Citation
Quality of attachment

Material Information

Title:
Quality of attachment stability and relationship to family environmental variables
Creator:
Isleib, Roberta Ann, 1953-
Publication Date:
Language:
English
Physical Description:
viii, 75 leaves : ; 29 cm.

Subjects

Subjects / Keywords:
Attachment behavior ( jstor )
Child development ( jstor )
Classification systems ( jstor )
Developmental psychology ( jstor )
Employment ( jstor )
Infants ( jstor )
Mothers ( jstor )
Psychological stress ( jstor )
Temperament ( jstor )
Toys ( jstor )
Child Development ( mesh )
Clinical and Health Psychology thesis Ph.D ( mesh )
Dissertations, Academic -- Clinical and Health Psychology -- UF ( mesh )
Parent-Child Relations ( mesh )
Social Adjustment -- Child ( mesh )
Social Adjustment -- Infant ( mesh )
Socialization ( mesh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 1985.
Bibliography:
Bibliography: leaves 69-74.
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Roberta Ann Isleib.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
023216839 ( ALEPH )
16996581 ( OCLC )
ACW5311 ( NOTIS )

Downloads

This item has the following downloads:

ERPQQY01L_WHYM8T.xml

AA00011823_00001.pdf

AA00011823_00001_0081.txt

qualityofattachm00isle_0027.txt

AA00011823_00001_0070.txt

AA00011823_00001_0039.txt

qualityofattachm00isle_0041.txt

qualityofattachm00isle_0017.txt

qualityofattachm00isle_0039.txt

AA00011823_00001_0046.txt

qualityofattachm00isle_0014.txt

AA00011823_00001_0008.txt

qualityofattachm00isle_0081.txt

qualityofattachm00isle_0012.txt

qualityofattachm00isle_0068.txt

qualityofattachm00isle_0023.txt

qualityofattachm00isle_0001.txt

qualityofattachm00isle_0031.txt

qualityofattachm00isle_0061.txt

qualityofattachm00isle_0078.txt

AA00011823_00001_0065.txt

AA00011823_00001_0053.txt

AA00011823_00001_0006.txt

qualityofattachm00isle_0042.txt

AA00011823_00001_0064.txt

qualityofattachm00isle_0022.txt

qualityofattachm00isle_0010.txt

AA00011823_00001_0007.txt

AA00011823_00001_0073.txt

qualityofattachm00isle_0005.txt

qualityofattachm00isle_0009.txt

qualityofattachm00isle_0050.txt

qualityofattachm00isle_0043.txt

AA00011823_00001_0034.txt

qualityofattachm00isle_0003.txt

qualityofattachm00isle_0052.txt

qualityofattachm00isle_0006.txt

qualityofattachm00isle_0028.txt

AA00011823_00001_0037.txt

qualityofattachm00isle_0049.txt

AA00011823_00001_0071.txt

AA00011823_00001_0019.txt

AA00011823_00001_0045.txt

qualityofattachm00isle_0076.txt

AA00011823_00001_0020.txt

AA00011823_00001_0022.txt

qualityofattachm00isle_0058.txt

AA00011823_00001_0002.txt

qualityofattachm00isle_0018.txt

qualityofattachm00isle_0036.txt

AA00011823_00001_0001.txt

AA00011823_00001_0063.txt

qualityofattachm00isle_0030.txt

qualityofattachm00isle_0072.txt

qualityofattachm00isle_0004.txt

AA00011823_00001_0011.txt

AA00011823_00001_0050.txt

qualityofattachm00isle_0026.txt

qualityofattachm00isle_0029.txt

AA00011823_00001_0026.txt

AA00011823_00001_0062.txt

qualityofattachm00isle_0013.txt

AA00011823_00001_0030.txt

qualityofattachm00isle_0056.txt

qualityofattachm00isle_0055.txt

AA00011823_00001_0043.txt

qualityofattachm00isle_0008.txt

AA00011823_00001_0076.txt

qualityofattachm00isle_0054.txt

AA00011823_00001_0040.txt

qualityofattachm00isle_0057.txt

AA00011823_00001_0033.txt

AA00011823_00001_0013.txt

AA00011823_00001_0038.txt

AA00011823_00001_0077.txt

AA00011823_00001_0025.txt

qualityofattachm00isle_0046.txt

AA00011823_00001_0028.txt

AA00011823_00001_0016.txt

qualityofattachm00isle_0077.txt

qualityofattachm00isle_0060.txt

AA00011823_00001_0027.txt

AA00011823_00001_0012.txt

qualityofattachm00isle_0083.txt

qualityofattachm00isle_0019.txt

AA00011823_00001_0032.txt

AA00011823_00001_0068.txt

qualityofattachm00isle_0063.txt

qualityofattachm00isle_0015.txt

AA00011823_00001_0018.txt

AA00011823_00001_0086.txt

AA00011823_00001_0031.txt

qualityofattachm00isle_0073.txt

qualityofattachm00isle_0011.txt

AA00011823_00001_0084.txt

qualityofattachm00isle_0040.txt

AA00011823_00001_0014.txt

AA00011823_00001_0015.txt

qualityofattachm00isle_0074.txt

AA00011823_00001_0061.txt

AA00011823_00001_0044.txt

AA00011823_00001_0042.txt

qualityofattachm00isle_0002.txt

qualityofattachm00isle_0047.txt

AA00011823_00001_0017.txt

qualityofattachm00isle_0067.txt

ERPQQY01L_WHYM8T_xml.txt

qualityofattachm00isle_0062.txt

qualityofattachm00isle_0032.txt

qualityofattachm00isle_0044.txt

qualityofattachm00isle_0053.txt

AA00011823_00001_0041.txt

AA00011823_00001_0085.txt

AA00011823_00001_0079.txt

AA00011823_00001_0080.txt

AA00011823_00001_0059.txt

qualityofattachm00isle_0035.txt

AA00011823_00001_0048.txt

qualityofattachm00isle_0079.txt

qualityofattachm00isle_0033.txt

AA00011823_00001_0054.txt

AA00011823_00001_0029.txt

AA00011823_00001_0047.txt

qualityofattachm00isle_0069.txt

qualityofattachm00isle_0066.txt

qualityofattachm00isle_0065.txt

AA00011823_00001_0009.txt

qualityofattachm00isle_0059.txt

AA00011823_00001_0060.txt

AA00011823_00001_0074.txt

qualityofattachm00isle_0064.txt

qualityofattachm00isle_0020.txt

qualityofattachm00isle_0045.txt

qualityofattachm00isle_0007.txt

qualityofattachm00isle_0037.txt

AA00011823_00001_0066.txt

qualityofattachm00isle_0070.txt

qualityofattachm00isle_0051.txt

qualityofattachm00isle_0034.txt

AA00011823_00001_0021.txt

AA00011823_00001_pdf.txt

AA00011823_00001_0069.txt

AA00011823_00001_0057.txt

qualityofattachm00isle_0021.txt

AA00011823_00001_0024.txt

AA00011823_00001_0004.txt

qualityofattachm00isle_0048.txt

AA00011823_00001_0049.txt

AA00011823_00001_0083.txt

AA00011823_00001_0067.txt

AA00011823_00001_0078.txt

AA00011823_00001_0056.txt

qualityofattachm00isle_0080.txt

AA00011823_00001_0051.txt

qualityofattachm00isle_0075.txt

qualityofattachm00isle_0038.txt

AA00011823_00001_0035.txt

AA00011823_00001_0003.txt

qualityofattachm00isle_0024.txt

qualityofattachm00isle_0000.txt

AA00011823_00001_0010.txt

qualityofattachm00isle_0071.txt

AA00011823_00001_0052.txt

AA00011823_00001_0023.txt

AA00011823_00001_0075.txt

qualityofattachm00isle_0025.txt

AA00011823_00001_0036.txt

qualityofattachm00isle_0016.txt

qualityofattachm00isle_0084.txt

qualityofattachm00isle_0085.txt

AA00011823_00001_0055.txt

AA00011823_00001_0082.txt

AA00011823_00001_0072.txt

qualityofattachm00isle_0082.txt

AA00011823_00001_0005.txt

AA00011823_00001_0058.txt


Full Text


















QUALITY OF ATTACHMENT:
STABILITY AND RELATIONSHIP TO FAMILY ENVIRONMENTAL VARIABLES








BY

ROBERTA ANN ISLEIB








A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA



1985




































To ny father

Charles Robert Isleib
















ACKNOWLEDGMENTS


The planning and execution of this manuscript have been

extremely complex and demanding. It could not have been

completed without the assistance of many people along the way.

Thanks go first to Suzanne Johnson, Roger Blashfield, Jim

Johnson, Pat Miller, Drew Bradlyn, and Randy Carter for their

supportive and constructive comments during the designing

stages of the study. Suzanne also served as an anchor

throughout my internship year, encouraging me to continue to

make progress in spite of various setbacks and the pull to

procrastinate.

Identifying and recruiting subjects was a major hurdle,

for which the input of Jim Siwi, Keith Berg, and Janice Benton

is gratefully acknowledged. Thanks also go to the many

friends who told others about my research and recruited

subjects by word of mouth. Prior to data collection, Jerry

Rudd contributed skill and labor while installing a one-way

mirror and Dan McNeil served as a cheerful video equipment

consultant.

Ross Thompson kindly reviewed several strange situation

videotapes and made helpful suggestions about the physical

environment, stranger behavior, and timing of the episodes.

Fi Kieffer and her son Christopher, and Dawn Bowers and her

daughter Martina agreed to serve as guinea pigs during the












initial training period. Special thanks are due to Penny

Phelps, Joy Colle, and Katie Kato, who played the difficult

role of the stranger. In addition, Rae Hendlin generously

filled in for one session on a last minute basis, and Roman

Urbana served as an experimenter throughout the study.

Scoring the videotapes proved to be an enormous task,

made possible initially by the training provided by Everett

Waters. Thanks are also due to Mary Jo Ward for her

reliability checks, and most certainly, to Susan Grajek,

without whom the data analysis would have seemed overwhelming.

Don Quinlan kindly allowed me access to his computer, such

that typing and editing the manuscript became manageable.

My most earnest thanks go to my father, C. Robert Isleib,

who painstakingly coded a seemingly endless number of

videotapes, and tolerated my irritability with equanimity in

times when the project lagged. This dissertation is dedicated

to him, not only for his invaluable assistance during this

year, but also in recognition of the unselfish support and

encouragement he has given me in many ways over the past 32

years.

Roberta A. Isleib
New Haven, CT.
September, 1985


















TABLE OF CONTENTS


Page

ACKNOWLEDGMENTS............... .........................iii

LIST OF TABLES ....... .... .............. ............... vi

ABSTRACT..... .......... ................................vii

INTRODUCTION..... ........... .......................... 1

The Bowlby-Ainsworth Theory of Attachment.......... 1
The Strange Situation............................... 4
Predictive Validity of the ABC Typology............ 11
Stability of the ABC Classification System.......... 13
Effects of Family Environmental Variables........... 24
Aims and Hypotheses of the Study................... 29

METHOD.... .... ... ....................................... 31

Subjects................ .. .......................... .. 31
Materials.......................................... 31
Procedure.............. ........................ 35

RESULTS............................................... .. 38

Reliability......... .............. ...... ......... 38
Stability......................................... 38
Family Environmental Variables..................... 50

DISCUSSION.. ...... ... ...... .......................... 51

APPENDIX A INFORMED CONSENT........................... 62

APPENDIX B STRUCTURED INTERVIEW....................... 65

APPENDIX C INSTRUCTIONS TO PARENTS.................... 67

REFERENCES.... .... ...... .............................. 69

BIOGRAPHICAL SKETCH.. .................................. 75

v
















LIST OF TABLES


Page

TABLE 1 Classification of Infants at Test and Retest
in Same Context/Changed Context Groups.......... 43

TABLE 2 Classification of Infants at Test and Retest
for the Current Study and Ainsworth's Sample... 45

TABLE 3 Subgroup Classification at Test and Retest..... 46

TABLE 4 Means and Standard Deviations for Interactive
Ratings............ ...................... ..... 49

TABLE 5 Comparison of Means of Strange Situation
Measures for Sessions One and Two and
Correlations Between Sessions.................. 50
















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


QUALITY OF ATTACHMENT:
STABILITY AND RELATIONSHIP TO FAMILY ENVIRONMENTAL VARIABLES

By

Roberta Ann Isleib

December 1985


Chairman: Suzanne Bennett Johnson
Major Department: Clinical Psychology

The present study tested the hypothesis that the stability

of attachment classification would be increased if an infant's

recognition of the assessment context was eliminated. Twenty-

nine white, middle-class mothers and their one-year-old

infants were seen twice in the "strange situation" and

classified according to Ainsworth's ABC taxonomy. Half of the

infants were seen two times in the same setting, with the

others seen in two very different settings, presented in

counterbalanced order. It was hypothesized that controlling

for context-specific stress would decrease the occurrence of

intensified attachment behavior in session two. However, no

differences between these groups were found in number of

infants who changed classification from test to retest, or in

four interactive ratings.

Given these findings, the two groups were collapsed in

order to examine short-term stability in the group as a whole.
vii.











Sixty-eight percent of the children retained their

classification over time (kappa = -0.116). Next, the

association between short-term stability of infant-mother

attachment and a number of family environmental and

demographic variables was assessed. No significant

differences between retained versus changed classification

subjects emerged on the following variables: social support,

life stress, maternal temperament, infant temperament,

caretaking changes, separations from mother over the first

year, infant sex, birth order, parental age, and maternal

work status.

The second aim of the study was to investigate the

relationship between quality of attachment and family

environmental variables. Using the classifications from the

first session only, the securely attached infants (n = 25)

were compared to those children who were judged anxiously

attached (n = 5). There were no differences on the variables

listed above, with the exception of maternal activity level on

which mothers of securely attached infants scored lower than

mothers of anxiously attached children.

These results are interpreted as lack of support for the

contention that short-term instability in the ABC system can

be explained as an effect of returning to the same setting

twice. The overall low stability in the entire sample is

presented as evidence that research should be directed at

exploring other models for describing and quantifying quality

of mother-infant attachment.
viii















INTRODUCTION


In 1958, Bowlby published an article entitled "The Nature

of the Child's Tie to his Mother" which was the first

exposition of the now widely recognized Bowlby-Ainsworth

ethological theory of attachment. Although many other

theories of attachment have been proposed (e.g., Cairns,

1966; Hoffman & DePaulo, 1977; Gewirtz, 1972), none have

generated as much empirical research or received as much

support from empirical evaluation, as has the Bowlby-Ainsworth

theory (Rajecki, Lamb, & Obmascher, 1978). Some of the focus

on this theory may be attributed to methodological

consistency, in that the "strange situation" paradigm

developed by Ainsworth and colleagues (Ainsworth, Blehar,

Waters, & Wall, 1978) has been widely adopted as the

attachment classification system.

The Bowlby-Ainsworth Theory of Attachment

In developing his theory of attachment, Bowlby drew on

ethological notions of instinctive behavior, control systems

theory, and Freud's concept of object relations, seeking an

explanatory framework for naturalistic observations of

children who had been separated from their mother-figures

(Bowlby, 1958; 1969; 1973). Previous research (e.g.,

Robertson, 1952) had noted that hospitalized children

progressed through predictable stages following separation,










characterized by active protest, despair or hopelessness, and

finally, detachment and apathy. As Ainsworth has been

instrumental in translating Bowlby's ideas into research, as

well as contributing to their elaboration and clarification

(e.g., Ainsworth, 1964; 1973; 1979a), the theory has become

known as the Bowlby-Ainsworth ethological model of attachment.

The Bowlby-Ainsworth model views attachment as a species

general behavior system whose function is to protect the

relatively helpless human infant from predation. Infants are

born with a biological predisposition to seek contact with

adults, which is manifested early on by signalling behaviors

such as crying, smiling, and calling, in addition to searching

and following behaviors as locomotion develops. Adults are

equipped with complementary predispositions to protect

infants. Bowlby originally suggested that the development of

a specific attachment was mediated by the amount of

interaction occurring between the infant and attachment

figure. As a cognitive appreciation for person permanence was

assumed to be a necessary prerequisite, the development of

attachment was not expected to occur until the second half of

the first year of life.

Drawing on control systems theory, Bowlby incorporated

the notion of a set-goal into his attachment model. The set-

goal, defined as the amount of distance from the attachment

figure which the infant will tolerate, is mediated by a

variety of internal (e.g., illness, hunger) and external

(e.g., stress, separation) factors. When the set-goal is

exceeded, the infant should respond with the activation of










attachment behaviors such as signalling or following, which

serve to bring the attachment figure into closer proximity.

Ainsworth (1979b) has suggested that "sensitive

responsiveness" is the quality most likely to foster secure

attachment. She maintained that if the mother-figure

responded appropriately to the infant's signals across the

first year of life, the infant would develop a working model

of the mother as responsive, reliable, and accessible. She

further emphasized that in addition to protection from

predation, attachment facilitates the use of the mother-figure

as a secure base for exploration of the environment. In other

words, an infant who feels confident of the accessibility of

the caretaker is able to expand the range of the set-goal and

engage in adaptive play, exploration, and problem-solving.

Sroufe and Waters (1977) have elaborated on Ainsworth's

conception of attachment, viewing it as an affective tie

between mother and infant, as well as a behavioral system.

Differences in attachment should theoretically be

distinguishable through observation of qualitative differences

in organization of behavior, rather than discrete behaviors.

As many so-called attachment behaviors are functionally

interchangeable, attachment patterns can only be assessed "by

reference to the organization of attachment behaviors with

respect to the caregiver and in consideration of context"

(Sroufe & Waters, 1977, p. 1188). These authors defined the

term affective bond as the infant's experience of security-

insecurity, which is influenced by a variety of factors such

as context, preceding events, and developmental level. The










degree of security which the infant experiences mediates both

proximity-seeking and exploration.

The Strange Situation

Ainsworth and her colleagues (Ainsworth & Wittig, 1969;

Ainsworth, Bell, & Stayton, 1971, 1972; Ainsworth, Blehar,

Waters, & Wall, 1978) have developed a standardized laboratory

paradigm to assess patterns of attachment in one to two year

old infants, called the "strange situation." The procedure

involves the infant, the mother, and a stranger, participating

in eight increasingly stressful episodes in which exploratory

behavior, response to the mother's absence, response to the

stranger, and response to reunion are observed. In episode 1,

the mother and baby are introduced to the experimental room by

an observer, who disappears shortly after this introduction.

During episode 2, the mother is instructed to attempt to

interest the infant in the available toys, and to refrain from

interactions other than responding to the infant's overtures.

Episode 3 involves the first appearance of the stranger, who

sits quietly, then chats with the mother, then becomes more

involved with the child. Finally, the mother leaves the room.

In episode 4, the stranger attempts to engage the baby in

interaction, intervening if the baby becomes distressed at the

mother's absence by distracting him with a toy, talking to

him, or picking him up. The primary behaviors of interest are

the amount of exploration that the baby engages in, as

compared with that occurring when the mother was present, as

well as reaction to separation (e.g., search behaviors,

crying), and response to the stranger. Episode 5 consists of










the mother's return to the room and the stranger's departure.

Of particular interest are the baby's response to the mother

upon her return and their subsequent interaction. When the

infant is sufficiently calm, the mother again leaves the room,

leaving the child alone for episode 6. Again, search

behaviors, vocalizations, and crying are noted. In episode 7,

the stranger renters the room and attempts to comfort and

reengage the infant in play. The baby's response to the

stranger is compared with his interactions with the mother in

previous reunion episodes. Finally, the mother returns,

pausing long enough to allow the baby time for spontaneous

reaction, and the stranger leaves. Each of the episodes lasts

approximately three minutes, with the exception of episode 1,

which lasts approximately 30 seconds. In early research,

trained observers viewed the process of the strange situation

through one-way mirrors and dictated running narratives

throughout, concentrating on the behaviors of the babies and

their interactions with both mother and stranger. The strange

situation was initially developed and standardized on four

samples, consisting of a total of 106 white, middle-class

subjects from the Baltimore area (Ainsworth et al., 1978).

Two observers were used for all but one of the studies, with

both narratives utilized in coding data. Four types of

measures were drawn from the strange situation narratives.

These included frequency counts of specific behaviors, number

of 15 second intervals in which behaviors occurred, scoring of

interactive behavioral variables along a seven point

continuum, and classification of infants into three groups and










eight subgroups, according to behavioral similarities and

differences.

Coding in these studies was undertaken by at least two

independent coders who had been carefully trained.

Reliability checks conducted on eight subjects in the second

study revealed reliability coefficients above 0.93 for four

frequency measures. Although frequency counts and interval

estimations were thought to describe attachment behaviors, the

authors felt they did not adequately represent interactional

behavior. Also, since a variety of attachment behaviors are

theoretically interchangeable in terms of functional

significance, frequency counts of single behaviors might

inaccurately reflect the activation of the infant's attachment

system. Furthermore, as the behavior of the mothers could not

be tightly controlled, a system which could be sensitive to

comparisons across situations in spite of individual

differences needed to be devised. Scores of interactive

behavior were designed with these concerns in mind. Six

behavioral variables were described by seven-point scales

which were anchored by behavioral descriptions garnered from

the actual narrative reports of the subjects in samples one

and two. These variables included proximity and contact

seeking, contact maintaining, resistance, avoidance, search,

and distance interaction. Five reliability checks from all of

the samples revealed interscorer reliability coefficients

ranging from 0.75 to 0.97, with a median value of 0.93.

Finally, a classification system was devised by grouping

infants according to their behavioral similarities in as many










of the strange situation episodes as possible. Three major

patterns of behavior were classified. Group A infants,

dubbed anxiously-avoidant, were characterized by a conspicuous

avoidance of interaction with their mothers in the reunion

episodes. Although they tended not to seek proximity with

their mothers, they did not demonstrate resistance to contact.

The stranger was treated similarly to the mother, and minimal

distress was evidenced during episodes of separation. Group

B, the securely attached infants, was characterized by active

proximity-seeking, active seeking of interaction with the

mother during reunion, clear preference for the mother over

the stranger, and, frequently, evidence of distress at the

mother's absence. Group C babies, called ambivalently or

resistantly anxious, displayed contact and interaction

resistance, especially during the final reunion episode.

