive comments by men and the avoidant women (/? = 0.80) than among more secure women
degree to which men attempted to avoid or downplay women's (0=0.21).
concerns (Step 2, with each variable used alone in separate Closely related findings emerged in an analysis of facial
analyses); and (c) the interaction of the first and second vari- touching and kissing on the part of women. This analysis
ables (Step 3). Results of these two analyses showed that women yielded a significant main effect for men's touching, F(l, 80) =
were more calmed when men made more supportive remarks, 5.37, p < .05, and a marginally significant interaction between
F(l, 64) = 57.57, p < .001. This trend, however, was contingent women's A/S scores and men's touching, F(l, 79) = 3.50, p <
on women's A/S scores, F(l, 63) = 4.56, p< .05. Contrary to the .07. The main effect indicates that more body touching by men
predictions of Hypothesis 4, increases in supportive comments was associated with more frequent facial touching and kissing
had a somewhat greater calming effect on more avoidant by women. The interaction suggests that this trend was more
women (0 = 0.59) than on less avoidant ones (# = 0.33). The pronounced for more secure women. The betas relating men's
other analysis showed no effects for women's A/S scores. It re- body touching to women's facial touching were 0.61 for more
vealed only main effects for men's downplaying of women's secure women and 0.10 for more avoidant women. The final
concerns, F(l, 64) = 43.15, p < .001. Not surprisingly, women analyses revealed that men's body touching was not associated
were less calmed if their concerns were avoided or downplayed. with the women's Smiling/Eye Contact factor.
The final analyses of conversations examined observer rat- The above analyses reveal a consistent pattern of effects in
ings of how calmed women appeared to be as predicted by which A/S scores moderate the reaction of women to body
women's A/S scores (Step 1), the extent to which they discussed touching by their partners. Analyses involving the men's other
their feelings (Step 2), and the interaction of these variables physical behavior factors indicated that women's reactions to
442 J. SIMPSON, W RHOLES, AND J. NELLIGAN