Resistance was generally manifested by striking out or

squirming away from an available adult or by rejecting

proffered toys. They also tended to seek proximity and

maintain contact, giving an ambivalent impression. In

general, these infants appeared to behave in more maladaptive

ways (either angry or passive) than did infants classified

into the other two groups. Tests of interjudge agreement

suggested that experienced judges identified infants highly

reliably according to the overall groups. Identification of

Group C infants appeared to present the most difficulty to the

judges.

Eight subgroups were also formed from the original three.

As these have been used in subsequent research, they will be










briefly described. Infants in subgroup Al displayed the most

conspicuous avoidance of their mothers, in association with

minimal proximity-seeking and contact-maintaining. Subgroup

A2 infants demonstrated a mixed response consisting of both

approach and avoidance. Of the four B subgroups, subgroup B1

babies showed the most distance interaction with their

mothers, and tended not to be distressed during separation.

B2 infants resembled those in group B1, except for an increase

in proximity-seeking. In subgroup B3, the infants were most

likely to seek physical contact with their mothers, while

demonstrating very little avoidance or resistance. (Subgroup

B3 contained the largest number of infants from the

development sample and was considered by the authors to be

the normative group, reflecting the most harmonious mother-

infant relationship.) Subgroup B4 infants appeared preoccupied

with their mothers throughout the strange situation, seeking

contact and proximity, but in a less effective fashion than

B3 babies. Subgroup C1 infants were distinguishable by their

angry appearance, marked both by strong attempts to gain

contact and conspicuous resistant behaviors. Finally, C2

babies appeared passive, involving themselves in limited

exploration and interaction, and using signalling behaviors to

gain proximity to their mothers. Their resistant behaviors

were not as visible as C1 infants. Although the

classification of infants into subgroups has not been

consistently reliable, Ainsworth et al. have emphasized that

instructing raters to attend to these detailed criteria

improved the overall reliability of the three group











classification. Also, the infants in the different subgroups

are hypothesized to have experienced different patterns of

caregiving, such that their attachment differs qualitatively.

In order to test the significance of the multivariate

differences between the three major categories of attachment,

a multiple discriminant function analysis was conducted. This

was also intended to explore whether the variables identified

in the classification instructions were those that best

discriminated the groups. The 73 original variables were

reduced to 22 by eliminating those which did not distinguish

between two groups at the 0.01 level of significance (using

analyses of variance), and those which did not contribute

significantly to 2-group discriminant function analyses (A vs.

B, and B vs. C). One hundred and five subjects were utilized

in the development sample, with two significant discriminant

functions emerging. The discriminant scores successfully

predicted classification of 96% of A infants, 91% of B

infants, and 92% of C infants. A serial cross-validation was

then conducted using unknown subjects from the development

sample. In this procedure, 92% of A infants, 85% of B

infants, and 69% of C infants were correctly classified.

The first discriminant function served to distinguish A

from non-A infants, and the strongest contributing variables

included proximity-seeking and contact-maintaining to the

mother, absence of resistance, interactive behavior with the

stranger, and exploratory behavior and crying. The second

discriminant function, labeled C vs. non-C, correlated highly

with resistance to mother, proximity- and contact-seeking,











resistance to the stranger, crying, and reduced exploratory

behavior. The authors concluded that these analyses validated

the utility of their classification system, although the

discrimination of C from non-C infants was most troublesome,

based on the low percentage of correct classification on cross

validation.

The multiple discriminant function analysis described

above suffers from several statistical flaws, and thus the

conclusion drawn by Ainsworth et al. regarding its validation

of their typology is questionable. First, as pointed out by

Connell and Goldsmith (1982), the variables entered into the

analysis were selected on the basis of their previously

identified ability to discriminate between the groups.

Therefore, the discriminant analysis was guaranteed to make

this differentiation and should have correctly classified

close to 100% of the sample. Second, the number of subjects

used in the analysis was considerably smaller than optimal,

considering the total number of variables involved. Given the

small number of subjects utilized, considerable instability of

the results might be expected, and less success predicted if a

replication was performed using a new data set. Finally, the

serial cross-validation procedure used by the authors was

designed to derive, rather than test, the functions, and was

therefore not adequate to assess their value in predicting

group membership for a new sample.

Predictive Validity of the ABC Typology

The long-range utility of the ABC typology depends on its

ability to differentially predict concurrent and future











performance in other arenas of infant competence. Studies

which have been conducted to evaluate the predictive validity

of the ABC classification system have uniformly supported the

more adequate performance of securely attached children on a

variety of social interaction and competence measures. Secure

infants (group B) obtained higher scores than anxious infants

(A and C) on measures of peer competence (Lieberman, 1977;

Pastor, 1981; Waters, Wippman, & Sroufe, 1979), imaginative

play (Matas, Arend, & Sroufe, 1978; Belsky, Garduque, &

Hrncir, 1984), ego resiliency (Arend, Gove, & Sroufe, 1979),

compliance and cooperation (Londerville & Main, 1981),

stranger sociability (Main & Weston, 1981), negotiation of the

environment (Cassidy & Ainsworth, 1983), and toleration of

brief separations at 18 months (Jacobson & Wille, 1984). Two

studies utilized subgroup classifications and found that B12

infants were more independent and interactive during play with

peers than either B3 or B4 children (Easterbrooks & Lamb,

1979), as well as more social with a friendly stranger than

both B34 and C subjects (Thompson & Lamb, 1983). Although

Jacobson, Wille, Tianen, and Aytch (1983) found ambivalent

infants to be more interactive with peers, they interpreted

this as the lack of expected or age-appropriate wariness. In

a longitudinal study, high-risk preschool children who had

maintained stable attachment classifications from 12 to 18

months were independently rated for behavior problems

(Erickson, Sroufe, & Egeland, in press). Group C infants were

rated as being less confident and having poorer social skills










than secure children, and Group A infants were more dependent

and less socially skilled. In general, anxiously attached

children as a group were more likely to be rated by observers

and teachers as having a variety of behavior problems.

In spite of the impressive global consistency of this

research, a major question has been raised which remains to be

addressed. In their critique of the strange situation

classification system, Connell and Goldsmith (1982) noted that

many predictive validity studies lumped ambivalent and

avoidant children into one large "insecure" category, when no

behavioral differences emerged between the groups on dependent

measures. In contrast, clear differences between the groups

did emerge in patterns of behavior in the strange situation.

Only 25% of the studies referred to above found differences

along the lines of the ABC classification, and none of these

were consistently interpretable based on theoretical

predictions. On the other hand, several authors have found

differences between subgroups (Easterbrooks & Lamb, 1979;

Thompson & Lamb, 1983). Given the difficulty researchers have

had in finding behavioral differences between the main groups,

these finer distinctions are puzzling. These inconsistencies

may in fact be reflecting inadequacies in the classification

system, rather than a lack of differences in the children.

Alternatively, differences between the three groups which are

clear in early life (e.g., Ainsworth et al., 1978; Belsky,

Rovine, & Taylor, 1984), may be moderated by later

environmental factors such that the ABC classification loses











predictive power. One aspect of the system which has been the

subject of some contention is stability of classification over

time. Research addressing this issue will be reviewed in the

following section.

Stability of the ABC Classification System

Much of the predictive validity of the ABC classification

system rests on claims for its demonstrated stability across

time. If infants cannot be consistently identified as

belonging to a particular group, the subsequent prediction

from group membership to their performance on a variety of

developmental tasks has little meaning.

In response to a critique by Masters and Wellman (1974)

suggesting that individual attachment behaviors had not been

found to be stable across time, Ainsworth et al. (1978)

assessed the reliability of the strange situation using a two

week test-retest procedure. They predicted that infants would

remember the first separation experience, and that distressed

infants would demonstrate more distress in the second session,

while undisturbed infants would appear even more comfortable

in the second assessment. In spite of these expected

behavioral changes from first to second assessment,

interactive behaviors were predicted to be significantly

correlated across sessions. Twenty-three white, middle-class

infants were involved in this test-retest study, conducted

when they were 50 and 52 weeks of age.

Overall, attachment behaviors increased significantly in

21 of 23 infants from the first to second session, including











proximity-seeking, contact-maintaining, search for mother, and

crying. Avoidance behaviors decreased in the second session

while resistance remained the same across time. In

interactions with the mother, proximity-seeking, contact-

maintaining, crying, searching, and resisting the stranger

were positively correlated across the two sessions, while

avoidance behaviors were not. As expected, frequency measures

of discrete behaviors such as smiling, vocalizing, and

touching did not demonstrate significant stability. The

overall stability of the ABC classification was 57% (kappa =

0.054), with the major changes accounted for by the switch of

all infants classified as Group A in the first session to

Group B in the second session. More specifically, ten of the

23 infants tested changed attachment groups upon retesting.

As Connell and Goldsmith (1982) have noted, the percentage of

correct reclassification expected by chance is 54%, which the

test-retest results only marginally exceeded. Ainsworth et al.

maintained that the differences in classification reflected

the oversensitivity of the system to short-term changes in the

infants' behavioral patterns (induced by the stress of

retesting), rather than reflecting inconsistencies in or

instability of individual differences in the organization of

attachment relationships. This explanation has apparently

been accepted by succeeding researchers, who have concentrated

later efforts on investigating the stability of the system

over longer intervals.










Only two other studies have repeated the strange

situation over a similarly short time interval. These results

will be briefly summarized, although both studies investigated

attachment towards mothers versus fathers, and neither

followed the Ainsworth scoring system faithfully. Willemsen,

Flaherty, Heaton, and Ritchey (1974) contrasted attachment

behaviors in one-year-old infants to mothers versus fathers.

These authors found more attachment behavior was directed to

the parent in the presence of the stranger in session 2, as

well as more distress (particularly crying) being exhibited.

As in the Ainsworth et al. test-retest study, these results

were attributed to the association of the experimental room

with the stress of separation. Lamb (1978) also investigated

patterns of attachment to mothers versus fathers during

testing sessions scheduled one week apart. Results

demonstrated no order effects; i.e., there was not a tendency

for any classification or behavior to be more or less common

in either session.

Connell (1978) tested 55 white, middle-class infants

using the strange situation at 12 and 18 months. He found

significant correlations in measures of contact-maintaining,

proximity- seeking, resistance, and avoidance across the two

sessions. Behavior towards the stranger was less consistent,

with significant correlations emerging in resistance,

avoidance, and proximity seeking. Approximately 81% of the

infants were classified in the same ABC group six months later

(kappa = 0.603), although Connell used techniques which











resulted in the elimination of eight subjects in subgroups B1

and B4 from the analyses.

In order to evaluate the construct validity of the

attachment concept, Waters (1978) examined the stability of

individual attachment behaviors, ratings of behavior

categories, and overall group classification in a middle-class

sample. Fifty infants were assessed at ages 12 and 18 months,

utilizing the Ainsworth strange situation paradigm. Time

samples of discrete behaviors and crying were recorded, as

well as behavioral ratings of five "attachment behavior

categories" (i.e., proximity-seeking, contact-maintaining,

proximity and interaction avoiding, contact resisting, and

distance interaction). Infants were also assigned to overall

categories (A, B, or C) as well as subgroups, based on their

strange situation performance. Behaviors were analyzed

separately for preseparation, separation, and reunion

episodes. Waters found that only four of 21 correlations

between discrete behavior scores at 12 and 18 months reached

the level of significance, and suggested that their occurrence

in the strange situation was frequently too low to allow for

adequate reliability. However, when ratings of interactive

behavior categories were considered, 13 of 18 correlations

were significant, including all of the behaviors directed at

the mothers in the reunion episodes. In terms of attachment

classification, 96% of the children maintained the same

attachment pattern over the six month period (kappa = 0.926),

while 30 of 50 subjects were also classified in the same










subgroups at both times. Waters interpreted this consistency

as support for the stability of attachment, conceptualized as

patterns of interactive behavior affected by variations in

context.

Vaughn, Egeland, Sroufe, and Waters (1979) investigated

the stability of attachment in a lower-class, urban sample (N

= 100) in order to assess the impact of stress and changing

life circumstances on the mother-infant relationship. As in

the previous study, infants were observed in the strange

situation at 12 and 18 months. In addition, mothers completed

a stressful life-events inventory to document events occurring

between the 12 to 18 month period. Although stability of

classification at the two assessments did reach the level of

significance (kappa = 0.321), considerably more children

changed groups than in the Waters (1978) study (38%, as

compared with 4%). In the evaluation of the relationship

between attachment and life events, avoidant and resistant

infants were grouped together and compared with securely

attached infants. Mothers of securely attached infants at 18

months reported fewer stressful life events than did mothers

of anxious infants. Furthermore, changes from secure to

anxious attachment were associated with higher stressful-

events scores when compared with subjects who maintained their

secure classification over the six month period. Changes in

the opposite direction (from anxious to secure attachment)

were not related to stressful events scores. Vaughn et al.

suggested that these data confirmed that stressful life

circumstances may have seriously disrupted the mother-child











relationship, although stable conditions did not appear to be

sufficient for improvements in interaction.

In a larger sample (N = 189) of low-income mother-infant

dyads which included the 100 subjects reported above, Egeland

and Farber (1984) found that 60% of the sample retained the

same attachment classification from 12 to 18 months (kappa =

0.299). In this study, as contrasted with the smaller sample,

virtually no relationship was found between stressful life

events and quality of attachment. The only significant

finding suggested that mothers of girls rated as resistant at

12 months reported higher life stress scores. In addition,

mothers of infants who changed from B to C reported increased

life stress as compared with mothers of stable B babies.

As part of a study investigating differences between

attachment to mothers versus fathers, Main and Weston (1981)

assessed changes in attachment classification at 12 and 20

months. These infants were from predominantly middle-class

homes and were selected for absence of birth complications,

no major separations from the parent within two months of the

assessments (two weeks or more), and 25 hours or less per week

of out-of-home daycare experiences. Fifteen infants were seen

with their mothers at both points, with the remainder seen

with fathers. The authors found that 80% of this sample was

classified into a similar category at both times (kappa =

0.69), although a fourth category called "unclassifiable" was

also utilized. (The unclassifiable category consisted of

infants who demonstrated conflicted or inconsistent behavioral










patterns which did not fit readily into one of the other

groups.) This stability was statistically significant, as was

stability of all interactive behavior scores, with the

exception of contact resistance in the mother sample. Only

proximity avoidance with fathers was significantly stable.

Thompson, Lamb, and Estes (1982) addressed the stability

of attachment patterns in an unselected middle-class sample,

maintaining that Waters' (1978) subjects had been selected for

minimal family stress and change. Forty-three infants were

observed in the strange situation at ages 12.5 and 19.5

months, and their mothers completed questionnaires regarding

changes in caregiving and family stability over the

experimental period. Fifty-three percent of the subjects

maintained their original classification status at the second

assessment (kappa = 0.033), and 26% of the children were also

classified into the same subgroup. Insecurely attached

infants at 12.5 months tended to change groups more often than

did securely attached infants. Changes in status were

significantly related to mothers returning to work and the

experience of caregiving by someone other than the mother.

However, changes in classification were bidirectional; i.e.,

infants experiencing these changes were equally likely to

shift from secure to insecure, as from insecure to secure.

Life events other than caregiving circumstances were not

related to attachment, suggesting that the critical variables

were those most likely to affect the ongoing quality (or

quantity) of the mother-infant relationship.











Waters (1983) has taken issue with the interpretation

offered by Thompson et al. for the considerable attachment

instability found in their middle-class sample. Maintaining

that the other studies (reviewed above) demonstrated that

attachment stability is a well-replicated finding, he

questioned the validity of these results, as the Thompson et

al. percentage of similar classification (53%) was only

slightly higher than that expected by chance, and was

considerably lower than the rate found by Vaughn et al. in a

disadvantaged sample (62%). Waters denied that his sample had

been specifically selected to reduce the effects of stressful

life circumstances, and therefore results so dissimilar from

the Thompson et al. sample would not have been predicted. He

suggested that the small sample size of the second study and

possible differences in scoring techniques may have resulted

in these disparate findings.

In answer to Waters, Thompson, Lamb, and Estes (1983)

have questioned whether attachment stability can be assumed

from the data presented to date. These authors dismissed the

results of the Connell (1976) and Main and Weston (1981)

studies on the basis of incomparable scoring and

classification techniques. They related their results to

differences in the types of stressful events experienced by

the families studied, rather than the frequency of events, as

assessed by Vaughn et al. Changes in patterns of caregiving

and the work status of the primary attachment figure would be

hypothesized to have impact on quality of attachment, as found










in the middle-class sample. Thompson et al. suggested that

middle-class families might have better resources to cope with

such stressors in a positive way, than would disadvantaged

families.

Owen, Easterbrooks, Chase-Lansdale, and Goldberg (1984)

studied quality of attachment at 12 and 20 months to both

mothers and fathers in relation to the mothers' employment

status. Seventy-eight percent of the children showed no

change in attachment classification with mothers (kappa =

0.174), while 62 % of the attachments remained similar with

fathers. Attachment classification was shown to be more

stable in children whose mothers returned to work during their

second year of life, as compared with mothers who maintained

full-time, part-time, or nonemployment over the 20 month

period. The authors suggested that concerns over the effects

of the change in employment status may have motivated the

mothers to maintain continuity in their relationships with

their infants. In addition, the authors cautioned that the

subjects were drawn from a highly educated, middle-class,

white sample, which limits its generalizability. Finally, the

incongruence of these results with those of Thompson et al.,

may relate to the timing of return to work and assessment of

attachment. In other words, a disruption in the relationship

may occur for a short period following change in employment,

which could inflate the overall estimate of attachment

instability.

To summarize, only one study to date has addressed test-

retest stability of the strange situation over a short











interval, using the same parent at both sessions (Ainsworth et

al., 1978). Although changes in patterns of behavior due to

the stress of a second assessment were predicted,

classification changes were largely accounted for by the

movement of avoidant infants to the secure category. The

relatively low percentage of infants who maintained similar

group membership suggests that the assessment paradigm may be

more sensitive than expected to temporally-contiguous

stressors involving separation, and less reflective of a

stable, enduring relationship quality. The two other studies

employing similar time intervals yielded contradictory

results: Lamb (1978) found no carryover from Session 1 to 2,

while Willemsen et al. (1974) found an increase in attachment

behaviors and distress from session 1 to 2.

Long-term (six month) test-retest research has produced

mixed results. The most striking feature of this research is

the wide variability in the estimates of stability which were

produced (kappa values ranged from 0.033 to 0.926). Two of

the studies reporting kappa values above 0.6 (Connell, 1976;

Main & Weston, 1981) used unstandardized variations of the

classification methods suggested by Ainsworth and her

colleagues. Vaughn et al. (1979) produced results reflecting

a lower proportion of stability, as predicted for a less

advantaged population. Their findings indicating that anxious

attachment and changes from secure to anxious attachment were

associated with high stress scores support this view.

Unfortunately, these findings were not replicated in their










larger study (Egeland & Farber, 1984), in which life stress

was not found to be consistently related to anxious attachment

or changes from secure to anxious attachment.

The Thompson et al. (1982) data also present

interpretative difficulties. Although their speculations

regarding the effects of caregiving circumstances on

attachment seem reasonable, the disparity between their sample

and the similar population studied by Waters is puzzling. As

Waters did not present data on the life circumstances of his

subjects, no firm conclusions can be drawn. Furthermore, Owen

et al. (1984) found a change in employment after the first

year to be related to more stable attachment, which is

inconsistent the Thompson et al. results. In general, these

data do not elucidate orderly conclusions about either the

stability of attachment classifications or their association

with life stress or changes in caregiving arrangements and

maternal work status.

Effects of Family Environmental Variables

Although debate continues regarding the role of

environmental variables in the formation of secure attachment,

several have received considerable attention. Research

investigating these factors, including maternal employment

status, infant temperament, social support, maternal

responsiveness, and life stress, will be addressed in this

section.

Both Belsky and Steinberg (1978) and Ainsworth et al.

(1978) have reviewed the available literature focusing on the

relationship of attachment security and day care. Belsky and











Steinberg concluded that very little evidence exists to

suggest that day care results in deleterious effects on the

infant-mother bond. They cautioned, however, that the

majority of studies have investigated high-quality,

university-affiliated centers which may not be representative

of the modal American environment. Similarly, Ainsworth and

her colleagues concluded that the majority of studies

exploring the effects of working versus non-working mothers

produced negative results. These authors have suggested that

the critical variable may be the stability and continuity of

alternative caregiving arrangements, rather than the working

vs. non-working mother dichotomy.

Several studies have addressed this issue since the

publication of these reviews. Chase-Lansdale (1981) studied

single- and dual-wage-earner families, and found a similar

proportion of secure attachments in both groups. Sons of

working mothers did demonstrate fewer secure attachments to

their fathers, as well as fewer instances of secure attachment

to both parents. Owen, Easterbrooks, Chase-Lansdale, and

Goldberg (1983) also evaluated the relationship between

attachment to each parent and maternal employment. No

differences in the proportion of secure attachments were found

between groups of children whose mothers were employed full-

time, part-time, or not at all. Finally, Schwartz (1983)

studied attachment behavior at 18 months in children who

attended day care on a part-time or full-time basis, and those

who did not attend at all. She found that full-time day care










children demonstrated more avoidant behavior during the final

reunion episode of the strange situation than did children in

either of the other groups. Attachment classifications and

their relationship to this finding were not reported. To

summarize, the three previous studies, all of which utilized

middle-class samples, did not produce evidence of strong

support for the relationship between quality of attachment and

maternal employment.

In contrast, Vaughn, Gove, and Egeland (1980) found that

regular, daily separations occasioned by out-of-home care in

the first year of life were associated with anxious-avoidant

attachment. These authors hypothesized that physical

inaccessibility of the mother might mirror psychological

inaccessibility in its effects on quality of attachment;

i.e., out-of-home care might contribute to anxious attachment

patterns. Three groups of mother-infant pairs from a

disadvantaged population were selected to meet criteria of

out-of-home care initiated before 12 months of age (n = 34),

out-of-home care begun between 12 and 18 months (n = 18), and

a home-only care control group (n = 52). A variety of

maternal psychological variables, measures of mother-infant

interactions at three and six months, and infant behavioral

observations did not differentiate the three groups. From

assessments with the strange situation at 12 and 18 months,

there emerged a significant relationship between work status

and attachment security, with more anxiously-avoidant infants

appearing in the early return to work group. At 18 months,

security of attachment was related both to early initiation of











out-of-home care and family intactness (presence of a male

figure in the household predicted secure attachment). These

findings are in line with those of Schwartz (1983), in

suggesting that full-time maternal employment may result in

avoidant patterns of behavior.