these forms of men's behavior were largely unrelated to their -.05). The independence of men's and women's scores on the
A/S scores. A/S index suggests that this dimension does not assess the
Men's physical behavior factors were analyzed in the same global quality of the current relationship.
way as were the women's behaviors. Although a few significant To marshal further evidence for discriminant validity, we
findings emerged, they did not present a coherent and easily reconducted all analyses on which significant effects for the
interpretable view of men's responses. In general, the physical A/S index emerged, partialing out scores on Rubin's Love scale
behaviors of men were unrelated to their A/S scores. and the RCI from each dependent measure in Step 1 of each
Relations among the adjective, conversational, and physical analysis. These reanalyses showed that the inclusion of these
behavior measures. The adjective, conversational, and behav- two measures had minimal impact on the findings reported
ioral measures were not necessarily designed to be orthogonal. above. We also conducted analyses (similar to the ones pre-
In fact, we expected some of them to be moderately correlated. sented above involving the A/S index) in which love and close-
The reason for assessing so many potentially correlated mea- ness measures were substituted for the A/S index. Neither love
sures was to ensure that we canvased the content of couples' nor closeness reproduced the set of significant effects described
interactions as thoroughly and broadly as possible (see Loe- above. In fact, only one analysis, which involved an interaction
vinger, 1957). between women's closeness and men's supportive comments,
Among women, four measures were correlated significantly. was significant. Viewed together, these results suggest that the
The Anxiety/Fear adjective factor was correlated with the ex- A/S index possesses good discriminant validity in relation to
tent to which women discussed feelings of distress (r = .69), measures of closeness and love.
such that more anxious women discussed their feelings more
extensively. The Anxiety/Fear factor also was correlated with Discussion
the degree to which women were calmed by their partners (r =
-.25), such that more anxious women were calmed somewhat This investigation tested several hypotheses concerning how
less. The Comfort/Support Seeking adjective factor was corre- spontaneous interaction between couples differs as a function
lated with the physical factor of Touching on Face (r = .57), with of each member's attachment style when one member of the
women who sought more comfort being more likely to touch dyad is confronted with an anxiety-provoking situation. Find-
their partners on the face. The Comfort/Support Seeking factor ings revealed that secure and avoidant persons differ in extent
also was correlated with the Smiling/Eye Contact factor (r = of support seeking and support giving as a function of the level
-.23), with greater support seeking being associated with less of anxiety displayed by their romantic partner. Specifically,
smiling/eye contact. Among men, all significant correlations more secure women tend to seek out more support as their level
involved the Reassurance/Emotional Support adjective factor. of anxiety increases, whereas more avoidant women tend to
Not surprisingly, more supportive men made more supportive seek less support with increasing anxiety. Moreover, more se-
comments in the conversations (r = .85) and made fewer re- cure men tend to offer more support as their partners display
marks avoiding or downplaying their partners' concerns (r = greater anxiety, whereas more avoidant men are less inclined to
-.16). No other measures were reliably correlated within men do so. No significant effects emerged for the anxious attach-
or women. ment style.
In summary, most of the adjective, conversational, and behav- Two features of the present findings should be highlighted.
ioral measures were relatively orthogonal. This suggests that First, as discussed above, attachment theory proposes that envi-
the effects reported above across the three types of measures ronmental events can and do influence the extent to which the
are not overly redundant. attachment system is activated and, therefore, the level of sup-
Relations between the A/S index andother measures: Discrimi- port seeking and support giving displayed by persons who
nant validity. The final analyses examined relations between differ in attachment style. Indeed, the behavioral characteris-
the A/S index, Rubin's (1970) Love scale, and the RCI (Bers- tics prototypic of securely, avoidantly, and anxiously attached
cheid et al., 1989). Among men, the A/S index was not reliably persons should be most clearly witnessed when the system is
correlated with either love or closeness (rs = —.18 and -.15, strongly activated. Consistent with this perspective, most of the
respectively). Among women, the A/S index was reliably corre- major results in this investigation involve interactions between
lated with both love and closeness (rs = -.25 and -.35, respec- self-reported attachment style and observer-rated anxiety dis-
tively, ps < .01), such that more avoidant women characterized played by women. Hence, one cannot fully characterize the
their current relationship as somewhat lower in love and close- typical behaviors displayed by secure and avoidant persons
ness. without specifying the context in which the interaction occurs.
Several findings, however, revealed that A/S scores were dis- More avoidant men, for example, cannot be characterized as
criminant from love and closeness. First, although the A/S in- cold, distant, or aloof in general. Although they do behave in
dex was somewhat correlated with Rubin's Love scale and the this manner when their partner experiences higher levels of
RCI, the correlations were relatively small and reliable only for distress, their behavior is not the same when their partner's
women. Second, within-dyad correlations on love, closeness, distress is lower.
and the A/S index revealed striking differences between the Second, the discriminant validation results suggest that the
A/S index and the other two measures. Love and closeness were present findings are attributable to attachment styles per se
highly correlated within dyads (within-dyad rs = .58 and .62 for rather than to more global indicators of interdependence, such
love and closeness, respectively, ps < .001). By contrast, the as love or closeness. Thesefindingsare consistent with our view
within-dyad correlation for the A/S index was negligible (r = that attachment propensities should be conceptualized in spe-
ATTACHMENT STYLES 443

cific rather than in general terms. Previous theory and research data, however, appear to contradict this impression. How can
indicates that attachment styles primarily serve to regulate how this be explained?
individuals interpret, understand, and cope with negative emo- To understand these findings, it is important to recognize
tional experiences during stressful situations (cf. Kobak & from the outset that our index of anxiety in women measured
Sceery, 1988; Simpson, 1990). Distressing situations, however, relative anxiety in the experimental situation. Even women who
represent a limited subset of the events that normally occur displayed less anxiety were, in all likelihood, at least somewhat
within and affect established relationships. Patterns of support anxious or uncertain and in need of some support. To account
seeking and support giving associated with attachment styles, for why support seeking and support giving decline among
of course, may influence other aspects of relationships, such as more avoidant persons when anxiety rises, we must return to
satisfaction, closeness, and love between partners. Moreover, the conflict model of avoidance (Ainsworth et al., 1978). This
such tendencies may influence certain kinds of relationships model holds that the behavior of avoidant persons is a product
more than others. For example, attachment styles may exert a of conflicting motives, namely a desire for, yet a simultaneous
stronger influence within relationships facing chronic stress, fear of, proximity. When these motives are both strongly
especially if stress accentuates the importance of effective sup- aroused under stressful conditions, infants do not seek proxim-
port seeking and support giving. Nonetheless, the present re- ity. This, however, does not mean that avoidant infants never
sults suggest that attachment styles have a unique domain of seek proximity or always display clearly avoidant behavior
applicability and that they are not merely global measures of (Ainsworth et al, 1978). Under less stressful conditions in the
relationship functioning or quality. home or in laboratory play sessions, avoidant infants some-
The findings of this investigation can be classified into three times do seek proximity with their mother and protest when
categories: support seeking, reactions to support, and support separated from her.
giving. In what follows, we discuss these domains in the context Extrapolating from Ainsworth et al.'s (1978) research, we
of existing research on attachment, noting some interesting par- speculate that desire for proximity in adults is aroused more
allels with research on attachment in children. Although longi- strongly than fear of proximity when environmental conditions
tudinal evidence for continuity in attachment style over time are only moderately threatening and emotional distress is at
and across different relationships is mounting (see Sroufe et al., lower levels. However, increases in perceived threat or distress
1990), any conclusions about continuity over the life span re- sharply accelerate the onset of fear of proximity, resulting in
main very tentative. The utility of Bowlby's (1969,1973,1980) decreased proximity seeking and giving. Moreover, at lower
attachment theory for adult relationships, however, does not levels of threat or distress, avoidant persons may typically over-
require that attachment styles observed in adulthood date back compensate vis-a-vis proximity because their motive to seek
to infancy. As long as the patterns of attachment identified in proximity is often frustrated and rarely satisfied. Both Bowlby
children are phenotypically similar to those that characterize (1969) and Ainsworth et al. argue that rebuffed efforts to
adults, and as long as the consequences of these styles for behav- achieve proximity and frequent failure to achieve satisfactory
ior and emotions are similar across different developmental resolution of aroused proximity needs can create psychological
levels, attachment theory remains a viable model for under- conditions in which proximity needs are characteristically
standing adult relationships. Given this perspective, general more "available" or more easily aroused.
comparisons between children and adults may be informative, The null effects for the anxious style in connection with sup-
regardless of whether continuity exists within individuals over port seeking could be attributable to the lower reliability of the
time. anxious attachment index compared to the avoidant/secure
one. Three pieces ofevidence, however, cast doubt on this expla-
nation. First, null results still emerged for the anxious attach-
Support Seeking ment index across all analyses, even when effect sizes were
corrected for attenuation that was due to the measure's lower
More securely attached women in this study used their reliability (see Cohen & Cohen, 1983). Second, unlike previous
partners as a source of comfort and reassurance as their anxiety studies using categorical methods to measure anxious attach-
increased, whereas more avoidant women retracted from their ment (eg., Hazan & Shaver, 1987), in this study the continuous
partners both emotionally and physically. These results appear measure of anxious attachment did not suffer from low endorse-
to parallel those found with infants in Ainsworth et al.'s (1978) ment rates; nullresults,therefore, are not due to problems asso-
Strange Situation. When confronted with anxiety, securely at- ciated with restricted range. Third, at least one previous study
tached children actively seek out close physical and emotional using the same continuous measure of anxious attachment has
contact with their attachment figure until anxiety abates, after revealed that this index does correlate highly and reliably with a
which normal behavior resumes. Avoidant children, on the diverse array of external criteria (see Simpson, 1990). Hence,
other hand, avoid and retract from their attachmentfigureeven the anxious attachment index has demonstrated predictive va-
though anxiety remains high. lidity in past research. Taken together, these considerations
The findings presented in Figures 1 and 2 are interesting in suggest that, rather than being a statistical artifact, the null
part because they suggest that, at lower levels of anxiety, more findings for the anxious attachment index may indeed be valid.
avoidant women seek more support from their partners than Like avoidant persons, anxiously attached persons also suffer
do more secure women, and more avoidant men provide more from internal conflict with regard to proximity needs. They
support than do more secure men. As noted earlier, avoidant need and desire proximity yet fear they will notreceiveas much
persons are often thought of as cold, distant, and aloof. These of it as they desire because previous attachment figures have
444 J. SIMPSON, W RHOLES, AND J. NELLIGAN