Clearly the study conducted by Vaughn et al. (1980)

implicates social support as an important mediating variable

in the formation of secure attachment. Crockenberg (1981)

studied 48 mother-infant pairs to assess the influence of

infant irritability, mother responsiveness, and social support

on attachment. Babies were administered the NBAS at days five

and ten following birth and evaluated in the strange situation

at 12 months. Home visits at three months were conducted to

observe maternal responsiveness (measured by time to respond

to infant distress signals), as well as to assess sources of

stress and support for the mothers through interviews. The

measure of social support combined both an estimate of number

of support persons available to the mother, as well as her

experience of significant stressors over the first year of

her infant's life. Low social support was found to be

significantly related to anxious attachment, for babies who

were rated as highly irritable. Similarly, low responsiveness

was related to insecure attachment in the high irritable

infant group only.

In further analyses of these data, Crockenberg (1985)

found that maternal responsiveness at 12 months had a greater

effect on irritability in the strange situation than did the










behavior of the mothers at three months. She suggested that

mothers of irritable babies tended to grow less responsive

over the first year, particularly under conditions of low

social support.

In line with Crockenberg's findings on irritability and

attachment, Belsky, Rovine, and Taylor (1984) evaluated the

relationship of reciprocal interaction at one, three, and nine

months to attachment classification at one year. Securely

attached infants experienced moderate levels of stimulation

over their first year of life, while avoidant infants were

overstimulated and resistant infants were understimulated.

Both types of insecure infants were found to be more fussy at

three and nine months, suggesting that infant temperament may

play a role in the development of maladaptive caretaking

patterns. Similarly, Egeland and Farber (1984) found that in

a lower SES sample, infants classified as resistantly attached

at one year had been rated as less alert and responsive during

feeding and play during their first few months of life.

Waters, Vaughn, and Egeland (1980) administered the NBAS

to 100 first-born neonates at age seven and ten days, then

administered the strange situation at one year. Results

indicated that anxious/resistant (Group C) infants scored

lower on orientation, motor maturity, and regulation items at

day seven. These differences had disappeared by day ten. The

authors suggested that these results supported other evidence

linking the more difficult nature of group C children with

difficult interaction and the development of anxious

attachment.











Finally, two studies investigated the relationship of

family stress, as measured by the Parenting Stress Index, and

patterns of attachment (Green, 1981; Hamilton, 1981). In

neither of these white, middle-class samples was a significant

relationship between these variables found.

To summarize, the majority of research done to date has

not found that maternal employment negatively affects security

of infant attachment. A major exception is the study by

Vaughn et al. (1980), which demonstrated that early out-of-

home care in a disadvantaged population resulted in a larger

proportion of anxiously attached children. Research

investigating long-term attachment stability (over a six month

period) has suggested that changes in attachment

classification may be associated with lower socioeconomic

status, high life stress, and changes in caregiving

circumstances. As a cautionary note, replication across

studies on these variables has been problematic, even within

research groups working with longitudinal studies. In

addition, some of the results have been directly contradictory

(e.g., Owen et al.'s (1984) finding that mothers who returned

to work in the second year had offspring with more stable

attachments, as contrasted with Thompson et al. (1982) who

found return to work to be associated with changes in

classification).

Aims and Hypotheses of the Study

The present study has two aims. First, given the rather

weak evidence available for the stability of the ABC typology










over both short and long time intervals, and the importance of

stability for the predictive validity of the classification

system, the study will address the issue of the contribution

of contextual effects to the stability of attachment over a

two week test-retest period. It is hypothesized that

controlling for context-specific stress will decrease the

occurrence of intensified attachment behavior in session two.

In other words, the stability of the ABC classification is

expected to be significantly higher than that reported by

Ainsworth et al. (1978) for infants seen in two different

contexts. In addition, interactive behavior categories are

expected to demonstrate significant stability across sessions,

in particular, for changed context subjects.

Family environmental variables have not been examined in

other research investigating short-term stability. Factors

which have been hypothesized to play a role in long term (six

to eight month) stability will be examined in relation to the

current sample. These variables will include maternal

employment, continuity of caregiving, and life stress.

Many of these same variables have been implicated in the

quality of mother-infant attachment which develops over the

first year. The second aim of this study is to investigate

the relationship of stressful life events, perceived social

support, infant temperament, maternal temperament, and

stability of caretaking arrangements to attachment

classification.

Given the research on environmental factors which has

been presented, it is predicted that changes in caretaking







30



circumstances will be negatively related to security of

attachment. Measures of social support and maternal

adaptiveness should show positive, though perhaps smaller

associations to secure attachment. The measure of maternal

temperament (including adaptiveness) will be utilized to

approximate the dimension of sensitive responsiveness. High

life stress scores may be related to anxious attachment,

although support for this hypothesis has been less clear.

Finally, irritable infant temperament is expected to be

related to anxious attachment.
















METHOD


Subjects

Thirty mothers and their one-year old infants served as

subjects in this study, although one mother-infant pair

completed only the initial interview and first session. The

mean age for mothers was 28.3 years (range = 20 38). One

mother was separated from her husband; all other couples had

intact marriages. The average family income was approximately

$20,000 (range = $5200 $60,000). Subjects were recruited

through several sources. These included ads in the

university newspaper and a local advertising flyer, notices

posted on bulletin boards in the teaching hospital and in

local stores, word of mouth, and a list of mothers who had

participated in a previous research project (unrelated to the

current study). Mothers were paid ten dollars for their

participation. The initial testing of the infants was

conducted within three weeks of their first birthday, with the

retest scheduled for two weeks later. The mean age for

infants at first testing was 363.5 days (range = 344 390),

and at retest was 379.9 days (range = 359 404). Thirteen

boys and seventeen girls participated in the study. Fifty-

three percent of the subjects were first born children (eight

boys and eight girls).

Materials

Interview. A structured interview was utilized to assess










demographic variables and the stability of caretaking

arrangements (see Appendix B). This interview was modeled

after one done by Thompson (personal communication, November,

1983) and included questions concerning the number of

babysitters and day care centers utilized, number of changes

in these arrangements, and number of planned events (e.g.,

social engagements, days of employment) which were disrupted

due to problems with caretaking provisions.

Social Support Questionnaire. Social support for the

mothers was assessed using the Social Support Questionnaire

(Sarason, Levine, Basham, & Sarason, 1983). This instrument

was composed of 27 hypothetical situations for which the

subject was asked to list potential support people and rate

satisfaction with that level of support. The questionnaire

yielded both a measure of total number of social supports (N)

and perceived satisfaction with support (S). A test-retest

study conducted over a four week period found correlations of

0.90 for N and 0.83 for S (Sarason et al., 1983).

Life Experiences Survey. Maternal life stress was

measured with a modified version of the Life Experiences

Survey (Sarason, Johnson, & Siegel, 1978), which produced an

estimate of the number of negative and positive life events

occurring over the previous year. This instrument was

selected because of the wide variety of events it lists,

including those concerning employment stressors. Subjects

simply identified events which had occurred in their lives

over the past year, rating them as either negative or

positive. Pearson product-moment correlations computed over a











five to six week test-retest period were found to range from

0.19 to 0.53 for positive life change and from 0.56 to 0.88

for negative life change (Sarason et al., 1978).

Infant Behavior Questionnaire. The Infant Behavior

Questionnaire, designed to measure infant temperament, yielded

six dimensions of temperament including activity level,

distress to approach novel stimuli (or fear), distress to

limitations, duration of orienting, soothability, and smiling

and laughter (Rothbart, 1981). The instrument required

frequency judgements on approximately 90 behavioral items

which might have occurred over the previous one to two weeks.

Stability coefficients over the three to 12 month age bracket

ranged from 0.06 to 0.80, with greater stability reported

above six months (Rothbart, 1981).

Dimensions of Temperament Survey. Maternal temperament

was measured with the Dimensions of Temperament Survey

(Lerner, Palermo, Spiro, & Nesselroade, 1982). This

instrument consisted of 34 behavioral items which subjects

marked as either true (accurate self-descriptions) or false

(inaccurate). The five temperamental factors identified by

this instrument were activity level, attention

span/distractibility, adaptability, rhythmicity, and

reactivity. Three day test-retest stability ranged from 0.60

to 0.93 (Lerner et al., 1982).

The Strange Situation. As described in the introduction,

the strange situation consisted of eight short episodes, in

which the infant's response to separation, reunion, and










overtures by a stranger were noted. The episodes were

designed to gradually increase the stress to which the infant

was exposed, with concomitant increases in attachment

behaviors expected. Strange situation assessments occurred in

two different experimental rooms, one located in Shands

Teaching Hospital and the other in the Department of Clinical

Psychology research trailer. These facilities were located

several miles apart and in very different settings, chosen to

minimize the children's recognition of the assessment

situation. The trailer was located on a quiet drive in a

woodsy setting, isolated from the rest of the university. In

contrast, the room at Shands Teaching Hospital was located

within a large medical center at a busy intersection of the

campus.

Each of the experimental rooms contained two chairs and a

number of toys appropriate for one-year-old infants. Toys at

the trailer included toy phone, foil pie plates, plastic

hammer, milk bottle with shapes, baby doll, musical apple,

felt chicken, plastic shapes, plastic toy rabbit, and toy keys

on a ring. Toys at the hospital consisted of large colored

ball, plastic bucket and shovel, colored plastic blocks, toy

camera music box, chicken rattle, cloth pig, plastic pull

train, cloth doll with rattle, bird squeeze toy, fish rattle,

and soft rubber elephant. The experimental room at the

hospital measured 10 by 12 feet, was carpeted, and contained a

couch and two small end tables in addition to the equipment

described above. The room at the trailer was smaller,

measuring 8 by 6, uncarpeted, and furnished only with a file











cabinet.

Procedure

Prior to the first assessment, the subjects were

interviewed by the principal investigator, either in their

homes or at the psychology clinic at Shands Teaching Hospital.

The procedures of the study were explained and informed

consent was obtained (Appendix A). Next, the structured

interview was conducted and mothers were instructed in the

procedures of the four inventories. Finally, the mothers' role

in the strange situation was outlined and each was given a

descriptive handout.

All subjects were tested in two sessions of the Ainsworth

Strange Situation, scheduled at two week intervals. Although

scheduling all children for assessments at the same time of

day proved impossible, an attempt was made to maintain

consistency in time of assessment between the two sessions of

each subject. For half of the subjects, the testing sessions

occurred in two different experimental rooms (at the research

trailer and Shands Teaching Hospital). The remaining subjects

were tested twice in the Clinical Psychology research trailer.

For subjects tested in two different settings, the order of

the experimental rooms was counterbalanced. When the subjects

arrived at the prescribed setting, the procedures of the study

were reviewed with them by the principal investigator or a

research assistant, and the mothers were given a cue sheet to

take into the room with them (Appendix C). Following each

session of the strange situation, mothers were asked (1) if










they were surprised by any aspect of their child's behavior,

and (2) whether anything unusual had occurred on the day of

testing which might have influenced the assessment.

Three different female strangers were utilized; the order

of their appearance was varied randomly. The strangers were

recruited from an undergraduate psychology class and they

received course credit for their participation in the study.

They engaged in four training sessions prior to beginning the

experiment. Although these students were familiar with

attachment theory in general, they were unaware of the

specific hypotheses of the study and were blind to the

behavior of each baby in any prior session. (Both the

experimental settings and stranger behavior were reviewed and

approved by Dr. Ross Thompson from the University of Nebraska

following the initial training sessions.)

The infants' behavior in the strange situation was

videotaped from behind one-way mirrors. Measures which were

scored from these records included classification into

groups and subgroups and ratings of proximity and contact-

seeking, contact-maintaining, avoidance, and resistance during

the two reunion episodes. These four interactive behaviors

have been identified previously as most salient and reliable

in differentiating patterns of attachment (Waters, personal

communication, April, 1985). Descriptive ratings were

provided by Ainsworth et al. (1978), with seven points on each

scale anchored to specific patterns of response.

Two raters (RAI and CRI) were trained by Dr. Everett

Waters in the scoring of the strange situation. Rater CRI was











blind to the experimental hypotheses of the study. Training

was undertaken using videotapes from an unrelated study. The

two raters each scored half of the session one and half of the

session two tapes, and were blind to the behavior and

classification of the infants in their other session. The

raters then scored the other half of the videotapes. Tapes

were rerated in groups of four if there was any disagreement

on group or subgroup classification, or if ratings of

interactive behaviors differed by more than two points.

Independent reliability was achieved with Dr. Mary Jo Ward who

scored a random sample of the tapes (N = 15). She was blind

to the hypotheses of the study, as well as to the behavior of

any infant in a previous session. Dr. Ward had been trained

by Dr. Alan Sroufe (Institute of Child Development,

Minneapolis, Minnesota) and had demonstrated consistent

reliability with his ratings.















RESULTS


Reliability

Two independent raters each scored 82% of the videotapes.

Ninety-eight percent agreement was achieved for ABC

classification (Kappa = 0.93). Eighty percent agreement was

reached on subgroup classification. Intraclass correlation

coefficients (Bartko, 1966) were calculated for the

interactive ratings. The values for proximity-seeking,

contact-maintaining, resistance, and avoidance were 0.93,

0.97, 0.86, and 0.90, respectively.

Reliability was also calculated between the final

consensus values used in the analyses and the ratings on a

random sample scored by Dr. Ward. Mean values for these

ratings were 0.80, 0.91, 0.92, and 0.60, respectively.

Stability

The group of infants tested twice in the same setting were

seen both times in the trailer, with no infants seen twice in

the hospital. Therefore, initial analyses were conducted to

ascertain that the effect of a particular setting did not

cloud the analyses of stability. First, a comparison of the

trailer/trailer group (TT, N = 15) and the trailer/hospital

group (TH, N = 8) was made for the distribution of the three

classifications at test and retest. Using chi-square tests,

there were no significant differences. Because the expected










cell frequencies were lower than optimal for the chi-square

statistic, comparisons were also made with Fisher's exact

test. For this comparison, the distribution of A and C

infants considered as one group was compared with group B,

with no significant differences found between the treatment

groups. The same two groups (TT/TH) were examined for

differences in number of infants who retained the same

classification over the two sessions. Again, a chi-square

test revealed no significant differences.

As a further check for setting differences, all infants

seen in the trailer at session one (n = 23) were compared to

those seen in the hospital (n = 6). There were no differences

between groups using the Fisher's exact test to examine

distribution of infants into A plus C versus B categories. T-

tests also demonstrated no differences in interactive ratings

during session one between these groups.

Next, the group of infants tested in the order

trailer/hospital (TH) was compared with that tested first in

the hospital, followed by the trailer (HT, N = 6). Chi-square

analyses revealed no differences between these groups in the

distribution of classifications either at the first or second

session. In addition, Fisher's exact tests demonstrated no

differences at test or retest in distribution of subjects into

A plus C versus B categories. Likewise, there were no

differences between these groups in the number of infants who

retained their classification across sessions.

In addition to examining the effects of setting on overall

group membership, the four interactive variables were











scrutinized. With one exception, no differences were noted on

t-tests comparing groups TT and TH in terms of proximity-

seeking, contact-maintaining, resistance, or avoidance in

either reunion episode, during both test and retest.

Avoidance in episode 5 of the retest was significantly higher

for infants in the TT group than the TH group (t = 2.131; p <

0.0489). Given the relatively large number of statistical

tests employed, as well as the lack of support for this

finding in other measures (and theoretically), it seems likely

that this is not a meaningful finding. Difference scores on

each interactive variable calculated across sessions (e.g.,

Davoid5 = Avoidance Episode 5, test Avoidance Episode 5,

retest) were also examined using t-tests. There were no

significant differences.

As with the overall group classification, comparisons

between the TH and HT groups were next made. There were no

differences between these two groups on measures of any of the

four interactive variables in each reunion in both test and

retest sessions. Several significant findings were noted in

difference scores. Babies in the TH group tended to increase

proximity-seeking in episodes 5 and 8 over the two sessions,

as compared with decreases in the HT group (t = -2.8820, p <

0.0138, and t = -2.3039, p < 0.0399, respectively). In

addition, avoidance in episode 8 tended to decrease across

sessions in the TH group, as compared with increases in the HT

group (t = 2.3569, p < 0.0363). All other difference score

comparisons were nonsignificant.











In light of the general lack of support for setting

effects, analyses were conducted to compare trailer/trailer

subjects with all changed context subjects (TH and HT). The

distribution of infants for the two groups at test and retest

is described in Table 1. First, the likelihood that a

particular classification was more frequent in one of the

groups at either test or retest was examined. Chi-square

tests were nonsignificant. Fisher's exact tests applied to

the A plus C versus B distribution at both sessions were also

nonsignificant. Similarly, chi-square tests demonstrated no

tendency for retaining classification to be more likely in one

group as compared with the other. As in the previous section,

t-tests were used to compare individual interactive ratings in

both episodes in each session. There were no differences. In

addition, no differences were found on comparisons of any of

the interactive rating difference scores.

As a final test of differences in stability between the

two groups, kappa was computed for each. The values were

-0.1409 and -0.1572, for TT and TH/HT, respectively. As

suggested by Cohen (1960), a two way comparison was made by

evaluating the normal curve deviate, with no significant

differences emerging (z = 0.03, p > 0.05).

As no significant differences in stability between TT and

TH/HT groups were found, all subjects were combined for the

following analyses. First, of the total sample in the first

session, 25 infants (83%) were classified as securely

attached, while three infants were rated as avoidant (10%),
















TABLE 1



Classification of Infants at Test and Retest
in Same Context/Changed Context Groups



Trailer/Trailer Group

Retest

A B C

0 1 0 1
(0.07)

1 10 2 13
(0.86)

0 1 0 1
(0.07)

1 12 2 15
(0.07) (0.80) (0.13)


Hospital/Trailer



A B

0 1


2 10


0 1


2 12
(0.14) (0.E


and Trailer/Hospital Group

Retest

C


0


0
86) (0.0)


1
(0.07)

12
(0.86)

1
(0.07)

14


Test A


B


C


Test A


B


C












and two were rated resistant (7%). Although this distribution

is weighted more heavily with secure infants than often found

in the literature, it is comparable to some studies using

middle class samples (e.g., Owen et al., 1984).

The overall stability was 69% (kappa = -0.116). Table 2

describes the distribution of infants at test and retest in

the current study, as compared with Ainsworth et al.'s (1978)

sample. There was no clear trend in direction of change.

Five children changed from secure to anxious attachment (three

B/A and two B/C) and four switched from anxious to secure (two

A/B and two C/B). Twenty-eight percent of the infants

retained the same subgroup classification from the first to

second session. (See Table 3.) Intraclass correlation

coefficients for proximity-seeking, contact-maintaining,

resistance and avoidance were 0.41, 0.56, 0.12, and 0.14,

respectively, for episode 5, and 0.42, 0.57, 0.28, and 0.14,

for episode 8.

Attachment behaviors (proximity-seeking and contact-

maintaining) tended to increase across sessions in episode

five (Dprox5 = -0.7241; Dcont5 = -0.8448) and decrease in

episode eight (Dprox8 = 0.7241; Dcont8 = 0.4828). Avoidance

tended to decrease across sessions in episode five (Davoid5 =

0.5517) and increase in episode eight (Davoid8 = -0.4138).

Means and standard deviations for the interactive ratings are

presented in Table 4. As illustrated, in session one,

proximity-seeking, contact-maintaining, and resistance

increased significantly from episode five to eight, while















TABLE 2



Classification of Infants at Test and Retest
for the Current Study and Ainsworth's Sample

Current Study

Retest

A B C

Test A 0 2 0 2
(0.07)

B 3 20 2 25
(0.86)

C 0 2 0 2
(0.07)

3 24 2 29
(0.10) (0.83) (0.07)



Ainsworth Sample

Retest

A B C

Test A 0 7 0 7
(0.30)

B 0 12 2 14
(0.61)

C 0 1 1 2
(0.09)

0 20 3 23


(0.00) (0.87)


(0.13)

















TABLE 3



Subgroup Classification at Test and Retest

Retest

Al A2 B1 B2 B3 B4 C1 C2

Test Al 0 0 1 1 0 0 0 0 2
(0.07)

A2 0 0 0 0 0 0 0 0 0
(0.0)

B1 0 0 1 3 0 0 0 0 4
(0.14)

B2 2 0 2 2 1 0 0 1 8
(0.28)

B3 1 0 1 0 5 5 1 0 13
(0.45)

B4 0 0 0 0 0 0 0 0 0
(0.0)

Cl 0 0 0 0 0 1 0 0 1
(0.03)

C2 0 0 0 0 1 0 0 0 1
(0.03)

3 0 5 6 7 6 1 1 29
(.10) (.17)(.21)(.24)(.21) (.03)(.03)












avoidance decreased. In contrast, the only significant change

across episodes in the retest was an increase in contact-

maintaining from five to eight. Combined means for episodes

five and eight for each session, as well as correlations

between these means, can be found in Table 5. Differences

between the means of the two sessions were nonsignificant for

all interactive ratings.

Further paired comparison t-tests were computed for

subjects sorted into two groups according to whether they

changed (n = 9) or retained (n = 20) their classifications

across sessions. Infants who changed classifications showed

increases from episode five to eight (in session one) in

proximity-seeking, contact-maintaining and resistance (p <

0.0327, 0.0112, and 0.0099, respectively). During the retest,

only a decrease in proximity-seeking from episode five to

eight was significant (p < 0.0404). For those infants who

retained their classification, proximity-seeking and contact-

maintaining increased across the episodes in the first session

(p < 0.0024 and p < 0.0009), while avoidance decreased (p <

0.0116). During the second session, the only significant

difference across episodes was an increase in contact-

maintaining (p < 0.0218).