been inconsistently available (Ainsworth et al., 1978). As a re- as partner anxiety increased. These findings corroborate and
sult, the need for proximity becomes associated with anger to- provide context for previous work which has shown that women
ward the attachmentfigure.In the avoidant style, internal con- describe secure men as being warm, self-disclosing, responsive,
flict typically leads to simple, unipolar avoidant behavior. In and good facilitators of communication in relationships (Col-
the anxious style, however, internal conflict is acted out more lins & Read, 1990). They also suggest that anxiety impedes the
directly in conflicted, contradictory behavior—at least among establishment of proximity in dyadic interactions involving
infants. When confronted with strong anxiety, they display avoidant persons, both when it is experienced directly and
marked conflict and ambivalence in the presence of their at- when it is experienced vicariously.
tachmentfigure,often blending contact-seeking behaviors with To date, relatively few studies have examined how attach-
angry, resistant, and dismissive ones (Ainsworth et al., 1978). ment styles affect support giving during spontaneous dyadic
Because anxiously attached adults strongly desire close, support- interaction. Two studies, however, have focused on support giv-
ive relationships (Collins & Read, 1990; Feeney & Noller, 1990; ing during mother-child interactions. Crowell and Feldman
Hazan & Shaver, 1987) but typically are not involved in them (1988) reported patterns of support giving among mothers that
(Collins & Read, 1990; Simpson, 1990), the behavior of highly are very similar to those observed among men in this study.
anxious people in the present study may reflect behavioral am- Specifically, mothers who possess a secure attachment style (as
bivalence in which tendencies to both approach and withdraw assessed by a clinical interview) provide their children with
from the partner effectively counterbalance one another. Fu- more support, help, and encouragement on difficult and stress-
ture research must determine whether the behavior of anxious ful tasks than do anxious/ambivalent and avoidant mothers.
persons in this setting reflects genuine ambivalence. Crowell and Feldman discovered a linear trend such that secure
mothers provided the most support, avoidant mothers offered
the least support, and anxious mothers demonstrated a level of
Reactions to Support support between these two groups. Although the paradigm and
Contrary to predictions, both more secure and more avoi- interactants used by Crowell and Feldman differ in many ways
dant women were calmed when their partners made supportive from ours, similar patterns of support giving appear to have
comments. In fact, avoidant women appeared to be somewhat emerged in these two different settings. Belsky, Rovine, and
more responsive to support than secure ones. This result is Taylor (1984) also reportfindingsthat parallel the present ones.
intriguing in light of the fact that avoidant women were less Mothers of avoidant infants, who themselves tend to be avoid-
likely to mention the stressful event to their partners. One possi- ant (Main et al., 1985), are highly involved with their infants
ble explanation for this unanticipated finding is that more avoi- under most circumstances, even more so than mothers of se-
dant women may choose or produce romantic partners who cure infants. Nonetheless, they respond much less supportively
routinely provide them with less frequent support. When sup- than mothers of secure infants when their infants are dis-
port is offered, however, it may have stronger and more positive tressed. In other words, the behaviors of these mothers are con-
impact on avoidant women than on secure ones. Regardless of tingent on their infants' emotional states in ways that are simi-
how it is interpreted, this finding clearly indicates that more lar to the contingencies seen among more avoidant men in the
avoidant persons do have proximity needs, for they appear to present study. These findings suggest that distressed infants
benefit from receiving support. and distressed adult partners present significant relationship
The extent to which the stressful event was discussed also was problems for avoidant persons.
associated with the degree to which more avoidant women were
calmed, independent of the behavior displayed by their Conclusions and Caveats
partners. Discussion alone, however, was not related to calming
among more secure women. These findings raise interesting Several recent studies have documented that securely at-
questions for future research about what kinds of support might tached persons have more committed, satisfying, interdepen-
be most effective for secure and avoidant persons. Among avoi- dent, and well-adjusted relationships relative to their avoidant
dant persons, support expressed verbally may be more effective and anxious counterparts (Collins & Read, 1990; Hendrick &
than support expressed through physical contact. For secure Hendrick, 1989; Kobak & Sceery, 1988; Levy & Davis, 1988;
persons, considerable support may be derived merely from the Simpson, 1990). Little, however, is known about what kinds of
knowledge that the partner is emotionally available and respon- specific patterns of interaction might generate these global ef-
sive if needed. fects. Although differences in relationship quality can stem
from many different factors (see Berscheid, 1985), the present
findings suggest that one precipitating factor may involve
Support Giving whether and to what extent romantic partners serve as mutual
As their partners' level of anxiety increased, more securely sources of comfort and support, particularly under stressful
attached men in this study offered greater reassurance and circumstances. Given that securely attached persons seek out
emotional support and made more supportive comments. and provide support when distress arises in their relationships,
More avoidant men did not show these same trends. The conver- these tendencies may facilitate and perhaps produce stronger
sational measures revealed only modest increases in supportive interdependence and positive affect. Because avoidantly at-
comments made by avoidant men as their female partners dis- tached persons fail to seek and offer such support, these pro-
cussed their feelings of distress more extensively, and the adjec- clivities may promote weaker interdependence and greater neg-
tive rating measures showed a decline in overall supportiveness ative affect. Future research must determine whether global
ATTACHMENT STYLES 445