Given the pattern of instability across sessions, the

association between short-term stability of infant-mother

attachment and family environmental and demographic variables

was assessed. For these analyses, subjects were sorted into

those who retained their classification over the two sessions











(N = 20) versus those infants who changed groups from test to

retest (N = 9). Using t-tests, no significant differences

were found between the groups on the following variables:

social support, life stress, maternal temperament, infant

temperament, caretaking changes, parental age and hours

worked, and separations from mother over the first year of

life. In contrast, infants who retained their classification

were found to come from lower income families (M = $17704)

than those who changed groups (M = $27700; t = 2.1503, p <

0.0403). This difference is contradictory to findings in the

literature suggesting a higher rate of stability in higher

income families, and is likely to be a spurious finding

resulting from the large number of statistical tests employed.

A series of chi-square tests evaluated the association

between stability of classification and categorical variables.

No differences were found based on sex of subject, birth

order, level of parental education, maternal work status, or

changes in caretakers over the first year.

To summarize the previous section, actual setting in which

the strange situation occurred had no effect on classification

in a particular setting or on stability of classification. In

addition, there were no differences in the stability of

classification in the same context group (TT) as compared with

the changed context group. Stability for the sample

considered as a whole was unimpressive (k = -0.166). Finally,

when infants who retained their group status were compared

with those who changed across sessions, no significant



















TABLE 4



Means and Standard Deviations for Interactive Ratings

Test


Episode 5


Mean


Episode 8


Paired
Comparisons


SD Mean SD


2.93 1.98 4.45 1.49


2.57

1.43

2.50


4.38

1.90

1.57

Retest


3.66 1.93


3.43 2.40

1.38 0.92

1.90 1.45


2.25

1.46

1.30


0.0002


0.0001

0.0377

0.0017


3.72 1.92 0.7636


3.95 2.32 0.0122

1.59 1.17 0.3392

1.97 1.38 0.7869


Proximity-
Seeking

Contact-
Maintaining

Resistance

Avoidance



Proximity-
Seeking

Contact-
Maintaining

Resistance

Avoidance

















TABLE 5



Comparison of Means of Strange Situation Measures
for Sessions One and Two and Correlations Between Sessions


Measure Session 1 Session 2 Significance r
Mean Mean of Difference

Proximity-
Seeking 3.69 3.69 n.s. 0.54
(0.0024)

Contact-
Maintaining 3.48 3.69 n.s. 0.65
(0.0001)

Resistance 1.67 1.48 n.s. 0.23
(0.2263)


Avoidance 2.03 1.93 n.s. 0.39
(0.0354)












differences emerged on demographic or family environmental

variables. One exception to this was family income. Due to

the small sample size, the group of infants which changed

classification could not be sorted according to direction of

change (e.g., secure to anxious versus anxious to secure).

Family Environmental Variables

The second aim of this study was to investigate the

relationship of a variety of family environmental variables

to quality of attachment. For these analyses, only the

classifications from the first session were utilized. The

group of securely attached infants (N = 25) was compared with

those children who were anxiously attached (N = 5), which

included both avoidant (n = 3) and resistant (n = 2) infants.

T-tests were used to compare these groups on measures of

social support, life stress, maternal temperament, infant

temperament, hours that mother and father spent caring for the

baby, caretaking changes, and income. There were no

significant differences except for the maternal temperament

factor of activity. On this variable, mothers of securely

attached infants tended to score lower than did mothers of

anxiously attached infants (t = -2.7282; p < 0.0109). A chi-

square test evaluating the relationship between attachment and

maternal work status also proved to be nonsignificant.
















DISCUSSION


The present study investigated the hypothesis that the

stability of attachment classification would be increased if

an infant's recognition of the assessment context was

eliminated. In order to evaluate this assumption, half of the

subjects were tested in two very different settings, presented

in counterbalanced order, with the remainder seen twice in the

same context. Results demonstrated that context effects had

no bearing on classification stability. The coefficients of

agreement (kappa) for the two groups (same context and changed

context) were similarly unimpressive, and nonsignificantly

different. This finding was supported by the lack of

differences between the groups in interactive ratings and

changes in those ratings across time. Although the relatively

small sample size and the large proportion of securely

attached infants found in the study preclude taking a

definitive stand on the meaning of these findings, it should

be noted that none of the statistical comparisons were even

close to significantly different. The lack of differences

between these groups suggests that if memory for the strange

situation is affecting classification stability, it is likely

to be recognition of the entire sequence of events rather than

of the physical environment alone that affects the children's

behavior in the second assessment.











As the recognition of context hypothesis received no

support, subjects were collapsed into one group in order to

examine stability in the sample as a whole. In this study,

there was a higher percentage of infants who retained their

classification across sessions as compared with the Ainsworth

et al. (1978) sample (69% and 57%, respectively). However, a

comparison of the kappa values computed for these two samples

revealed that when the amount of agreement expected by chance

was considered, stability in the current study was even lower

than in Ainsworth's results.

Temporal instability would not necessarily undermine the

validity of the classification system, if changes in the

quality of attachment were either the predictable result of

anxiety engendered in the first session spilling over into the

second, or if the changes could be related predictably to

environmental changes. Ainsworth and her colleagues were able

to develop reasonable explanations for their findings, based

on the pattern of changes they observed. First, the bulk of

changes in classification were accounted for by avoidant

infants appearing secure in the second session. In addition,

proximity-seeking and contact-maintaining were significantly

higher in session 2, while avoidance decreased from time one

to time two. The authors stated: "it is clear that the

effect of the repetition of the strange situation after a

brief two-week interval was to increase distress and the

intensity of attachment behavior, while at the same time the

negative behaviors of avoidance and resistance decreased or












remained at about the same level of intensity" (Ainsworth et

al., 1978, p.222).

The present study does not support these conclusions. Of

the nine infants who changed classification, four moved from

anxiously to securely attached, and the remainder changed from

secure to anxious. Also in contrast to the Ainsworth study,

changes in interactive behaviors across sessions were not

consistently in the expected directions. In general, the

explanation posed by other researchers to explain short-term

instability lacks validity in relation to these results: the

hypothesis that the anxiety engendered by encountering a

second "strange situation" results in increased attachment

behaviors and a general progression to secure attachment is

not supported.

As Ainsworth and her colleagues suggested, it seems

unlikely that the quality of the mother-infant attachment

would change dramatically over a two week period, unless there

were gross changes in the factors which influence the mother-

infant interaction. Post-experimental questions administered

after each assessment in the current study provided anecdotal

evidence to the effect that infants who changed

classifications had not experienced any noteworthy disruptions

or trauma in the two week interim, relative to those infants

who remained in their original group. Mothers occasionally

reported illness in the child or an unrestful nap, but these

comments were equally distributed across groups and time of

testing.













In a further attempt to understand the pattern of

instability which emerged in the present study, family

environmental variables which have been implicated in long-

term stability were examined. It was not expected that gross

changes in the factors measured would have occurred between

sessions, and in fact, the questionnaires were completed prior

to both assessments. However, based on six to eight month

test-retest studies, it seemed that certain variables such as

infant difficulty, life stress, or caregiving continuity might

affect short-term stability. Clearly, there was no evidence

to support these hypothesized relationships.

To summarize, the current study did not support the

contention that short-term instability in the ABC

classification system was related to stress-inducing effects

of returning twice to the same assessment context.

Furthermore, instability in this sample as a whole was greater

than that found in the Ainsworth sample. As Lamb (1978) had

reported with a mother/father test-retest study, there was no

predictable pattern in the direction of changes which

occurred. Infants were as likely to change from secure to

anxious as they were from anxious to secure. Finally, none of

the personality or environmental variables which were

investigated differentiated between infants who changed versus

retained classification. As mentioned previously, due to the

small number of subjects, the changed classification group

could not be analyzed according to direction of change.












The conclusion which emerges from these results is that

classification of attachment quality over a two week period

lacks stability. In contrast to Ainsworth's findings, but in

line with Lamb's, this instability appeared random in

direction, rather than following a trend to normative or

secure attachment.

This study does have limitations which constrain

overgeneralization of results. The inclusion of a treatment

group comprised of infants seen twice in the hospital setting

would have eliminated the need for analyses of specific

setting effects. In addition, the relatively small sample

size and the securely weighted distribution of infants

presented difficulties for statistical analysis.

Overriding these concerns was the fact that none of the

statistical comparisons were even marginally significant.

Converging evidence from both classification status and

interactive ratings strongly supported the finding of lack of

stability. Further research to document these results should

include a larger sample, both to increase the power of

statistical evaluation and avoid the problems raised by the

ABC sample distribution. There is little reason to believe,

however, that the presence of more anxiously attached infants

in the first session would improve stability estimates.

Both the Ainsworth data and results of long-term studies

(e.g., Thompson et al., 1983) suggest that anxiously attached

children are less likely to retain their classification in a

retest than are secure infants. Finally, future research











might also investigate stability at an interval longer than

two weeks. As a cautionary note, however, as the time between

test and retest increases, the risk of- confounds from

developmental changes and alterations in environmental

circumstances increases as well.

Unfortunately, very little can be concluded regarding the

second aim of this study. In this case, the large percentage

of securely attached infants weighed against the probability

of identifying differences between groups on family

environmental variables.

Keeping in mind that the sample size and group

distribution limit firm conclusions, what might these results

mean for the ABC classification system? One speculation would

be that the ABC taxonomy is not the most parsimonious and

accurate model for describing differences in quality of

attachment. Conceivably, differences within each group are

larger than stylistic differences between groups. More

specifically, infants in different subgroups may have very

different styles of responding to increases in anxiety, which

influence patterns of temporal stability. Although Ainsworth

et al. did not report subgroup membership, their most notable

finding was the movement of seven infants from avoidantly to

securely attached. In the present study, only two infants

were classified as avoidant in the first session, and both

moved to a secure classification (B1 or B2)'at time two. In

terms of the B subgroups, several tentative directional trends

can be identified. B3 infants tended either to remain in that

group or move to B4. B2 infants tended to remain stable or












change to a B1 or Al classification. B1 infants tended to

look like B2's in their second session.

Lamb, Thompson, Gardner, Charnov, and Estes (1984) have

suggested that cluster analysis does not provide strong

support for the ABC classification system. In applying

cluster analysis to five samples which varied in age tested

and cultural background, Gardner, Lamb, and Thompson

(1985) found that boundaries between groups generally

did not fall along the A/B and B/C distinctions. They

reported the C group to appear extremely heterogeneous and

that differences between some B subgroups were often greater

than differences between A and B subgroups. More

specifically, "distal" B's (BI and B2) frequently clustered

with A infants, while proximall" B's (B3 and B4) were often

grouped with C's. In addition, the fit of the ABC

classification and the empirically derived clusters were found

to vary according to the age and cultural background of the

sample, with the closest fit shown in an American, one-year-

old population. The authors conclude by recommending the

consideration of continuous measures comprised of (1) a

composite summing proximity-seeking and contact-maintaining

and subtracting avoidance, and (2) the resistance scale.

The general trend of the movement between subgroups in the

present study is largely consistent with Gardner et al.'s

findings using cluster analysis. In other words, a change in

classification from B1 or B2 to group A (or vice versa) may

actually be reflective of behavioral similarity and continuity











among those groups, rather than suggesting temporal

instability of individual differences, as the current

classification system would dictate.

An examination of the overall pattern of interactive

rating change in the present study demonstrated that there

were significant changes in all four ratings from episode five

to eight in the first session. Two weeks later, only contact-

maintaining showed a significant increase. By sorting the

infants according to stability of classification, an attempt

was made to evaluate whether certain infants had habituated to

the stress of the strange situation, while others had not.

The two groups showed reasonably similar patterns of change in

proximity-seeking and contact-maintaining in session one,

though differing on the other two variables. Differences in

the pattern of change across episodes continued in the retest.

In a larger sample which would allow investigation of

direction of classification change, analyses along these lines

might clarify which types of infants show predictable

behavioral reactions to the stress of separation.

The final section of this paper focuses on practical and

sociological observations of the ABC classification system.

As reviewed in the introduction, the strange situation and its

ensuing ABC taxonomy are plagued with a number of

problems. Questions of both short-term and long-term

stability remain unanswered. Results of the present study

illustrate the questions surrounding short-term temporal

stability. Efforts which have been directed to understanding

six to ten month stability have produced inconsistent results.












First, there has been wide variability reported across

studies. Second, the differences found by various researchers

are not clearly explicable on the basis of sample

characteristics (e.g., note the wide discrepancies in

estimates of stability found in similar middle-class samples).

Finally, attempts to link stability/instability to external

variables have not been definitive: while Vaughn et al.

(1980) found changes from secure to anxious attachment to be

associated with life stress, Egeland and Farber (1984) were

unable to replicate this in a larger sample. Likewise,

Thompson et al. (1982) reported bidirectional changes in

attachment status to be related to changes in mothers'

employment status and caregiving disruption, while Owen et al.

(1984) found increased stability in infants whose mothers

returned to work during the second year.

In spite of problems such as those cited above, the

strange situation is a highly popular and visible metholology.

A brief review of the abstracts from the Society for Research

in Child Development catalog (biennial meeting, 1985)

demonstrates that of the 32 articles discussing attachment, 31

or 97% utilized this methodology. The only author who did not

use the ABC taxonomy was Everett Waters, who is in the process

of developing a Q-sort instrument for assessing attachment.

A cursory look at the major figures producing attachment

research reveals that they fall roughly into those associated

with Ainsworth through training or collaboration (e.g.,

Sroufe, Waters, Vaughn, Main, Egeland) and those not (e.g.,











Lamb, Thompson, Belsky, Crockenberg). The authors who have

been most critical of the methodology while continuing to be

extremely involved in the area are Michael Lamb and his

students. In general, critiques of the taxonomy tend to be

ignored or to generate considerable heat (e.g., Thompson et

al., 1982; Waters, 1983; Thompson et al., 1983). It seems

noteworthy that the major review paper produced by Lamb et al.

(1984) which has a format specifically designed for peer

commentary, stimulated no response from any of the Ainsworth

related group.

Frequently, the focus of the controversy over this

methodology comes to rest on the scoring system for the

strange situation videotapes. In several discussion sections,

results which contradict the ABC model were deemphasized or

dismissed on the basis of inadequate scoring (e.g., Waters,

1983; Ainsworth et al., 1978). Practically speaking, learning

the scoring system is a cumbersome, time-consuming, and

expensive task for researchers who are not collaborating with

a major figure. Clearly, working only from the written

material provided in the Ainsworth et al. book is inadequate

for establishing reliability. Waters (1978) made reference to

a set of standardized videotapes which would be available for

training and reliability. In fact, these were never

developed, presumably because of disagreements among the

prospective authors (Waters, personal communication, 1985).

Given how firmly entrenched the ABC system has become in the

attachment literature, the standardization of the methodology

is critical. Resolution of the controversy surrounding the








61


taxonomy rests on the provision of standardized materials

which would allow the assumption of uniform ratings across

research laboratories and theoretical camps.














APPENDIX A
INFORMED CONSENT



Informed consent to Participate in Research

J. Hillis Miller Health Center
University of Florida
Gainesville, Florida 32610

You are being asked to volunteer as a participant in a

research study. This form is designed to provide you with

information about this study and to answer any of your

questions.

TITLE: Mother-infant attachment: Stability and relationship

to family environmental variables

PROJECT DIRECTOR: Roberta A. Isleib, M.S. 392-2944

PURPOSE: The purpose of the present study is to understand

the effects of variables such as stressful life events, social

support, mother and child temperament, and caretaking

arrangements on infant social behavior and the mother-child

relationship. The study will also investigate the stability

of the classification system which is currently used to

describe infant behavior.

PROCEDURES: Each participant will be briefly visited in their

home, where the study will be explained and questionnaires

concerning life stress, social support networks, temperament,

and caretaking arrangements will be completed. In addition,

each mother-infant pair will visit the laboratory two times,

at two weeks apart, in order to participate in a 20 minute

session in which the baby's reaction to the mother, a friendly


. I











stranger, and being briefly left alone will be filmed.

POTENTIAL RISKS OR DISCOMFORTS: The only potential risk to

subjects for participation in this study would be the

relatively mild distress experienced by some infants during

separation from their mothers for a three minute period.

Mothers will be allowed to watch their infants from behind a

mirror during this time and may terminate the study at any

point, if desired.

POTENTIAL BENEFITS TO YOU OR TO OTHERS: There are no clear

benefits to be gained by experimental subjects. The primary

benefits of the research involve increasing our understanding

of the factors involved in infant social behavior, and

indirectly, to improve the quality of preventative mental

health care in this important area.

GENERAL CONDITIONS: I understand that I will receive ten

dollars for my participation in this study. I understand that

I will not be charged additional expenses for my participation

in this study. I understand that I am free to withdraw my/my

child's consent and discontinue participation in this research

project at any time without this decision affecting my child's

medical care. In the event of my/my child's sustaining a

physical injury which is proximately caused by this

experiment, no professional medical care received at the J.

Hillis Miller Health Center exclusive of hospital expenses

will be provided me without charge. It is understood that no

form of compensation exists other than those described above.

All data obtained from this research will remain

confidential. The University of Florida will protect the








64



confidentiality of this document and your records from this

research to the extent provided by law.



I have fully explained to the

nature and purpose of the above-described procedure and the

benefits and risks that are involved in its performance. I

have answered and will answer all questions to the best of

my ability. I may be contacted at telephone #


Signature of person obtaining consent Date



I have been fully informed of the above-described procedure

with its possible benefits and risks and I have received a

copy of this description. I have given permission for my/my

child's participation in this study.


Signature of relative, parent, or guardian Date


Witness to signature


Date















APPENDIX B
STRUCTURED INTERVIEW

Baby's birthdate Sex M F

Mother's age Father's age

Ages of other children living in the home

Names and relationships of others living in the home

Level of education: Mother Father

Occupation: Mother_ Father

Describe any changes in the father's employment status since

the baby's birth (such as promotion, job change, laid off)



Mother returned to work since the birth? y n

If so, on what date? Occupation?

How many hours weekly? Still working?

If not, stopped working when?

Who regularly (7 hours or more daily, 4 or more days per week)

cares for the baby when the baby is awake? mother

father relative_ babysitter or neighbor

day care center_ two share equally

On the average, how many hours weekly does the mother care for

the baby when the baby is awake? Father?

On the average, how much time weekly is the baby with a

caregiver who is not the parent?

Have you changed babysitters/day care centers in the past year

for any reason? y n # times___ Reason








66


Has either parent missed work in the past year due to

caretaking problems? y n # days mother # days

father Reasons

Have you cancelled any social engagements because of

caretaking problems? y n # times __ Reason

Has the child ever been separated from the mother for a period

of 24 hours or longer (such as for a vacation,

hospitalization, etc.)? y n If so, date

Reason__ Length of separation

Caretaker during this time

Does your child regularly have contact with other infants or

young children? y n Circumstances

Total family income














APPENDIX C
INSTRUCTIONS TO PARENTS

Duration
of Episode Description of Episode

30 seconds 1. MOTHER, BABY, AND EXPERIMENTER

Experimenter will accompany you and your
child to the playroom, and show you your
seat.

3 minutes 2. MOTHER AND BABY

Try to interest your baby in toy play,
then return to your chair. Respond
naturally and appropriately to your
child's initiatives, but try not to
initiate interaction on your own.

3 minutes 3. MOTHER, BABY, AND STRANGER

First minute: stranger sits quietly
Second minute: stranger chats with
mother
Third minute: stranger begins to play
with baby

3 minutes 4. BABY AND STRANGER
(or less)
Knock on window signals your departure.
Walk quietly to the observation room
door and leave the room, closing the
door behind you.

3 minutes 5. MOTHER AND BABY

Talk loudly before opening the door so
baby knows you're coming. Pause in
doorway to allow baby to greet you.
Comfort and soothe your child, if
necessary. When baby is ready, try to
reinterest him/her in toy play, and
return to your chair.

3 minutes 6. BABY ALONE
(or less)
Knock on window signals your departure.
Walk quietly to the observation room
door, say "good-bye" to your baby, and
leave the room, closing the door.








68


3 minutes 7. BABY AND STRANGER
(or less)
Stranger will try to comfort baby, if
needed.

3 minutes 8. MOTHER AND BABY

Talk loudly before opening the door so
baby knows you're coming. Pause in
doorway to allow baby to greet you.
Comfort and soothe your child, if
needed. When baby is ready, try to
reinterest him/her in toy play, and
return to your chair.

Remain in the room; we will return
shortly to end the session.


















REFERENCES

Ainsworth, M.D. (1964). Patterns of attachment behavior shown
by the infant in interaction with his mother. Merrill-Palmer
Quarterly, 10, 51-58.

Ainsworth, M.D. (1973). The development of infant-mother
attachment. In Betty Caldwell & H. Ricciuti (eds.), Review of
child development research (Vol. 3). Chicago: University of
Chicago Press, 1-94.

Ainsworth, M.D. (1979a). Attachment as related to mother-
infant interaction. In J.S. Rosenblatt, R.A. Hinde, C.Beer, &
M.C. Busnel (Eds.), Advances in the study of behavior, vol. 9.
New York: Academic Press, 1-51.

Ainsworth, M.D. (1979b). Infant-mother attachment. American
Psychologist, 34, 932-937.

Ainsworth, M.D., Bell, S.M., & Stayton, D.J. (1971).
Individual differences in strange situation behavior of one-
year-olds. In H.R. Schaffer (Ed.), The origins of human
social relations. London: Academic Press, 17-57.

Ainsworth, M.D., Bell, S.M., & Stayton, D.J. (1972).
Individual differences in the development of some attachment
behaviors. Merrill-Palmer Quarterly, 18, 123-143.

Ainsworth, M.D., Blehar, M.C., Waters, E., & Wall, S. (1978).
Patterns of attachment: A psychological study of the strange
situation. Hillsdale, N.J.: Lawrence Erlbaum Associates,
Inc.

Ainsworth, M.D., & Wittig, B.A. (1969). Attachment and
exploratory behavior of one-year-olds in a strange situation.
In B.M. Foss (Ed.), Determinants of infant behaviour IV.
London: Methuen, 111-153.

Arend, R., Gove, F.L., & Sroufe, L.A. (1979). Continuity of
individual adaptation from infancy to kindergarten: A
predictive study of ego-resiliency and curiosity in
preschoolers. Child Development, 50, 950-959.

Bartko, J. (1966). The intraclass correlation coefficient as
a measure of reliability. Psychological Reports, 19, 3-11.











Belsky, J., Garduque, L., & Hrncir, E. (1984). Assessing
performance, competence, and executive capacity in infant
play: Relations to home environment and attachment.
Developmental Psychology, 20, 406-417.

Belsky, J., Rovine, M., & Taylor, D. (1984). The Pennsylvania
infant and family development project, III: The origins of
individual differences in infant-mother attachment: Maternal
and child contributions. Child Development, 55, 718-728.

Belsky, J., & Steinberg, L.D. (1978). The effects of day
care: A critical review. Child Development, 49, 929-949.

Bowlby, J. (1958). The nature of a child's tie to his mother.
International Journal of Psychoanalysis, 39, 350-373.

Bowlby, J. (1969). Attachment and loss. Volume 1:
Attachment. New York: Basic Books.

Bowlby, J. (1973). Attachment and loss. Volume 2:
Separation. New York: Basic Books.

Bowlby, J. (1977). The making and breaking of affectional
bonds. I. Aetiology and psychopathology in the light of
attachment theory. British Journal of Psychiatry, 130, 201-
210.

Cairns, R.B. (1966). Attachment behavior of mammals.
Psychological Review, 73, 409-426.

Cassidy, J., & Ainsworth, M.D. (1983). The toddler's ability
to negotiate the environment: Its relation to quality of play
and quality of attachment. Presented at the Society for
Research in Child Development, Detroit.

Chase-Lansdale, P.L. (1981). Maternal employment in a family
context: Effects on infant-mother and infant-father
attachments. Dissertation Abstracts International, 42, 2562-A.

Cohen, J. (1960). A coefficient of agreement for nominal
scales. Educational and Psychological Measurement, 1, 37-46.

Connell, D.B. (1978). Individual differences in attachment:
Long-term stability and relationships to early language
development. Dissertation Abstracts International, 38,
3954-B.

Connell, J.P., & Goldsmith, H. (1982). A structural modeling
approach to the study of attachment and strange situation
behaviors. In R. Emde & R. Harmon (Eds.), The development of
attachment and affiliative systems. New York: Plenum Press,
213-243.











Crockenberg, Susan B. (1981). Infant irritability, mother
responsiveness, and social support influences on the security
of infant-mother attachment. Child Development, 52, 857-865.

Crockenberg, S. (1985). Predicting infant attachment from
early and current maternal behavior. Paper presented to the
Society for Research in Child Development, Toronto.

Easterbrooks, A., & Lamb, M. (1979). The relationship between
quality of infant-mother attachment and infant competence in
initial encounters with peers. Child Development, 50, 380-
387.

Egeland, B., & Farber, A. (1984). Infant-mother attachment:
Factors related to its development and changes over time.
Child Development, 55, 753-771.

Erickson, M., Sroufe, L., & Egeland, B. (in press). The
relationship between quality of attachment and behavior
problems in preschool in a high-risk sample. In I. Bretherton
& E. Waters (Eds.), Child development monographs. Chicago:
University of Chicago Press.

Gardner, W., Lamb, M., & Thompson, R. (1985). On the
quantitative consistency of the strange situation
classification system. Unpublished manuscript.

Gewirtz, J.L. (1972). Attachment, dependence, and a
distinction in terms of stimulus control. In J.L. Gewirtz
(Ed.), Attachment and dependency. Washington, D.C.: Winston.

Green, J.H. (1981). The relationship of stress to infant-
mother attachment. Dissertation Abstracts International, 41,
3557-B.

Hamilton, E. (1981). The relationship of maternal patterns of
stress, coping, and support to quality of early infant-mother
attachment. Dissertation Abstracts International, 42, 1172-A.

Hoffman, H.S., & DePaulo, P. (1977). Behavioral control by an
imprinting stimulus. American Scientist, 65, 58-66.

Jacobson, L., & Wille, D. (1984). Influence of attachment and
separation experience on separation distress at 18 months.
Developmental Psychology, 20, 477-484.

Jacobson, J.L., Wille, D.E., Tianen, R.L., & Aytch, D.M.
(1983). The influence of infant-mother attachment on toddler
sociability. Paper presented at the Society for Research in
Child Development, Detroit.

Lamb, M.E. (1978). Qualitative aspects of mother- and father-
infant attachments. Infant Behavior and Development, 1, 265-
275.











Lamb, M., Thompson, R., Gardner, W., Charnov, E., & Estes, D.
(1984). Security of infantile attachment as assessed in the
"strange situation": Its study and biological interpretation.
The Behavioral and Brain Sciences, 7, 127-171.

Lerner, L.M., Palermo, M., Spiro III, A., & Nesselroade, J.R.
(1982). Assessing the dimensions of temperamental
individuality across the life span: The dimensions of
temperament survey (DOTS). Child Development, 53, 149-159.

Lieberman, A.F. (1977). Competence with a peer: Relations
with attachment and peer experience. Child Development, 48,
1277-1287.

Londerville, S., & Main, M. (1981). Security of attachment,
compliance, and maternal training methods in the second year
of life. Developmental Psychology, 3, 289-299.

Main, M., & Weston, D.R. (1981). The quality of the toddler's
relationship to mother and to father: Relationship to
conflict behavior and the readiness to establish new
relationships. Child Development, 52, 932-940.

Masters, J.C., & Wellman, H.M. (1974). The study of human
infant attachment: A procedural critique. Psychological
Bulletin, 81, 218-237.

Matas, L., Arend, R., & Sroufe, L.A. (1978). Continuity of
adaptation in the second year: The relationship between
quality of attachment and later competence. Child
Development, 49, 547-556.

Owen, M., Easterbrooks, A., Chase-Lansdale, L., & Goldberg, W.
(1983). Infancy to toddlerhood: Effects of maternal
employment on infant-mother and infant-father attachments.
Paper presented to the Society for Research in Child
Development, Detroit.

Owen, M., Easterbrooks, M., Chase-Lansdale, L., & Goldberg, W.
(1984). The relationship between maternal employment status
and the stability of attachments to mother and to father.
Child Development, 55, 1894-1901.

Pastor,. D.L. (1981). The quality of mother-infant attachment
and its relationship to toddlers' initial sociability with
peers. Developmental Psychology, 17, 326-335.

Rajecki, D., Lamb, M., & Obmascher, P. (1978). Toward a
general theory of infantile attachment: A comparative review
of aspects of the social bond. Behavioral and Brain Sciences,
3, 417-464.

Robertson, J. (1952). Film: A two-year-old goes to the
hospital. London: Tavistock Child Development Research Unit.











Rothbart, M.K. (1981). Measurement of temperament in infancy.
Child Development, 52, 569-578.

Sarason, I., Johnson, J., & Siegel, J. (1978). Assessing the
impact of life changes: Development of the Life Experiences
Survey. Journal of Consulting and Clinical Psychology, 46,
932-946.

Sarason, I., Levine, H., Basham, R., & Sarason, B. (1983).
Assessing social support: The Social Support Questionnaire.
Journal of Personality and Social Psychology, 44, 127-139.

Schwartz, P. (1983). Length of day-care attendance and
attachment behavior in eighteen-month-old infants. Child
Development, 54, 1073-1078.

Sroufe, L.A., & Waters, E. (1977). Attachment as an
organizational construct. Child Development, 48, 1184-1199.

Thompson, R., & Lamb, M. (1983). Security of attachment and
stranger sociability in infancy. Developmental Psychology,
19, 184-191.

Thompson, R., Lamb, M., & Estes, D. (1982). Stability of
infant-mother attachment and its relationship to changing life
circumstances in an unselected middle-class sample. Child
Development, 53, 144-148.

Thompson, R., Lamb, M., & Estes, D. (1983). Harmonizing
discordant notes: A reply to Waters. Child Development, 54,
521-524.

Vaughn, B., Egeland, B., Sroufe, L.A., & Waters, E. (1979).
Individual differences in infant-mother attachment at twelve
and eighteen months: Stability and change in families under
stress. Child Development, 50, 971-975.

Vaughn, B., Gove, F., & Egeland, B. (1980). The relationship
between out-of-home care and the quality of infant-mother
attachment in an economically disadvantaged population. Child
Development, 51, 1203-1214.

Waters, E. (1978). The reliability and stability of
individual differences in infant-mother attachment. Child
Development, 49, 483-494.

Waters, E. (1983). The stability of individual differences in
infant attachment: Comments on the Thompson, Lamb, and Estes
contribution. Child Development, 54, 516-520.

Waters, E., Vaughn, B., & Egeland, B. (1980). Individual
differences in infant-mother attachment relationships at age
one: Antecedents in neonatal behavior in an urban,
economically disadvantaged sample. Child Development, 51,
208-216.








74



Waters, E., Wippman, J., & Sroufe, L.A. (1979). Attachment,
positive affect, and competence in the peer group: Two
studies in construct validation. -Child Development, 50, 821-
829.

Willemsen, E., Flaherty, D., Heaton, C., & Ritchey, G. (1974).
Attachment behavior of one-year-olds as a function of mother
vs. father, sex of child, session and toys. Genetic
Psychology Monographs, 90, 305-324.


















BIOGRAPHICAL SKETCH


Roberta Ann Isleib was born on January 14, 1953, in

Hackensack, New Jersey. She was the second of four children

born to Janet Burdette and Charles Robert Isleib. She and her

family adapted to several cross country moves by spending time

together: reading, camping, traveling. She entered Princeton

University in 1971 with plans to prepare for medical school,

but amended these in favor of majoring in French literature

and art history. After graduation and a year of travel and

work, she attended the University of Tennessee and obtained a

master's degree in vocational rehabilitation counseling. The

following two years were spent in the foothills of the Great

Smoky Mountains as a rural rehabilitation counselor. In

1980, she moved to Florida to enter the doctoral program in

clinical psychology. She recently completed her internship at

Yale University, working as a primary clinician on an

adolescent inpatient unit. Currently, Ms. Isleib is employed

at the Yale University Health Plan, Division of Mental

Hygiene, as a postgraduate fellow.













I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.



Suzaine Bpnett Johnon, Chairman
Associate Professor of Clinical
Psychology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.



Jaes H. Johnson U
As ciate Professor of Clinical
P ychology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.



Anarew Bradlyn
Assistant Professor oiClinical
Psychology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.



Patricia Miller
Associate Professor of Psychology

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.



Randy Cart r
Associate Professor of Statistics












This dissertation was submitted to the Graduate Faculty of the
College of Health Related Professions and to the Graduate
School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.

December, 1985 L S_____
Dean, College of Health Related
Professions


Dean, Graduate School

















































3 1262 lll 554 724 llllllllltI
3 1262 08554 7924




Full Text
73
Rothbart, M.K. (1981). Measurement of temperament in infancy.
Child Development, 52, 569-578.
Sarason, I., Johnson, J., & Siegel, J. (1978). Assessing the
impact of life changes: Development of the Life Experiences
Survey. Journal of Consulting and Clinical Psychology, 46,
932-946.
Sarason, I., Levine, H., Basham, R., & Sarason, B. (1983).
Assessing social support: The Social Support Questionnaire.
Journal of Personality and Social Psychology, 44, 127-139.
Schwartz, P. (1983). Length of day-care attendance and
attachment behavior in eighteen-month-old infants. Child
Development, 54, 1073-1078.
Sroufe, L.A., & Waters, E. (1977). Attachment as an
organizational construct. Child Development, 48, 1184-1199.
Thompson, R., & Lamb, M. (1983). Security of attachment and
stranger sociability in infancy. Developmental Psychology,
J_9, 184-191.
Thompson, R., Lamb, M., & Estes, D. (1982). Stability of
infant-mother attachment and its relationship to changing life
circumstances in an unselected middle-class sample. Child
Development, 53, 144-148.
Thompson, R., Lamb, M., & Estes, D. (1983). Harmonizing
discordant notes: A reply to Waters. Child Development, 54,
521-524.
Vaughn, B., Egeland, B., Sroufe, L.A., & Waters, E. (1979).
Individual differences in infant-mother attachment at twelve
and eighteen months: Stability and change in families under
stress. Child Development, 50, 971-975.
Vaughn, B., Gove, F., & Egeland, B. (1980). The relationship
between out-of-home care and the quality of infant-mother
attachment in an economically disadvantaged population. Child
Development, 51 1203-1 21 4.
Waters, E. (1978). The reliability and stability of
individual differences in infant-mother attachment. Child
Development, 49, 483-494.
Waters, E. (1983). The stability of individual differences in
infant attachment: Comments on the Thompson, Lamb, and Estes
contribution. Child Development, 54, 516-520.
Waters, E., Vaughn, B., & Egeland, B. (1980). Individual
differences in infant-mother attachment relationships at age
one: Antecedents in neonatal behavior in an urban,
economically disadvantaged sample. Child Development, 51 ,
208-216.


APPENDIX A
INFORMED CONSENT
Informed consent to Participate in Research
J. Hillis Miller Health Center
University of Florida
Gainesville, Florida 32610
You are being asked to volunteer as a participant in a
research study. This form is designed to provide you with
information about this study and to answer any of your
questions.
TITLE: Mother-infant attachment: Stability and relationship
to family environmental variables
PROJECT DIRECTOR: Roberta A. Isleib, M.S. 392-2944
PURPOSE: The purpose of the present study is to understand
the effects of variables such as stressful life events, social
support, mother and child temperament, and caretaking
arrangements on infant social behavior and the mother-child
relationship. The study will also investigate the stability
of the classification system which is currently used to
describe infant behavior.
PROCEDURES: Each participant will be briefly visited in their
home, where the study will be explained and questionnaires
concerning life stress, social support networks, temperament,
and caretaking arrangements will be completed. In addition,
each mother-infant pair will visit the laboratory two times,
at two weeks apart, in order to participate in a 20 minute
session in which the baby's reaction to the mother, a friendly


METHOD
Subj ects
Thirty mothers and their one-year old infants served as
subjects in this study, although one mother-infant pair
completed only the initial interview and first session. The
mean age for mothers was 28.3 years (range =20 -38). One
mother was separated from her husband; all other couples had
intact marriages. The average family income was approximately
$20,000 (range = $5200 $60,000). Subjects were recruited
through several sources. These included ads in the
university newspaper and a local advertising flyer, notices
posted on bulletin boards in the teaching hospital and in
local stores, word of mouth, and a list of mothers who had
participated in a previous research project (unrelated to the
current study). Mothers were paid ten dollars for their
participation. The initial testing of the infants was
conducted within three weeks of their first birthday, with the
retest scheduled for two weeks later. The mean age for
infants at first testing was 363.5 days (range = 344 390),
and at retest was 379.9 days (range = 359 404). Thirteen
boys and seventeen girls participated in the study. Fifty-
three percent of the subjects were first born children (eight
boys and eight girls).
Materials
Interview. A structured interview was utilized to assess


RESULTS
Reliability
Two independent raters each scored 82% of the videotapes.
Ninety-eight percent agreement was achieved for ABC
classification (Kappa = 0.93). Eighty percent agreement was
reached on subgroup classification. Intraclass correlation
coefficients (Bartko, 1966) were calculated for the
interactive ratings. The values for proximity-seeking,
contact-maintaining, resistance, and avoidance were 0.93,
0.97, 0.86, and 0.90, respectively.
Reliability was also calculated between the final
consensus values used in the analyses and the ratings on a
random sample scored by Dr. Ward. Mean values for these
ratings were 0.80, 0.91, 0.92, and 0.60, respectively.
Stability
The group of infants tested twice in the same setting were
seen both times in the trailer, with no infants seen twice in
the hospital. Therefore, initial analyses were conducted to
ascertain that the effect of a particular setting did not
cloud the analyses of stability. First, a comparison of the
trailer/trailer group (TT, N = 15) and the trailer/hospital
group (TH, N = 8) was made for the distribution of the three
classifications at test and retest. Using chi-square tests,
there were no significant differences. Because the expected


Sixty-eight
percent
of
the
children
retained
their
classification over
time
(kappa
= -0.116). Next,
the
association
between
short-
term
stability
of infant-
mother
attachment
and a
number
of
family i
environmental
and
demographic variables was assessed. No significant
differences between retained versus changed classification
subjects emerged on the following variables: social support,
life stress, maternal temperament, infant temperament,
caretaking changes, separations from mother over the first
year, infant sex, birth order, parental age, and maternal
work status.
The second aim of the study was to investigate the
relationship between quality of attachment and family
environmental variables. Using the classifications from the
first session only, the securely attached infants (n = 25)
were compared to those children who were judged anxiously
attached (n = 5). There were no differences on the variables
listed above, with the exception of maternal activity level on
which mothers of securely attached infants scored lower than
mothers of anxiously attached children.
These results are interpreted as lack of support for the
contention that short-term instability in the ABC system can
be explained as an effect of returning to the same setting
twice. The overall low stability in the entire sample is
presented as evidence that research should be directed at
exploring other models for describing and quantifying quality
of mother-infant attachment.
viii


57
change to a B1 or A1 classification. B1 infants tended to
look like B2s in their second session.
Lamb, Thompson, Gardner, Charnov, and Estes (1984) have
suggested that cluster analysis does not provide strong
support for the ABC classification system. In applying
cluster analysis to five samples which varied in age tested
and cultural background, Gardner, Lamb, and Thompson
(1985) found that boundaries between groups generally
did not fall along the A/B and B/C distinctions. They
reported the C group to appear extremely heterogeneous and
that differences between some B subgroups were often greater
than differences between A and B subgroups. More
specifically, "distal" B's (B1 and B2) frequently clustered
with A infants, while "proximal" B's (B3 and B4) were often
grouped with C's. In addition, the fit of the ABC
classification and the empirically derived clusters were found
to vary according to the age and cultural background of the
sample, with the closest fit shown in an American, one-year-
old population. The authors conclude by recommending the
consideration of continuous measures comprised of (1) a
composite summing proximity-seeking and contact-maintaining
and subtracting avoidance, and (2) the resistance scale.
The general trend of the movement between subgroups in the
present study is largely consistent with Gardner et al.'s
findings using cluster analysis. In other words, a change in
classification from B1 or B2 to group A (or vice versa) may
actually be reflective of behavioral similarity and continuity


45
Test
TABLE 3
Subgroup Classification at Test and Retest
A1
A1
0
A2
0
11
1
Retest
B2 B3
1 0
B4
0
ci
0
C2
0
2
A2
0
0
0
0
0
0
0
0
(0.07)
0
11
0
0
1
3
0
0
0
0
(0.0)
4
B2
2
0
2
2
1
0
0
1
(0.14)
8
B3
1
0
1
0
5
5
1
0
(0.28)
13
B4
G
0
0
0
0
0
0
0
(0.45)
0
C1
0
0
0
0
0
1
0
0
(0.0)
1
C2
0
0
0
0
1
0
0
0
(0.03)
1
3
0
5
6
7
6
1
1
(0.03)
29
(.10)
(
1 7 ) (
.21 )(
. 24 ) (
.21 )
(.03 ) (
.03)
29


LIST OF TABLES
Page
TABLE 1 Classification of Infants at Test and Retest
in Same Context/Changed Context Groups 43
TABLE 2 Classification of Infants at Test and Retest
for the Current Study and Ainsworth's Sample... 45
TABLE 3 Subgroup Classification at Test and Retest 46
TABLE 4 Means and Standard Deviations for Interactive
Ratings 49
TABLE 5 Comparison of Means of Strange Situation
Measures for Sessions One and Two and
Correlations Between Sessions 50
vi


56
might also investigate stability at an interval longer than
two weeks. As a cautionary note, however, as the time between
test and retest increases, the risk of- confounds from
developmental changes and alterations in environmental
circumstances increases as well.
Unfortunately, very little can be concluded regarding the
second aim of this study. In this case, the large percentage
of securely attached infants weighed against the probability
of identifying differences between groups on family
environmental variables.
Keeping in mind that the sample size and group
distribution limit firm conclusions, what might these results
mean for the ABC classification system? One speculation would
be that the ABC taxonomy is not the most parsimonious and
accurate model for describing differences in quality of
attachment. Conceivably., differences within each group are
larger than stylistic differences between groups. More
specifically, infants in different subgroups may have very
different styles of responding to increases in anxiety, which
influence patterns of temporal stability. Although Ainsworth
et al. did not report subgroup membership, their most notable
finding was the movement of seven infants from avoidantly to
securely attached. In the present study, only two infants
were classified as avoidant in the first session, and both
moved to a secure classification (B1 or B2) at time two. In
terms of the B subgroups, several tentative directional trends
can be identified. B3 infants tended either to remain in that
group or move to B4. B2 infants tended to remain stable or


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
QUALITY OF ATTACHMENT:
STABILITY AND RELATIONSHIP TO FAMILY ENVIRONMENTAL VARIABLES
By
Roberta Ann Isleib
December 1 985
Chairman: Suzanne Bennett Johnson
Major Department: Clinical Psychology
The present study tested the hypothesis that the stability
of attachment classification would be increased if an infant's
recognition of the assessment context was eliminated. Twenty-
nine white, middle-class mothers and their one-year-old
infants were seen twice in the "strange situation" and
classified according to Ainsworth's ABC taxonomy. Half of the
infants were seen two times in the same setting, with the
others seen in two very different settings, presented in
counterbalanced order. It was hypothesized that controlling
for context-specific stress would decrease the occurrence of
intensified attachment behavior in session two. However, no
differences between these groups were found in number of
infants who changed classification from test to retest, or in
four interactive ratings.
Given these findings, the two groups were collapsed in
order to examine short-term stability in the group as a whole.
vii