impressions of relationship quality directly produce these dif- Cattell, R. B. (1966). The scree test for the number of factors. Multi-
ferent styles of interaction, whether these styles produce the variate Behavioral Research, 1, 245-276.
global impressions, or whether other variables mediate the two. Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation
This research advances previous work on adult attachment in analysis for the behavioral sciences (2nded.). Hillsdale, NJ: Erlbaum.
two important ways. First, it provides some of thefirstevidence Collins, N. L., & Read, S. J. (1990). Adult attachment, working models,
that adult attachment styles have clear and theoretically mean- and relationship quality in dating couples. Journal ofPersonality and
Social Psychology, 58, 644-663.
ingful effects on actual behavior during spontaneous dyadic in-
Crittenden, P. M., & Ainsworth, M. (1989). Child maltreatment and
teractions. Second, this investigation represents one of the first
attachment theory. In D. Cicchetti & Y Carlson (Eds.), Child mal-
attempts to examine whether and how attachment styles moder- treatment: Theory and research on the causes and consequences of
ate dyadic interaction under stressful conditions in adults. Prior child abuse and neglect (pp. 432-463). Cambridge, England: Cam-
research has focused almost entirely on how attachment styles bridge University Press.
in infants and children differentially affect behavior under Crowell, J. A., & Feldman, S. S. (1988). Mothers' internal models of
stressful conditions. relationships and childrens' behavioral and developmental status: A
The results of this study should be interpreted bearing three study of mother-child interaction. Child Development, 59, 1273-
caveats in mind. First, only women were subjected to anxiety, 1285.
and only men were given the opportunity to provide support in Elicker, J., & Sroufe, L. A. (in press). Predicting peer competence and
this investigation. Although there are no a priori reasons to peer relationships in childhood from early parent-child relation-
expect that our effects should be specific to the differing "roles" ships. In R. Parke & G. Ladd (Eds.), Family-peer relationships:
enacted by each sex, the findings for support seeking cannot Modes oflinkage. Hillsdale, NJ: Erlbaum.
necessarily be generalized to men, and those for support giving Erickson, M. E, Sroufe, L. A., & Egeland, B. (1985). The relationship
cannot be generalized to women. Second, despite the fact that between quality of attachment and behavior problems in preschool
in a high-risk sample. Monographs of the Society for Research in
strong theoreticalreasonsexist to believe that attachment styles
Child Development, 50(1 & 2), 147-166.
may have directly influenced the behaviors and interaction
Feeney, J. A., & Noller, P. (1990). Attachment style as a predictor of
styles we observed, the correlational nature of the data do not adult romanticrelationships.Journal of Personality and Social Psy-
permit unambiguous causal inferences. Finally, even though chology, 58, 281-291.
adults and children appear to exhibit similar patterns of sup- Hazan, C, & Shaver, P. (1987). Romantic love conceptualized as an
port seeking and support giving when distressed as a function attachment process. Journal of Personality and Social Psychology,
of their attachment style, this investigation does not and cannot 52,511-524.
address whether these patterns reflect the operation of attach- Hendrick, C, & Hendrick, S. S. (1989). Research on love: Does it mea-
ment styles that were first formed in early childhood. sure up? Journal ofPersonality and Social Psychology, 56, 784-794.
Jackson, D. N. (1971). The dynamics of structured personality tests:
1971. Psychological Review, 78, 229-248.
Knapp, M. L. (1978). Nonverbal communication in human interaction
References (2nd ed.). New York: Holt, Rinehart & Winston.
Kobak, R. R., & Sceery, A. (1988). Attachment in late adolescence:
Ainsworth, M. (1985). Attachments across the life span. Bulletin ofthe Working models, affect regulation, andrepresentationsof self and
New York Academy ofMedicine, 61, 792-812. others. Child Development, 59,135-146.
Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978). Patterns of Levy, M. B., & Davis, K. E. (1988). Lovestyles and attachment styles
attachment. Hillsdale, NJ: Erlbaum. compared: Their relations to each other and to various relationship
Belsky, J., Rovine, M., & Taylor, D. G. (1984). The Pennsylvania infant characteristics. Journal ofSocial and Personal Relationships, 5,439-
and family development project, III: The origins of individual dif-
471.
ferences in infant-mother attachment: Maternal and infant contri-
butions. Child Development, 55, 718-728. Lewis, M., Feiring, C, McGuffog, C, & Jaskir, J. (1984). Predicting
psychopathology in six-year-olds from early social relations. Child
Berscheid, E. (1985). Interpersonal attraction. In G. Lindzey & E. Ar-
onson (Eds.), Handbook ofsocial psychology (3rd ed., pp. 413-484). Development, 55,123-136.
New York: Random House. Loehlin, J. C. (1990). Component analysis versus common factor analy-
Berscheid, E., Snyder, M., & Omoto, A. M. (1989). The Relationship sis: A case of disputed authorship. Multivariate Behavioral Research,
Closeness Inventory: Assessing the closeness of interpersonal rela- 25,29-31.
tionships. Journal ofPersonality and Social Psychology, 57,792-807. Loevinger, J. (1957). Objective tests as instruments of psychological
Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York: theory. Psychological Reports, 3, 635-694.
Basic Books. Londerville, S., & Main, M. (1981). Security of attachment, compli-
Bowlby, J. (1973). Attachment and loss: Vol. 2. Separation: Anxiety and ance, and maternal training methods in the second year oflife. Devel-
anger. New York: Basic Books. opmental Psychology, 17, 289-299.
Bowlby, J. (1979). The making and breaking of affectional bonds. Lon- Main, M., & Goldwyn, R. (1984). Predicting rejection of her infant
don: Tavistock. from mother's representation of her own experience: Implications
Bowlby, J. (1980). Attachment and loss: Vol. 3. Loss: Sadness and depres- for the abused-abusing intergenerational cycle. Child Abuse and Ne-
sion. New York: Basic Books. glect, 8, 203-217.
Bretherton, I. (1985). Attachment theory: Retrospect and prospect. In Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, child-
I. Bretherton & E. Waters (Eds.), Growing points of attachment hood, and adulthood: A move to the level of representation. Mono-
theory and research. Monographs ofthe Society for Research in Child graphs of the Society for Research in Child Development, 50(1 & 2),
Development, 50(1 & 2), 3-35. 66-104.
446 J. SIMPSON, W RHOLES, AND J. NELLIGAN