APPENDIX B
STRUCTURED INTERVIEW
Baby' s birthdate Sex M F
Mother 1s age Father 's age
Ages of other children living in the home
Names and relationships of others living in the home
Level of education: Mother Father
Occupation: Mother Father
Describe any changes in the father's employment status since
the baby's birth (such as promotion, job change, laid off)
Mother returned to work since the birth? y n
If so, on what date? Occupation?
How many hours weekly? Still working?
If not, stopped working when?
Who regularly (7 hours or more daily, 4 or more days per week)
cares for the baby when the baby is awake? mother
father relative babysitter or neighbor
day care center two share equally
On the average, how many hours weekly does the mother care for
the baby when the baby is awake? Father?
On the average, how much time weekly is the baby with a
caregiver who is not the parent?
Have you changed babysitters/day care centers in the past year
for any reason? y n # times Reason


26
out-of-home care and family intactness (presence of a male
figure in the household predicted secure attachment). These
findings are in line with those of Schwartz (1983), in
suggesting that full-time maternal employment may result in
avoidant patterns of behavior.
Clearly the study conducted by Vaughn et al. (1980)
implicates social support as an important mediating variable
in the formation of secure attachment. Crockenberg (1981)
studied 48 mother-infant pairs to assess the influence of
infant irritability, mother responsiveness, and social support
on attachment. Babies were administered the NBAS at days five
and ten following birth and evaluated in the strange situation
at 12 months. Home visits at three months were conducted to
observe maternal responsiveness (measured by time to respond
to infant distress signals), as well as to assess sources of
stress and support for the mothers through interviews. The
measure of social support combined both an estimate of number
of support persons available to the mother, as well as her
experience of significant stressors over the first year of
her infant's life. Low social support was found to be
significantly related to anxious attachment, for babies who
were rated as highly irritable. Similarly, low responsiveness
was related to insecure attachment in the high irritable
infant group only.
In further analyses of these data, Crockenberg (1985)
found that maternal responsiveness at 12 months had a greater
effect on irritability in the strange situation than did the


29
over both short and long time intervals, and the importance of
stability for the predictive validity of the classification
system, the study will address the issue of the contribution
of contextual effects to the stability of attachment over a
two week test-retest period* It is hypothesized that
controlling for context-specific stress will decrease the
occurrence of intensified attachment behavior in session two.
In other words, the stability of the ABC classification is
expected to be significantly higher than that reported by
Ainsworth et al. (1978) for infants seen in two different
contexts. In addition, interactive behavior categories are
expected to demonstrate significant stability across sessions,
in particular, for changed context subjects.
Family environmental variables have not been examined in
other research investigating short-term stability. Factors
which have been hypothesized to play a role in long term (six
to eight month) stability will be examined in relation to the
current sample. These variables will include maternal
employment, continuity of caregiving, and life stress.
Many of these same variables have been implicated in the
quality of mother-infant attachment which develops over the
first year. The second aim of this study is to investigate
the relationship of stressful life events, perceived social
support, infant temperament, maternal temperament, and
stability of caretaking arrangements to attachment
classification.
Given the research on environmental factors which has
been presented, it is predicted that changes in caretaking


63
stranger, and being briefly left alone will be filmed.
POTENTIAL RISKS OR DISCOMFORTS: The only potential risk to
subjects for participation in this study would be the
relatively mild distress experienced by some infants during
separation from their mothers for a three minute period.
Mothers will be allowed to watch their infants from behind a
mirror during this time and may terminate the study at any
point, if desired.
POTENTIAL BENEFITS TO YOU OR TO OTHERS: There are no clear
benefits to be gained by experimental subjects. The primary
benefits of the research involve increasing our understanding
of the factors involved in infant social behavior, and
indirectly, to improve the quality of preventative mental
health care in this important area.
GENERAL CONDITIONS: I understand that I will receive ten
dollars for my participation in this study. I understand that
I will not be charged additional expenses for my participation
in this study. I understand that I am free to withdraw my/my
child's consent and discontinue participation in this research
project at any time without this decision affecting my child's
medical care. In the event of my/my child's sustaining a
physical injury # which is proximately caused by this
experiment, no professional medical care received at the J.
Hillis Miller Health Center exclusive of hospital expenses
will be provided me without charge. It is understood that no
form of compensation exists other than those described above.
All data obtained from this research will remain
confidential. The University of Florida will protect the


performance in other arenas of infant competence. Studies
which have been conducted to evaluate the predictive validity
of the ABC classification system have uniformly supported the
more adequate performance of securely attached children on a
variety of social interaction and competence measures. Secure
infants (group B) obtained higher scores than anxious infants
(A and C) on measures of peer competence (Lieberman, 1977;
Pastor, 1981; Waters, Wippman, & Sroufe, 1979), imaginative
play (Matas, Arend, & Sroufe, 1978; Belsky, Garduque, &
Hrncir, 1984), ego resiliency (Arend, Gove, & Sroufe, 1979),
compliance and cooperation (Londerville & Main, 1981),
stranger sociability (Main & Weston, 1981), negotiation of the
environment (Cassidy & Ainsworth, 1983), and toleration of
brief separations at 18 months (Jacobson & Wille, 1984). Two
studies utilized subgroup classifications and found that B12
infants were more independent and interactive during play with
peers than either B3 or B4 children (Easterbrooks & Lamb,
1979), as well as more social with a friendly stranger than
both B34 and C subjects (Thompson & Lamb, 1983). Although
Jacobson, Wille, Tianen, and Aytch (1983) found ambivalent
infants to be more interactive with peers, they interpreted
this as the lack of expected or age-appropriate wariness. In
a longitudinal study, high-risk preschool children who had
maintained stable attachment classifications from 12 to 18
months were independently rated for behavior problems
(Erickson, Sroufe, & Egeland, in press). Group C infants were
rated as being less confident and having poorer social skills


37
blind to the experimental hypotheses of the study. Training
was undertaken using videotapes from an unrelated study. The
two raters each scored half of the session one and half of the
session two tapes, and were blind to the behavior and
classification of the infants in their other session. The
raters then scored the other half of the videotapes. Tapes
were rerated in groups of four if there was any disagreement
on group or subgroup classification, or if ratings of
interactive behaviors differed by more than two points.
Independent reliability was achieved with Dr. Mary Jo Ward who
scored a random sample of the tapes (N = 15). She was blind
to the hypotheses of the study, as well as to the behavior of
any infant in a previous session. Dr. Ward had been trained
by Dr. Alan Sroufe (Institute of Child Development,
Minneapolis, Minnesota) and had demonstrated consistent
reliability with his ratings.


12
than secure children, and Group A infants were more dependent
and less socially skilled. In general, anxiously attached
children as a group were more likely to be rated by observers
and teachers as having a variety of behavior problems.
In spite of the impressive global consistency of this
research, a major question has been raised which remains to be
addressed. In their critique of the strange situation
classification system, Connell and Goldsmith (1982) noted that
many predictive validity studies lumped ambivalent and
avoidant children into one large "insecure" category, when no
behavioral differences emerged between the groups on dependent
measures. In contrast, clear differences between the groups
did emerge in patterns of behavior in the strange situation.
Only 25% of the studies referred to above found differences
along the lines of the ABC classification, and none of these
were consistently interpretable based on theoretical
predictions. On the other hand, several authors have found
differences between subgroups (Easterbrooks & Lamb, 1979;
Thompson & Lamb, 1983). Given the difficulty researchers have
had in finding behavioral differences between the main groups,
these finer distinctions are puzzling. These inconsistencies
may in fact be reflecting inadequacies in the classification
system, rather than a lack of differences in the children.
Alternatively, differences between the three groups which are
clear in early life (e.g., Ainsworth et al., 1978; Belsky,
Rovine, & Taylor, 1984), may be moderated by later
environmental factors such that the ABC classification loses


proximity-seeking, contact-maintaining, search for mother, and
crying. Avoidance behaviors decreased in the second session
while resistance remained the same across time. In
interactions with the mother, proximity-seeking, contact-
maintaining, crying, searching, and resisting the stranger
were positively correlated across the two sessions, while
avoidance behaviors were not. As expected, frequency measures
of discrete behaviors such as smiling, vocalizing, and
touching did not demonstrate significant stability. The
overall stability of the ABC classification was 57% (kappa =
0.054), with the major changes accounted for by the switch of
all infants classified as Group A in the first session to
Group B in the second session. More specifically, ten of the
23 infants tested changed attachment groups upon retesting.
As Connell and Goldsmith (1982) have noted, the percentage of
correct reclassification expected by chance is 54%, which the
test-retest results only marginally exceeded. Ainsworth et al.
maintained that the differences in classification reflected
the oversensitivity of the system to short-term changes in the
infants' behavioral patterns (induced by the stress of
retesting), rather than reflecting inconsistencies in or
instability of individual differences in the organization of
attachment relationships. This explanation has apparently
been accepted by succeeding researchers, who have concentrated
later efforts on investigating the stability of the system
over longer intervals.


To my father
Charles Robert Isleib


QUALITY OF ATTACHMENT:
STABILITY AND RELATIONSHIP TO FAMILY ENVIRONMENTAL VARIABLES
BY
ROBERTA ANN ISLEIB
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1985


55
The conclusion which emerges from these results is that
classification of attachment quality over a two week period
lacks stability. In contrast to Ainsworth's findings, but in
line with Lamb's, this instability appeared random in
direction, rather than following a trend to normative or
secure attachment.
This study does have limitations which constrain
overgeneralization of results. The inclusion of a treatment
group comprised of infants seen twice in the hospital setting
would have eliminated the need for analyses of specific
setting effects. In addition, the relatively small sample
size and the securely weighted distribution of infants
presented difficulties for statistical analysis.
Overriding these concerns was the fact that none of the
statistical comparisons were even marginally significant.
Converging evidence from both classification status and
interactive ratings strongly supported the finding of lack of
stability. Further research to document these results should
include a larger sample, both to increase the power of
statistical evaluation and avoid the problems raised by the
ABC sample distribution. There is little reason to believe,
however, that the presence of more anxiously attached infants
in the first session would improve stability estimates.
Both the Ainsworth data and results of long-term studies
(e.g., Thompson et al., 1983) suggest that anxiously attached
children are less likely to retain their classification in a
retest than are secure infants. Finally, future research


3
attachment behaviors such as signalling or following, which
serve to bring the attachment figure into closer proximity,
Ainsworth (1979b) has suggested that sensitive
responsiveness is the quality most likely to foster secure
attachment. She maintained that if the mother-figure
responded appropriately to the infants signals across the
first year of life, the infant would develop a working model
of the mother as responsive, reliable, and accessible. She
further emphasized that in addition to protection from
predation, attachment facilitates the use of the mother-figure
as a secure base for exploration of the environment. In other
words, an infant who feels confident of the accessibility of
the caretaker is able to expand the range of the set-goal and
engage in adaptive play, exploration, and problem-solving.
Sroufe and Waters (1977) have elaborated on Ainsworth's
conception of attachment, viewing it as an affective tie
between mother and infant, as well as a behavioral system.
Differences in attachment should theoretically be
distinguishable through observation of qualitative differences
in organization of behavior, rather than discrete behaviors.
As many so-called attachment behaviors are functionally
interchangeable, attachment patterns can only be assessed by
reference to the organization of attachment behaviors with
respect to the caregiver and in consideration of context"
(Sroufe & Waters, 1977, p. 1188). These authors defined the
term affective bond as the infant's experience of security-
insecurity, which is influenced by a variety of factors such
as context, preceding events, and developmental level. The


42
TABLE 1
Classification of Infants at Test and Retest
in Same Context/Changed Context Groups
Test A
B
C
Trailer/Trailer Group
Retest
B C
1 0
A
0
1
0
1
(0.07)
10
1
12
(0.80)
2
0
2
(0.13)
1
(0.
07)
1
3
(0.
86)
1
(0.
07)
15
Test A
Hospital/Trailer and Trailer/Hospital Group
Retest
A B C
0 1 0
10
2 12
(0.14) (0.86)
0
(0.0)
1
(0.07)
12
(0.86)
1
(0.07)
14


18
relationship, although stable conditions did not appear to be
sufficient for improvements in interaction.
In a larger sample (N = 189) of low-income mother-infant
dyads which included the 100 subjects reported above, Egeland
and Farber (1984) found that 60% of the sample retained the
same attachment classification from 12 to 18 months (kappa =
0.299). In this study, as contrasted with the smaller sample,
virtually no relationship was found between stressful life
events and quality of attachment. The only significant
finding suggested that mothers of girls rated as resistant at
12 months reported higher life stress scores. In addition,
mothers of infants who changed from B to C reported increased
life stress as compared with mothers of stable B babies.
As part of a study investigating differences between
attachment to mothers versus fathers, Main and Weston (1981)
assessed changes in attachment classification at 12 and 20
months. These infants were from predominantly middle-class
homes and were selected for absence of birth complications,
no major separations from the parent within two months of the
assessments (two weeks or more), and 25 hours or less per week
of out-of-home daycare experiences. Fifteen infants were seen
with their mothers at both points, with the remainder seen
with fathers. The authors found that 80% of this sample was
classified into a similar category at both times (kappa =
0.69), although a fourth category called "unclassifiable" was
also utilized. (The unclassifiable category consisted of
infants who demonstrated conflicted or inconsistent behavioral


54
In a further attempt to understand the pattern of
instability which emerged in the present study, family
environmental variables which have been implicated in long
term stability were examined. It was not expected that gross
changes in the factors measured would have occurred between
sessions, and in fact, the questionnaires were completed prior
to both assessments. However, based on six to eight month
test-retest studies, it seemed that certain variables such as
infant difficulty, life stress, or caregiving continuity might
affect short-term stability. Clearly, there was no evidence
to support these hypothesized relationships.
To summarize, the current study did not support the
contention that short-term instability in the ABC
classification system was related to stress-inducing effects
of returning twice to the same assessment context.
Furthermore, instability in this sample as a whole was greater
than that found in the Ainsworth sample. As Lamb (1978) had
reported with a mother/father test-retest study, there was no
predictable pattern in the direction of changes which
occurred. Infants were as likely to change from secure to
anxious as they were from anxious to secure. Finally, none of
the personality or environmental variables which were
investigated differentiated between infants who changed versus
retained classification. As mentioned previously, due to the
small number of subjects, the changed classification group
could not be analyzed according to direction of change.


22
interval, using the same parent at both sessions (Ainsworth et
al., 1978). Although changes in patterns of behavior due to
the stress of a second assessment were predicted,
classification changes were largely accounted for by the
movement of avoidant infants to the secure category. The
relatively low percentage of infants who maintained similar
group membership suggests that the assessment paradigm may be
more sensitive than expected to temporally-contiguous
stressors involving separation, and less reflective of a
stable, enduring relationship quality. The two other studies
employing similar time intervals yielded contradictory
results: Lamb (1978) found no carryover from Session 1 to 2,
while Willemsen et al. (1974) found an increase in attachment
behaviors and distress from session 1 to 2.
Long-term (six month) test-retest research has produced
mixed results. The most striking feature of this research is
the wide variability in the estimates of stability which were
produced (kappa values ranged from 0.033 to 0.926). Two of
the studies reporting kappa values above 0.6 (Connell, 1976;
Main & Weston, 1981) used unstandardized variations of the
classification methods suggested by Ainsworth and her
colleagues. Vaughn et al. (1979) produced results reflecting
a lower proportion of stability, as predicted for a less
advantaged population. Their findings indicating that anxious
attachment and changes from secure to anxious attachment were
associated with high stress scores support this view.
Unfortunately, these findings were not replicated in their


35
cabinet.
Procedure
Prior to the first assessment, the subjects were
interviewed by the principal investigator, either in their
homes or at the psychology clinic at Shands Teaching Hospital.
The procedures of the study were explained and informed
consent was obtained (Appendix A). Next, the structured
interview was conducted and mothers were instructed in the
procedures of the four inventories. Finally, the mothers' role
in the strange situation was outlined and each was given a
descriptive handout.
All subjects were tested in two sessions of the Ainsworth
Strange Situation, scheduled at two week intervals. Although
scheduling all children for assessments at the same time of
day proved impossible, an attempt was made to maintain
consistency in time of assessment between the two sessions of
each subject. For half of the subjects, the testing sessions
occurred in two different experimental rooms (at the research
trailer and Shands Teaching Hospital). The remaining subjects
were tested twice in the Clinical Psychology research trailer.
For subjects tested in two different settings, the order of
the experimental rooms was counterbalanced. When the subjects
arrived at the prescribed setting, the procedures of the study
were reviewed with them by the principal investigator or a
research assistant, and the mothers were given a cue sheet to
take into the room with them (Appendix C). Following each
session of the strange situation, mothers were asked (1) if


68
3 minutes 7. BABY AND STRANGER
(or less)
Stranger will try to comfort baby, if
needed.
3 minutes 8. MOTHER AND BABY
Talk loudly before opening the door so
baby knows you're coming. Pause in
doorway to allow baby to greet you.
Comfort and soothe your child, if
needed. When baby is ready, try to
reinterest him/her in toy play, and
return to your chair.
Remain in the room; we will return
shortly to end the session.


32
demographic variables and the stability of caretaking
arrangements (see Appendix B). This interview was modeled
after one done by Thompson (personal communication, November,
1983) and included questions concerning the number of
babysitters and day care centers utilized, number of changes
in these arrangements, and number of planned events (e.g.,
social engagements, days of employment) which were disrupted
due to problems with caretaking provisions.
Social Support Questionnaire. Social support for the
mothers was assessed using the Social Support Questionnaire
(Sarason, Levine, Basham, & Sarason, 1983). This instrument
was composed of 27 hypothetical situations for which the
subject was asked to list potential support people and rate
satisfaction with that level of support. The questionnaire
yielded both a measure of total number of social supports (N)
and perceived satisfaction with support (S). A test-retest
study conducted over a four week period found correlations of
0.90 for N and 0.83 for S (Sarason et al., 1983).
Life Experiences Survey. Maternal life stress was
measured with a modified version of the Life Experiences
Survey (Sarason, Johnson, & Siegel, 1978), which produced an
estimate of the number of negative and positive life events
occurring over the previous year. This instrument was
selected because of the wide variety of events it lists,
including those concerning employment stressors. Subjects
simply identified events which had occurred in their lives
over the past year, rating them as either negative or
positive. Pearson product-moment correlations computed over a


25
children demonstrated more avoidant behavior during the final
reunion episode of the strange situation than did children in
either of the other groups. Attachment classifications and
their relationship to this finding were not reported. To
summarize, the three previous studies, all of which utilized
middle-class samples, did not produce evidence of strong
support for the relationship between quality of attachment and
maternal employment.
In contrast, Vaughn, Gove, and Egeland (1980) found that
regular, daily separations occasioned by out-of-home care in
the first year of life were associated with anxious-avoidant
attachment. These authors hypothesized that physical
inaccessibility of the mother might mirror psychological
inaccessibility in its effects on quality of attachment;
i.e., out-of-home care might contribute to anxious attachment
patterns. Three groups of mother-infant pairs from a
disadvantaged population were selected to meet criteria of
out-of-home care initiated before 12 months of age (n = 34),
out-of-home care begun between 12 and 18 months (n = 18), and
a home-only care control group (n = 52). A variety of
maternal psychological variables, measures of mother-infant
interactions at three and six months, and infant behavioral
observations did not differentiate the three groups. From
assessments with the strange situation at 12 and 18 months,
there emerged a significant relationship between work status
and attachment security, with more anxiously-avoidant infants
appearing in the early return to work group. At 18 months,
security of attachment was related both to early initiation of


5
the mother's return to the room and the stranger's departure.
Of particular interest are the baby's response to the mother
upon her return and their subsequent interaction. When the
infant is sufficiently calm, the mother again leaves the room,
leaving the child alone for episode 6. Again, search
behaviors, vocalizations, and crying are noted. In episode 7,
the stranger reenters the room and attempts to comfort and
reengage the infant in play. The baby's response to the
stranger is compared with his interactions with the mother in
previous reunion episodes. Finally, the mother returns,
pausing long enough to allow the baby time for spontaneous
reaction, and the stranger leaves. Each of the episodes lasts
approximately three minutes, with the exception of episode 1,
which lasts approximately 30 seconds. In early research,
trained observers viewed the process of the strange situation
through one-way mirrors and dictated running narratives
throughout, concentrating on the behaviors of the babies and
their interactions with both mother and stranger. The strange
situation was initially developed and standardized on four
samples, consisting of a total of 106 white, middle-class
subjects from the Baltimore area (Ainsworth et al., 1978).
Two observers were used for all but one of the studies, with
both narratives utilized in coding data. Four types of
measures were drawn from the strange situation narratives.
These included frequency counts of specific behaviors, number
of 15 second intervals in which behaviors occurred, scoring of
interactive behavioral variables along a seven point
continuum, and classification of infants into three groups and -


30
circumstances will be negatively related to security of
attachment. Measures of social support and maternal
adaptiveness should show positive, though perhaps smaller
associations to secure attachment. The measure of maternal
temperament (including adaptiveness) will be utilized to
approximate the dimension of sensitive responsiveness. High
life stress scores may be related to anxious attachment,
although support for this hypothesis has been less clear.
Finally, irritable infant temperament is expected to be
related to anxious attachment.


REFERENCES
Ainsworth, M.D. (1964). Patterns of attachment behavior shown
by the infant in interaction with his mother. Merrill-Palmer
Quarterly, 10, 51-58.
Ainsworth, M.D. (1973). The development of infant-mother
attachment. In Betty Caldwell & H. Ricciuti (eds.), Review of
child development research (Vol. 3). Chicago: University of
Chicago Press, 1-94.
Ainsworth, M.D. (1979a). Attachment as related to mother-
infant interaction. In J.S. Rosenblatt, R.A. Hinde, C.Beer, &
M.C. Busnel (Eds.), Advances in the study of behavior, vol. 9.
New York: Academic Press, 1-51.
Ainsworth, M.D. (1979b). Infant-mother attachment
Psychologist, 34, 932-937.
American
Ainsworth, M.D., Bell, S.M., & Stayton, D.J. (1971).
Individual differences in strange situation behavior of one-
year-olds. In H.R. Schaffer (Ed.), The origins of human
social relations. London: Academic Press, 17-57.
Ainsworth, M.D., Bell, S.M., & Stayton, D.J. (1972).
Individual differences in the development of some attachment
behaviors. Merrill-Palmer Quarterly, 18, 123-143.
Ainsworth, M.D., Blehar, M.C., Waters, E., & Wall, S. (1978).
Patterns of attachment: A psychological study of the strange
situation.
Inc.
Hillsdale, N.J.: Lawrence Erlbaum Associates,
Ainsworth, M.D., & Wittig, B.A. (1969). Attachment and
exploratory behavior of one-year-olds in a strange situation.
In B.M. Foss (Ed.), Determinants of infant behaviour IV.
London: Methuen, 111-153.
Arend, R., Gove, F.L., & Sroufe, L.A. (1979). Continuity of
individual adaptation from infancy to kindergarten: A
predictive study of ego-resiliency and curiosity in
preschoolers. Child Development, 50, 950-959.
Bartko, J. (1966). The intraclass correlation coefficient as
a measure of reliability. Psychological Reports, 19, 3-11.


17
subgroups at both times. Waters interpreted this consistency
as support for the stability of attachment, conceptualized as
patterns of interactive behavior affected by variations in
context.
Vaughn, Egeland, Sroufe, and Waters (1979) investigated
the stability of attachment in a lower-class, urban sample (N
= 100) in order to assess the impact of stress and changing
life circumstances on the mother-infant relationship. As in
the previous study, infants were observed in the strange
situation at 12 and 18 months. In addition, mothers completed
a stressful life-events inventory to document events occurring
between the 12 to 18 month period. Although stability of
classification at the two assessments did reach the level of
significance (kappa = 0.321), considerably more children
changed groups than in the Waters (1978) study (38%, as
compared with 4%). In the evaluation of the relationship
between attachment and life events, avoidant and resistant
infants were grouped together and compared with securely
attached infants. Mothers of securely attached infants at 18
months reported fewer stressful life events than did mothers
of anxious infants. Furthermore, changes from secure to
anxious attachment were associated with higher stressful-
events scores when compared with subjects who maintained their
secure classification over the six month period. Changes in
the opposite direction (from anxious to secure attachment)
were not related to stressful events scores. Vaughn et al.
suggested that these data confirmed that stressful life
circumstances may have seriously disrupted the mother-child


20
Waters (1983) has taken issue with the interpretation
offered by Thompson et al. for the considerable attachment
instability found in their middle-class sample. Maintaining
that the other studies (reviewed above) demonstrated that
attachment stability is a well-replicated finding, he
questioned the validity of these results, as the Thompson et
al. percentage of similar classification (53%) was only
slightly higher than that expected by chance, and was
considerably lower than the rate found by Vaughn et al. in a
disadvantaged sample (62%). Waters denied that his sample had
been specifically selected to reduce the effects of stressful
life circumstances, and therefore results so dissimilar from
the Thompson et al. sample would not have been predicted. He
suggested that the small sample size of the second study and
possible differences in scoring techniques may have resulted
in these disparate findings.
In answer to Waters, Thompson, Lamb, and Estes (1983)
have questioned whether attachment stability can be assumed
from the data presented to date. These authors dismissed the
results of the Connell (1976) and Main and Weston (1981)
studies on the basis of incomparable scoring and
classification techniques. They related their results to
differences in the types of stressful events experienced by
the families studied, rather than the frequency of events, as
assessed by Vaughn et al. Changes in patterns of caregiving
and the work status of the primary attachment figure would be
hypothesized to have impact on quality of attachment, as found


8
briefly described. Infants in subgroup A1 displayed the most
conspicuous avoidance of their mothers, in association with
minimal proximity-seeking and contact-maintaining. Subgroup
A2 infants demonstrated a mixed response consisting of both
approach and avoidance. Of the four B subgroups, subgroup B1
babies showed the most distance interaction with their
mothers, and tended not to be distressed during separation.
B2 infants resembled those in group B1, except for an increase
in proximity-seeking. In subgroup B3, the infants were most
likely to seek physical contact with their mothers, while
demonstrating very little avoidance or resistance. (Subgroup
B3 contained the largest number of infants from the
development sample and was considered by the authors to be
the normative group, reflecting the most harmonious mother-
infant relationship.) Subgroup B4 infants appeared preoccupied
with their mothers throughout the strange situation, seeking
contact and proximity, but in a less effective fashion than
B3 babies. Subgroup Cl infants were distinguishable by their
angry appearance, marked both by strong attempts to gain
contact and conspicuous resistant behaviors. Finally, C2
babies appeared passive, involving themselves in limited
exploration and interaction, and using signalling behaviors to
gain proximity to their mothers. Their resistant behaviors
were not as visible as C1 infants. Although the
classification of infants into subgroups has not been
consistently reliable, Ainsworth et al. have emphasized that
instructing raters to attend to these detailed criteria
improved the overall reliability of the three group


19
patterns which did not fit readily into one of the other
groups.) This stability was statistically significant, as was
stability of all interactive behavior scores, with the
exception of contact resistance in the mother sample. Only
proximity avoidance with fathers was significantly stable.
Thompson, Lamb, and Estes (1982) addressed the stability
of attachment patterns in an unselected middle-class sample,
maintaining that Waters' (1978) subjects had been selected for
minimal family stress and change. Forty-three infants were
observed in the strange situation at ages 12.5 and 19.5
months, and their mothers completed questionnaires regarding
changes in caregiving and family stability over the
experimental period. Fifty-three percent of the subjects
maintained their original classification status at the second
assessment (kappa = 0.033), and 26% of the children were also
classified into the same subgroup. Insecurely attached
infants at 12.5 months tended to change groups more often than
did securely attached infants. Changes in status were
significantly related to mothers returning to work and the
experience of caregiving by someone other than the mother.
However, changes in classification were bidirectional; i.e.,
infants experiencing these changes were equally likely to
shift from secure to insecure, as from insecure to secure.
Life events other than caregiving circumstances were not
related to attachment, suggesting that the critical variables
were those most likely to affect the ongoing quality (or
quantity) of the mother-infant relationship.


4
degree of security which the infant experiences mediates both
proximity-seeking and exploration.
The Strange Situation
Ainsworth and her colleagues (Ainsworth & Wittig, 1969;
Ainsworth, Bell, & Stayton, 1971, 1972; Ainsworth, Blehar,
Waters, & Wall, 1978) have developed a standardized laboratory
paradigm to assess patterns of attachment in one to two year
old infants, called the "strange situation." The procedure
involves the infant, the mother, and a stranger, participating
in eight increasingly stressful episodes in which exploratory
behavior, response to the mothers absence, response to the
stranger, and response to reunion are observed. In episode 1,
the mother and baby are introduced to the experimental room by
an observer, who disappears shortly after this introduction.
During episode 2, the mother is instructed to attempt to
interest the infant in the available toys, and to refrain from
interactions other than responding to the infant's overtures.
Episode 3 involves the first appearance of the stranger, who
sits quietly, then chats with the mother, then becomes more
involved with the child. Finally, the mother leaves the room.
In episode 4, the stranger attempts to engage the baby in
interaction, intervening if the baby becomes distressed at the
mother's absence by distracting him with a toy, talking to
him, or picking him up. The primary behaviors of interest are
the amount of exploration that the baby engages in, as
compared with that occurring when the mother was present, as
well as reaction to separation (e.g., search behaviors,
crying), and response to the stranger. Episode 5 consists of


24
Steinberg concluded that very little evidence exists to
suggest that day care results in deleterious effects on the
infant-mother bond. They cautioned, however, that the
majority of studies have investigated high-quality,
university-affiliated centers which may not be representative
of the modal American environment. Similarly, Ainsworth and
her colleagues concluded that the majority of studies
exploring the effects of working versus non-working mothers
produced negative results. These authors have suggested that
the critical variable may be the stability and continuity of
alternative caregiving arrangements, rather than the working
vs. non-working mother dichotomy.
Several studies have addressed this issue since the
publication of these reviews. Chase-Lansdale (1981) studied
single- and dual-wage-earner families, and found a similar
proportion of secure attachments in both groups. Sons of
working mothers did demonstrate fewer secure attachments to
their fathers, as well as fewer instances of secure attachment
to both parents. Owen, Easterbrooks, Chase-Lansdale, and
Goldberg (1983) also evaluated the relationship between
attachment to each parent and maternal employment. No
differences in the proportion of secure attachments were found
between groups of children whose mothers were employed full
time, part-time, or not at all. Finally, Schwartz (1983)
studied attachment behavior at 18 months in children who
attended day care on a part-time or full-time basis, and those
who did not attend at all. She found that full-time day care


60
Lamb, Thompson, Belsky, Crockenberg). The authors who have
been most critical of the methodology while continuing to be
extremely involved in the area are Michael Lamb and his
students. In general, critiques of the taxonomy tend to be
ignored or to generate considerable heat (e.g., Thompson et
al., 1982; Waters, 1983; Thompson et al., 1983). It seems
noteworthy that the major review paper produced by Lamb et al.
(1984) which has a format specifically designed for peer
commentary, stimulated no response from any of the Ainsworth
related group.
Frequently, the focus of the controversy over this
methodology comes to rest on the scoring system for the
strange situation videotapes. In several discussion sections,
results which contradict the ABC model were deemphasized or
dismissed on the basis of inadequate scoring (e.g., Waters,
1983; Ainsworth et al., 1978). Practically speaking, learning
the scoring system is a cumbersome, time-consuming, and
expensive task for researchers who are not collaborating with
a major figure. Clearly, working only from the written
material provided in the Ainsworth et al. book is inadequate
for establishing reliability. Waters (1978) made reference to
a set of standardized videotapes which would be available for
training and reliability. In fact, these were never
developed, presumably because of disagreements among the
prospective authors (Waters, personal communication, 1985).
Given how firmly entrenched the ABC system has become in the
attachment literature, the standardization of the methodology
is critical. Resolution of the controversy surrounding the


10
resistance to the stranger, crying, and reduced exploratory
behavior. The authors concluded that these analyses validated
the utility of their classification system, although the
discrimination of C from non-C infants was most troublesome,
based on the low percentage of correct classification on cross
validation.
The multiple discriminant function analysis described
above suffers from several statistical flaws, and thus the
conclusion drawn by Ainsworth et al. regarding its validation
of their typology is questionable. First, as pointed out by
Connell and Goldsmith (1982), the variables entered into the
analysis were selected on the basis of their previously
identified ability to discriminate between the groups.
Therefore, the discriminant analysis was guaranteed to make
this differentiation and should have correctly classified
close to 100% of the sample. Second, the number of subjects
used in the analysis was considerably smaller than optimal,
considering the total number of variables involved. Given the
small number of subjects utilized, considerable instability of
the results might be expected, and less success predicted if a
replication was performed using a new data set. Finally, the
serial cross-validation procedure used by the authors was
designed to derive, rather than test, the functions, and was
therefore not adequate to assess their value in predicting
group membership for a new sample.
Predictive Validity of the ABC Typology
The long-range utility of the ABC typology depends on its
ability to differentially predict concurrent and future


UNIVERSITY OF FLORIDA
iiiimiiiiiii
3 1262 08554 7924


23
larger study (Egeland & Farber, 1984), in which life stress
was not found to be consistently related to anxious attachment
or changes from secure to anxious attachment.
The Thompson et al. (1982) data also present
interpretative difficulties. Although their speculations
regarding the effects of caregiving circumstances on
attachment seem reasonable, the disparity between their sample
and the similar population studied by Waters is puzzling. As
Waters did not present data on the life circumstances of his
subjects, no firm conclusions can be drawn. Furthermore, Owen
et al. (1984) found a change in employment after the first
year to be related to more stable attachment, which is
inconsistent the Thompson et al. results. In general, these
data do not elucidate orderly conclusions about either the
stability of attachment classifications or their association
with life stress or changes in caregiving arrangements and
maternal work status.
Effects of Family Environmental Variables
Although debate continues regarding the role of
environmental variables in the formation of secure attachment,
several have received considerable attention. Research
investigating these factors, including maternal employment
status, infant temperament, social support, maternal
responsiveness, and life stress, will be addressed in this
section.
Both Belsky and Steinberg (1978) and Ainsworth et al.
(1978) have reviewed the available literature focusing on the
relationship of attachment security and day care. Belsky and


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Suzatfine Bpnett Johnson, Chairman
Associate Professor of Clinical
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Andrew Bradlyn
Assistant Professor
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Associate Professor of Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
y Carter
Randy Carter
Associate Professor of Statistics


6
eight subgroups, according to behavioral similarities and
differences -
Coding in these studies was undertaken by at least two
independent coders who had been carefully trained.
Reliability checks conducted on eight subjects in the second
study revealed reliability coefficients above 0.93 for four
frequency measures. Although frequency counts and interval
estimations were thought to describe attachment behaviors, the
authors felt they did not adequately represent interactional
behavior. Also, since a variety of attachment behaviors are
theoretically interchangeable in terms of functional
significance, frequency counts of single behaviors might
inaccurately reflect the activation of the infant's attachment
system. Furthermore, as the behavior of the mothers could not
be tightly controlled, a system which could be sensitive to
comparisons across situations in spite of individual
differences needed to be devised. Scores of interactive
behavior were designed with these concerns in mind. Six
behavioral variables were described by seven-point scales
which were anchored by behavioral descriptions garnered from
the actual narrative reports of the subjects in samples one
and two. These variables included proximity and contact
seeking, contact maintaining, resistance, avoidance, search,
and distance interaction. Five reliability checks from all of
the samples revealed interscorer reliability coefficients
ranging from 0.75 to 0.97, with a median value of 0.93.
Finally, a classification system was devised by grouping
infants according to their behavioral similarities in as many


7
of the strange situation episodes as possible. Three major
patterns of behavior were classified. Group A infants,
dubbed anxiously-avoidant, were characterized by a conspicuous
avoidance of interaction with their mothers in the reunion
episodes. Although they tended not to seek proximity with
their mothers, they did not demonstrate resistance to contact.
The stranger was treated similarly to the mother, and minimal
distress was evidenced during episodes of separation. Group
B, the securely attached infants, was characterized by active
proximity-seeking, active seeking of interaction with the
mother during reunion, clear preference for the mother over
the stranger, and, frequently, evidence of distress at the
mother's absence. Group C babies, called ambivalently or
resistantly anxious, displayed contact and interaction
resistance, especially during the final reunion episode.
Resistance was generally manifested by striking out or
squirming away from an available adult or by rejecting
proffered toys. They also tended to seek proximity and
maintain contact, giving an ambivalent impression. In
general, these infants appeared to behave in more maladaptive
ways (either angry or passive) than did infants classified
into the other two groups. Tests of interjudge agreement
suggested that experienced judges identified infants highly
reliably according to the overall groups. Identification of
Group C infants appeared to present the most difficulty to the
judges.
Eight subgroups were also formed from the original three.
As these have been used in subsequent research, they will be


53
remained at about the same level of intensity" (Ainsworth et
al., 1978, p.222).
The present study does not support these conclusions. Of
the nine infants who changed classification, four moved from
anxiously to securely attached, and the remainder changed from
secure to anxious. Also in contrast to the Ainsworth study,
changes in interactive behaviors across sessions were not
consistently in the expected directions. In general, the
explanation posed by other researchers to explain short-term
instability lacks validity in relation to these results: the
hypothesis that the anxiety engendered by encountering a
second "strange situation" results in increased attachment
behaviors and a general progression to secure attachment is
not supported.
As Ainsworth and her colleagues suggested, it seems
unlikely that the quality of the mother-infant attachment
would change dramatically over a two week period, unless there
were gross changes in the factors which influence the mother-
infant interaction. Post-experimental questions administered
after each assessment in the current study provided anecdotal
evidence to the effect that infants who changed
classifications had not experienced any noteworthy disruptions
or trauma in the two week interim, relative to those infants
who remained in their original group. Mothers occasionally
reported illness in the child or an unrestful nap, but these
comments were equally distributed across groups and time of
testing.


36
they were surprised by any aspect of their child's behavior,
and (2) whether anything unusual had occurred on the day of
testing which might have influenced the assessment.
Three different female strangers were utilized; the order
of their appearance was varied randomly. The strangers were
recruited from an undergraduate psychology class and they
received course credit for their participation in the study.
They engaged in four training sessions prior to beginning the
experiment. Although these students were familiar with
attachment theory in general, they were unaware of the
specific hypotheses of the study and were blind to the
behavior of each baby in any prior session. (Both the
experimental settings and stranger behavior were reviewed and
approved by Dr. Ross Thompson from the University of Nebraska
following the initial training sessions.)
The infants' behavior in the strange situation was
videotaped from behind one-way mirrors. Measures which were
scored from these records included classification into
groups and subgroups and ratings of proximity and contact
seeking, contact-maintaining, avoidance, and resistance during
the two reunion episodes. These four interactive behaviors
have been identified previously as most salient and reliable
in differentiating patterns of attachment (Waters, personal
communication, April, 1985). Descriptive ratings were
provided by Ainsworth et al. (1978), with seven points on each
scale anchored to specific patterns of response.
Two raters (RAI and CRI) were trained by Dr. Everett
Waters in the scoring of the strange situation. Rater CRI was


34
overtures by a stranger were noted. The episodes were
designed to gradually increase the stress to which the infant
was exposed, with concomitant increases in attachment
behaviors expected. Strange situation assessments occurred in
two different experimental rooms, one located in Shands
Teaching Hospital and the other in the Department of Clinical
Psychology research trailer. These facilities were located
several miles apart and in very different settings, chosen to
minimize the children's recognition of the assessment
situation. The trailer was located on a quiet drive in a
woodsy setting, isolated from the rest of the university. In
contrast, the room at Shands Teaching Hospital was located
within a large medical center at a busy intersection of the
campus.
Each of the experimental rooms contained two chairs and a
number of toys appropriate for one-year-old infants. Toys at
the trailer included toy phone, foil pie plates, plastic
hammer, milk bottle with shapes, baby doll, musical apple,
felt chicken, plastic shapes, plastic toy rabbit, and toy keys
on a ring. Toys at the hospital consisted of large colored
ball, plastic bucket and shovel, colored plastic blocks, toy
camera music box, chicken rattle, cloth pig, plastic pull
train, cloth doll with rattle, bird squeeze toy, fish rattle,
and soft rubber elephant. The experimental room at the
hospital measured 10 by 12 feet, was carpeted, and contained a
couch and two small end tables in addition to the equipment
described above. The room at the trailer was smaller,
measuring 8 by 6, uncarpeted, and furnished only with a file


9
classification. Also, the infants in the different subgroups
are hypothesized to have experienced different patterns of
caregiving, such that their attachment differs qualitatively.
In order to test the significance of the multivariate
differences between the three major categories of attachment,
a multiple discriminant function analysis was conducted. This
was also intended to explore whether the variables identified
in the classification instructions were those that best
discriminated the groups. The 73 original variables were
reduced to 22 by eliminating those which did not distinguish
between two groups at the 0.01 level of significance (using
analyses of variance), and those which did not contribute
significantly to 2-group discriminant function analyses (A vs.
B, and B vs. C). One hundred and five subjects were utilized
in the development sample, with two significant discriminant
functions emerging. The discriminant scores successfully
predicted classification of 96% of A infants, 91% of B
infants, and 92% of C infants. A serial cross-validation was
then conducted using unknown subjects from the development
sample. In this procedure, 92% of A infants, 85% of B
infants, and 69% of C infants were correctly classified.
The first discriminant function served to distinguish A
from non-A infants, and the strongest contributing variables
included proximity-seeking and contact-maintaining to the
mother, absence of resistance, interactive behavior with the
stranger, and exploratory behavior and crying. The second
discriminant function, labeled C vs. non-C, correlated highly
with resistance to mother, proximity- and contact-seeking,


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ERPQQY01L_WHYM8T INGEST_TIME 2012-09-24T14:25:57Z PACKAGE AA00011823_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


33
five to six week test-retest period were found to range from
0.19 to 0.53 for positive life change and from 0.56 to 0.88
for negative life change (Sarason et al., 1978).
Infant Behavior Questionnaire. The Infant Behavior
Questionnaire, designed to measure infant temperament, yielded
six dimensions of temperament including activity level,
distress to approach novel stimuli (or fear), distress to
limitations, duration of orienting, soothability, and smiling
and laughter (Rothbart, 1981). The instrument required
frequency judgements on approximately 90 behavioral items
which might have occurred over the previous one to two weeks.
Stability coefficients over the three to 12 month age bracket
ranged from 0.06 to 0.80, with greater stability reported
above six months (Rothbart, 1981).
Dimensions of Temperament Survey. Maternal temperament
was measured with the Dimensions of Temperament Survey
(Lerner, Palermo, Spiro, & Nesselroade, 1982). This
instrument consisted of 34 behavioral items which subjects
marked as either true (accurate self-descriptions) or false
(inaccurate). The five temperamental factors identified by
this instrument were activity level, attention
span/distractibility, adaptability, rhythmicity, and
reactivity. Three day test-retest stability ranged from 0.60
to 0.93 (Lerner et al., 1982).
The Strange Situation. As described in the introduction,
the strange situation consisted of eight short episodes, in
which the infant's response to separation, reunion, and


This dissertation was submitted to the Graduate Faculty of the
College of Health Related Professions and to the Graduate
School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
December, 1985
Dean, College of Health Related
Professions
Dean, Graduate School


71
Crockenberg, Susan B. (1981). Infant irritability, mother
responsiveness, and social support influences on the security
of infant-mother attachment. Child Development, 52, 857-865.
Crockenberg, S. (1985). Predicting infant attachment from
early and current maternal behavior. Paper presented to the
Society for Research in Child Development, Toronto.
Easterbrooks, A., & Lamb, M. (1979). The relationship between
quality of infant-mother attachment and infant competence in
initial encounters with peers. Child Development, 50, 380-
387.
Egeland, B., & Farber, A. (1984). Infant-mother attachment:
Factors related to its development and changes over time.
Child Development, 55, 753-771.
Erickson, M., Sroufe, L., & Egeland, B. (in press). The
relationship between quality of attachment and behavior
problems in preschool in a high-risk sample. In I. Bretherton
& E. Waters (Eds.), Child development monographs. Chicago:
University of Chicago Press.
Gardner, W., Lamb, M., & Thompson, R. (1985). On the
quantitative consistency of the strange situation
classification system. Unpublished manuscript.
Gewirtz, J.L. (1972). Attachment, dependence, and a
distinction in terms of stimulus control. In J.L. Gewirtz
(Ed.), Attachment and dependency. Washington, D.C.: Winston.
Green, J.H. (1981). The relationship of stress to infant-
mother attachment. Dissertation Abstracts International, 41,
3557-B.
Hamilton, E. (1981). The relationship of maternal patterns of
stress, coping, and support to quality of early infant-mother
attachment. Dissertation Abstracts International, 42, 1172-A.
Hoffman, H.S., & DePaulo, P. (1977). Behavioral control by an
imprinting stimulus. American Scientist, 65, 58-66.
Jacobson, L., & Wille, D. (1984). Influence of attachment and
separation experience on separation distress at 18 months.
Developmental Psychology, 20, 477-484.
Jacobson, J.L., Wille, D.E., Tianen, R.L., & Aytch, D.M.
(1983). The influence of infant-mother attachment on toddler
sociability. Paper presented at the Society for Research in
Child Development, Detroit.
Lamb, M.E. (1978). Qualitative aspects of mother- and father-
infant attachments. Infant Behavior and Development, 1_, 265-
275.