Rubin, Z. (1970). Measurement of romantic love. Journal ofPersonality Velicher, W E, & Jackson, D. N. (1990). Component analysis versus
and Social Psychology, 16, 265-273. common factor analysis: Some issues in selecting an appropriate
Simpson, J. A. (1990). Influence of attachment styles on romantic rela- procedure. Multivariate Behavioral Research, 25,1-28.
tionships. Journal ofPersonality and Social Psychology, 59,971-980. Waters, E., Wippman, J., & Sroufe, L. A. (1979). Attachment, positive
Sroufe, L. A. (1983). Infant-caregiver attachment and patterns of adap- affect, and competence in the peer group: Two studies in construct
tation in preschool: The roots of maladaptation and competence. In validation. Child Development, 50, 821-829.
M. Perlmutter (Ed.), Minnesota Symposium on Child Psychology West, S. G., & Finch, J. E (in press). Measurement and analysis issues in
(Vol. 16, pp. 41-83). Hillsdale, NJ: Erlbaum. the investigation of personality structure. In S. Briggs, R. Hogan, &
Sroufe, L. A., Egeland, B., & Kreutzer, T. (1990). The fate of early W Jones (Eds.), Handbook ofpersonality psychology. San Diego, CA:
experience following developmental change: Longitudinal ap- Academic Press.
proaches to individual adaptation in childhood. Child Development,
61, 1363-1373.
Tinsley, H. E. A., & Weiss, D. J. (1975). Interrater reliability and agree- Received December 11,1990
ment of subjective judgments. Journal ofCounseling Psychology, 22, Revision received October 10,1991
358-376. Accepted October 23,1991 •

Neuropsychology to Be an APA Journal

In January 1993, Neuropsychology, which has been published by the Educational


Publishing Foundation (a subsidiary publishing program of the American Psycho-
logical Association), will be published by the American Psychological Association.
The Publications and Communications Board of the APA has appointed Nelson
Butters as editor of Neuropsychology. As of January 1,1992, manuscripts should be
submitted to
Nelson Butters
Chief, Psychology Service (116B)
Department of Veterans Affairs Medical Center
3350 La Jolla Village Drive
La Jolla.CA 92161

Manuscripts considered by the incoming editor will be published beginning in the


January 1993 issue. Submitted manuscripts should fall within the following new
editorial policy statement:

The mission of Neuropsychology is to foster (a) basic research, (b) the


integration of basic and applied research, and (c) improved practice in the field
of neuropsychology, broadly conceived. The primary function of Neuropsy-
chology is to publish original, empirical papers in the field. Occasionally,
scholarly reviews and theoretical papers will also be published—all with the
goal of promoting empirical research on the relation between brain and human
cognitive, emotional, and behavioral function. Sought are submissions of
human experimental, cognitive, and behavioral research with implications for
neuropsychological theory and practice. Papers that increase our understand-
ing of neuropsychological functions in both normal and disordered states and
across the lifespan are encouraged. Applied, clinical research that will
stimulate systematic experimental, cognitive, and behavioral investigations as
well as improve the effectiveness, range, and depth of application is germane.
Neuropsychology seeks to be the vehicle for the best research and ideas in the
field.

View publication stats

S-ar putea să vă placă și