72
Lamb, M., Thompson, R., Gardner, W., Charnov, E., & Estes, D.
(1984). Security of infantile attachment as assessed in the
"strange situation": Its study and biological interpretation.
The Behavioral and Brain Sciences, l_t 127-171.
Lerner, L.M., Palermo, M., Spiro III, A., & Nesselroade, J.R.
(1982). Assessing the dimensions of temperamental
individuality across the life span: The dimensions of
temperament survey (DOTS). Child Development, 53, 149-159.
Lieberman, A.F. (1977). Competence with a peer: Relations
with attachment and peer experience. Child Development, 48,
1 277-1287. ""
Londerville, S., & Main, M. (1981). Security of attachment,
compliance, and maternal training methods in the second year
of life. Developmental Psychology, 289-299.
Main, M., & Weston, D.R. (1981). The quality of the toddler's
relationship to mother and to father: Relationship to
conflict behavior and the readiness to establish new
relationships. Child Development, 52, 932-940.
Masters, J.C., & Wellman, H.M. (1974). The study of human
infant attachment: A procedural critique. Psychological
Bulletin, 81 218-237.
Matas, L., Arend, R., & Sroufe, L.A. (1978). Continuity of
adaptation in the second year: The relationship between
quality of attachment and later competence. Child
Development, 49, 547-556.
Owen, M., Easterbrooks, A., Chase-Lansdale, L., & Goldberg, W.
(1983). Infancy to toddlerhood: Effects of maternal
employment on infant-mother and infant-father attachments.
Paper presented to the Society for Research in Child
Development, Detroit.
Owen, M., Easterbrooks, M., Chase-Lansdale, L., & Goldberg, W.
(1984). The relationship between maternal employment status
and the stability of attachments to mother and to father.
Child Development, 55, 1894-1901.
Pastor, D.L. (1981). The quality of mother-infant attachment
and its relationship to toddlers initial sociability with
peers. Developmental Psychology, 17, 326-335.
Rajecki, D., Lamb, M., & Obmascher, P. (1978). Toward a
general theory of infantile attachment: A comparative review
of aspects of the social bond. Behavioral and Brain Sciences
3, 417-464.
Robertson, J. (1952). Film: A two-year-old goes to the
hospital. London: Tavistock Child Development Research Unit.


DISCUSSION
The present study investigated the hypothesis that the
stability of attachment classification would be increased if
an infant's recognition of the assessment context was
eliminated. In order to evaluate this assumption, half of the
subjects were tested in two very different settings, presented
in counterbalanced order, with the remainder seen twice in the
same context. Results demonstrated that context effects had
no bearing on classification stability. The coefficients of
agreement (kappa) for the two groups (same context and changed
context) were similarly unimpressive, and nonsignificantly
different. This finding was supported by the lack of
differences between the groups in interactive ratings and
changes in those ratings across time. Although the relatively
small sample size and the large proportion of securely
attached infants found in the study preclude taking a
definitive stand on the meaning of these findings, it should
be noted that none of the statistical comparisons were even
close to significantly different. The lack of differences
between these groups suggests that if memory for the strange
situation is affecting classification stability, it is likely
to be recognition of the entire sequence of events rather than
of the physical environment alone that affects the children's
behavior in the second assessment.


40
scrutinized. With one exception, no differences were noted on
t-tests comparing groups TT and TH in terms of proximity
seeking, contact-maintaining, resistance, or avoidance in
either reunion episode, during both test and retest.
Avoidance in episode 5 of the retest was significantly higher
for infants in the TT group than the TH group (t = 2.131; p <
0.0489). Given the relatively large number of statistical
tests employed, as well as the lack of support for this
finding in other measures (and theoretically), it seems likely
that this is not a meaningful finding. Difference scores on
each interactive variable calculated across sessions (e.g.,
Davoid5 = Avoidance Episode 5, test Avoidance Episode 5,
retest) were also examined using t-tests. There were no
significant differences.
As with the overall group classification, comparisons
between the TH and HT groups were next made. There were no
differences between these two groups on measures of any of the
four interactive variables in each reunion in both test and
retest sessions. Several significant findings were noted in
difference scores. Babies in the TH group tended to increase
proximity-seeking in episodes 5 and 8 over the two sessions,
as compared with decreases in the HT group (t = -2.8820, p <
0.0138, and t = -2.3039, p < 0.0399, respectively). In
addition, avoidance in episode 8 tended to decrease across
sessions in the TH group, as compared with increases in the HT
group (t = 2.3569, p < 0.0363). All other difference score
comparisons were nonsignificant.


46
avoidance decreased. In contrast, the only significant change
across episodes in the retest was an increase in contact-
maintaining from five to eight. Combined means for episodes
five and eight for each session, as well as correlations
between these means, can be found in Table 5. Differences
between the means of the two sessions were nonsignificant for
all interactive ratings.
Further paired comparison t-tests were computed for
subjects sorted into two groups according to whether they
changed (n = 9) or retained (n = 20) their classifications
across sessions. Infants who changed classifications showed
increases from episode five to eight (in session one) in
proximity-seeking, contact-maintaining and resistance (p <
0.0327, 0.0112, and 0.0099, respectively). During the retest,
only a decrease in proximity-seeking from episode five to
eight was significant (p < 0.0404). For those infants who
retained their classification, proximity-seeking and contact-
maintaining increased across the episodes in the first session
(p < 0.0024 and p < 0.0009), while avoidance decreased (p <
0.0116). During the second session, the only significant
difference across episodes was an increase in contact-
maintaining (p < 0.0218).
Given the pattern of instability across sessions, the
association between short-term stability of infant-mother
attachment and family environmental and demographic variables
was assessed. For these analyses, subjects were sorted into
those who retained their classification over the two sessions


21
in the middle-class sample. Thompson et al. suggested that
middle-class families might have better resources to cope with
such stressors in a positive way, than would disadvantaged
families.
Owen, Easterbrooks, Chase-Lansdale, and Goldberg (1984)
studied quality of attachment at 12 and 20 months to both
mothers and fathers in relation to the mothers' employment
status. Seventy-eight percent of the children showed no
change in attachment classification with mothers (kappa =
0.174), while 62 % of the attachments remained similar with
fathers. Attachment classification was shown to be more
stable in children whose mothers returned to work during their
second year of life, as compared with mothers who maintained
full-time, part-time, or nonemployment over the 20 month
period. The authors suggested that concerns over the effects
of the change in employment status may have motivated the
mothers to maintain continuity in their relationships with
their infants. In addition, the authors cautioned that the
subjects were drawn from a highly educated, middle-class,
white sample, which limits its generalizability. Finally, the
incongruence of these results with those of Thompson et al.,
may relate to the timing of return to work and assessment of
attachment. In other words, a disruption in the relationship
may occur for a short period following change in employment,
which could inflate the overall estimate of attachment
instability.
To summarize, only one study to date has addressed test-
retest stability of the strange situation over a short


39
cell frequencies were lower than optimal for the chi-square
statistic, comparisons were also made with Fisher's exact
test. For this comparison, the distribution of A and C
infants considered as one group was compared with group B,
with no significant differences found between the treatment
groups. The same two groups (TT/TH) were examined for
differences in number of infants who retained the same
classification over the two sessions. Again, a chi-square
test revealed no significant differences.
As a further check for setting differences, all infants
seen in the trailer at session one (n = 23) were compared to
those seen in the hospital (n = 6). There were no differences
between groups using the Fisher's exact test to examine
distribution of infants into A plus C versus B categories. T-
tests also demonstrated no differences in interactive ratings
during session one between these groups.
Next, the group of infants tested in the order
trailer/hospital (TH) was compared with that tested first in
the hospital, followed by the trailer (HT, N = 6). Chi-square
analyses revealed no differences between these groups in the
distribution of classifications either at the first or second
session. In addition, Fisher's exact tests demonstrated no
differences at test or retest in distribution of subjects into
A plus C versus B categories. Likewise, there were no
differences between these groups in the number of infants who
retained their classification across sessions.
In addition to examining the effects of setting on overall
group membership,
the four interactive variables
were


INTRODUCTION
In 1958, Bowlby published an article entitled "The Nature
of the Child's Tie to his Mother" which was the first
exposition of the now widely recognized Bowlby-Ainsworth
ethological theory of attachment. Although many other
theories of attachment have been proposed (e.g., Cairns,
1966; Hoffman & DePaulo, 1977; Gewirtz, 1972), none have
generated as much empirical research or received as much
support from empirical evaluation, as has the Bowlby-Ainsworth
theory (Rajecki, Lamb, & Obmascher, 1978). Some of the focus
on this theory may be attributed to methodological
consistency, in that the "strange situation" paradigm
developed by Ainsworth and colleagues (Ainsworth, Blehar,
Waters, & Wall, 1978) has been widely adopted as the
attachment classification system.
The Bowlby-Ainsworth Theory of Attachment
In developing his theory of attachment, Bowlby drew on
ethological notions of instinctive behavior, control systems
theory, and Freud's concept of object relations, seeking an
explanatory framework for naturalistic observations of
children who had been separated from their mother-figures
(Bowlby, 1958; 1969; 1973). Previous research (e.g.,
Robertson, 1952) had noted that hospitalized children
progressed through predictable stages following separation,


52
As the recognition of context hypothesis received no
support, subjects were collapsed into one group in order to
examine stability in the sample as a whole. In this study,
there was a higher percentage of infants who retained their
classification across sessions as compared with the Ainsworth
et al. (1978) sample (69% and 57%, respectively). However, a
comparison of the kappa values computed for these two samples
revealed that when the amount of agreement expected by chance
was considered, stability in the current study was even lower
than in Ainsworth's results.
Temporal instability would not necessarily undermine the
validity of the classification system, if changes in the
quality of attachment were either the predictable result of
anxiety engendered in the first session spilling over into the
second, or if the changes could be related predictably to
environmental changes. Ainsworth and her colleagues were able
to develop reasonable explanations for their findings, based
on the pattern of changes they observed. First, the bulk of
changes in classification were accounted for by avoidant
infants appearing secure in the second session. In addition,
proximity-seeking and contact-maintaining were significantly
higher in session 2, while avoidance decreased from time one
to time two. The authors stated: "it is clear that the
effect of the repetition of the strange situation after a
brief two-week interval was to increase distress and the
intensity of attachment behavior, while at the same time the
negative behaviors of avoidance and resistance decreased or


66
Has either parent missed work in the past year due to
caretaking problems? y n # days mother # days
father Reasons
Have you cancelled any social engagements because of
caretaking problems? y n # times Reason
Has the child ever been separated from the mother for a period
of 24 hours or longer (such as for a vacation,
hospitalization, etc.)? y n If so, date
Reason Length of separation
Caretaker during this time
Does your child regularly have contact with other infants or
young children? y n Circumstances
Total family income


58
among those groups, rather than suggesting temporal
instability of individual differences, as the current
classification system would dictate.
An examination of the overall pattern of interactive
rating change in the present study demonstrated that there
were significant changes in all four ratings from episode five
to eight in the first session. Two weeks later, only contact-
maintaining showed a significant increase. By sorting the
infants according to stability of classification, an attempt
was made to evaluate whether certain infants had habituated to
the stress of the strange situation, while others had not.
The two groups showed reasonably similar patterns of change in
proximity-seeking and contact-maintaining in session one,
though differing on the other two variables. Differences in
the pattern of change across episodes continued in the retest.
In a larger sample which would allow investigation of
direction of classification change, analyses along these lines
might clarify which types of infants show predictable
behavioral reactions to the stress of separation.
The final section of this paper focuses on practical and
sociological observations of the ABC classification system.
As reviewed in the introduction, the strange situation and its
ensuing ABC taxonomy are plagued with a number of
problems. Questions of both short-term and long-term
stability remain unanswered. Results of the present study
illustrate the questions surrounding short-term temporal
stability. Efforts which have been directed to understanding
six to ten month stability have produced inconsistent results.


13
predictive power. One aspect of the system which has been the
subject of some contention is stability of classification over
time. Research addressing this issue will be reviewed in the
following section.
Stability of the ABC Classification System
Much of the predictive validity of the ABC classification
system rests on claims for its demonstrated stability across
time. If infants cannot be consistently identified as
belonging to a particular group, the subsequent prediction
from group membership to their performance on a variety of
developmental tasks has little meaning.
In response to a critique by Masters and Wellman (1974)
suggesting that individual attachment behaviors had not been
found to be stable across time, Ainsworth et al. (1978)
assessed the reliability of the strange situation using a two
week test-retest procedure. They predicted that infants would
remember the first separation experience, and that distressed
infants would demonstrate more distress in the second session,
while undisturbed infants would appear even more comfortable
in the second assessment. In spite of these expected
behavioral changes from first to second assessment,
interactive behaviors were predicted to be significantly
correlated across sessions. Twenty-three white, middle-class
infants were involved in this test-retest study, conducted
when they were 50 and 52 weeks of age.
Overall, attachment behaviors increased significantly in
21 of 23 infants from the first to second session, including


QUALITY OF ATTACHMENT:
STABILITY AND RELATIONSHIP TO FAMILY ENVIRONMENTAL VARIABLES
BY
ROBERTA ANN ISLEIB
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1 985

To ny father
Charles Robert Isleib

ACKNOWLEDGMENTS
The planning and execution of this manuscript have been
extremely complex and demanding. It could not have been
completed without the assistance of many people along the way.
Thanks go first to Suzanne Johnson, Roger Blashfield, Jim
Johnson, Pat Miller, Drew Bradlyn, and Randy Carter for their
supportive and constructive comments during the designing
stages of the study. Suzanne also served as an anchor
throughout my internship year, encouraging me to continue to
make progress in spite of various setbacks and the pull to
procrastinate.
Identifying and recruiting subjects was a major hurdle,
for which the input of Jim Siwi, Keith Berg, and Janice Benton
is gratefully acknowledged. Thanks also go to the many
friends who told others about my research and recruited
subjects by word of mouth. Prior to data collection, Jerry
Rudd contributed skill and labor while installing a one-way
mirror and Dan McNeil served as a cheerful video equipment
consultant.
Ross Thompson kindly reviewed several strange situation
videotapes and made helpful suggestions about the physical
environment, stranger behavior, and timing of the episodes.
Fi Kieffer and her son Christopher, and Dawn Bowers and her
daughter Martina agreed to serve as guinea pigs during the
iii

initial training period. Special thanks are due to Penny
Phelps, Joy Colle, and Katie Kato, who played the difficult
role of the stranger. In addition, Rae Hendlin generously
filled in for one session on a last minute basis, and Roman
Urbana served as an experimenter throughout the study.
Scoring the videotapes proved to be an enormous task,
made possible initially by the training provided by Everett
Waters. Thanks are also due to Mary Jo Ward for her
reliability checks, and most certainly, to Susan Grajek,
without whom the data analysis would have seemed overwhelming.
Don Quinlan kindly allowed me access to his computer, such
that typing and editing the manuscript became manageable.
My most earnest thanks go to my father, C. Robert Isleib,
who painstakingly coded a seemingly endless number of
videotapes, and tolerated my irritability with equanimity in
times when the project lagged. This dissertation is dedicated
to him, not only for his invaluable assistance during this
year, but also in recognition of the unselfish support and
encouragement he has given me in many ways over the past 32
years.
Roberta A. Isleib
New Haven, CT.
September, 1985
IV

TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS iii
LIST OF TABLES vi
ABSTRACT vii
INTRODUCTION 1
The Bowlby-Ainsworth Theory of Attachment 1
The Strange Situation 4
Predictive Validity of the ABC Typology 11
Stability of the ABC Classification System 13
Effects of Family Environmental Variables 24
Aims and Hypotheses of the Study 29
METHOD 31
Subjects 31
Materials 31
Procedure 35
RESULTS 38
Reliability 38
Stability 38
Family Environmental Variables 50
DISCUSSION 51
APPENDIX A INFORMED CONSENT 62
APPENDIX B STRUCTURED INTERVIEW 65
APPENDIX C INSTRUCTIONS TO PARENTS 67
REFERENCES 69
BIOGRAPHICAL SKETCH 75
v

LIST OF TABLES
Page
TABLE 1 Classification of Infants at Test and Retest
in Same Context/Changed Context Groups 43
TABLE 2 Classification of Infants at Test and Retest
for the Current Study and Ainsworth's Sample... 45
TABLE 3 Subgroup Classification at Test and Retest 46
TABLE 4 Means and Standard Deviations for Interactive
Ratings 49
TABLE 5 Comparison of Means of Strange Situation
Measures for Sessions One and Two and
Correlations Between Sessions 50
vi

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
QUALITY OF ATTACHMENT:
STABILITY AND RELATIONSHIP TO FAMILY ENVIRONMENTAL VARIABLES
By
Roberta Ann Isleib
December 1985
Chairman: Suzanne Bennett Johnson
Major Department: Clinical Psychology
The present study tested the hypothesis that the stability
of attachment classification would be increased if an infant's
recognition of the assessment context was eliminated. Twenty-
nine white, middle-class mothers and their one-year-old
infants were seen twice in the "strange situation" and
classified according to Ainsworth's ABC taxonomy. Half of the
infants were seen two times in the same setting, with the
others seen in two very different settings, presented in
counterbalanced order. It was hypothesized that controlling
for context-specific stress would decrease the occurrence of
intensified attachment behavior in session two. However, no
differences between these groups were found in number of
infants who changed classification from test to retest, or in
four interactive ratings.
Given these findings, the two groups were collapsed in
order to examine short-term stability in the group as a whole.
vii-

Sixty-eight
percent
of
the
children
retained
their
classification over
time
(kappa
= -0.116). Next,
the
association
between
short-
term
stability
of infant-
mother
attachment
and a
number
of
family
environmental
and
demographic variables was assessed. No significant
differences between retained versus changed classification
subjects emerged on the following variables: social support,
life stress, maternal temperament, infant temperament,
caretaking changes, separations from mother over the first
year, infant sex, birth order, parental age, and maternal
work status.
The second aim of the study was to investigate the
relationship between quality of attachment and family
environmental variables. Using the classifications from the
first session only, the securely attached infants (n = 25)
were compared to those children who were judged anxiously
attached (n = 5). There were no differences on the variables
listed above, with the exception of maternal activity level on
which mothers of securely attached infants scored lower than
mothers of anxiously attached children.
These results are interpreted as lack of support for the
contention that short-term instability in the ABC system can
be explained as an effect of returning to the same setting
twice. The overall low stability in the entire sample is
presented as evidence that research should be directed at
exploring other models for describing and quantifying quality
of mother-infant attachment.
viii

INTRODUCTION
In 1958, Bowlby published an article entitled "The Nature
of the Child's Tie to his Mother" which was the first
exposition of the now widely recognized Bowlby-Ainsworth
ethological theory of attachment. Although many other
theories of attachment have been proposed (e.g., Cairns,
1966; Hoffman S DePaulo, 1977; Gewirtz, 1972), none have
generated as much empirical research or received as much
support from empirical evaluation, as has the Bowlby-Ainsworth
theory (Rajecki, Lamb, & Obmascher, 1978). Some of the focus
on this theory may be attributed to methodological
consistency, in that the "strange situation" paradigm
developed by Ainsworth and colleagues (Ainsworth, Blehar,
Waters, X Wall, 1978) has been widely adopted as the
attachment classification system.
The Bowlby-Ainsworth Theory of Attachment
In developing his theory of attachment, Bowlby drew on
ethological notions of instinctive behavior, control systems
theory, and Freud's concept of object relations, seeking an
explanatory framework for naturalistic observations of
children who had been separated from their mother-figures
(Bowlby, 1958; 1969; 1973). Previous research (e.g.,
Robertson, 1952) had noted that hospitalized children
progressed through predictable stages following separation,

2
characterized by active protest, despair or hopelessness, and
finally, detachment and apathy. As Ainsworth has been
instrumental in translating Bowlby's ideas into research, as
well as contributing to their elaboration and clarification
(e.g., Ainsworth, 1964; 1973; 1979a), the theory has become
known as the Bowlby-Ainsworth ethological model of attachment.
The Bowlby-Ainsworth model views attachment as a species
general behavior system whose function is to protect the
relatively helpless human infant from predation. Infants are
born with a biological predisposition to seek contact with
adults, which is manifested early on by signalling behaviors
such as crying, smiling, and calling, in addition to searching
and following behaviors as locomotion develops. Adults are
equipped with complementary predispositions to protect
infants. Bowlby originally suggested that the development of
a specific attachment was mediated by the amount of
interaction occurring between the infant and attachment
figure. As a cognitive appreciation for person permanence was
assumed to be a necessary prerequisite, the development of
attachment was not expected to occur until the second half of
the first year of life.
Drawing on control systems theory, Bowlby incorporated
the notion of a set-goal into his attachment model. The set-
goal, defined as the amount of distance from the attachment
figure which the infant will tolerate, is mediated by a
variety of internal (e.g., illness, hunger) and external
(e.g., stress, separation) factors. When the set-goal is
exceeded, the infant should respond with the activation of

3
attachment behaviors such as signalling or following, which
serve to bring the attachment figure into closer proximity.
Ainsworth (1979b) has suggested that "sensitive
responsiveness" is the quality most likely to foster secure
attachment. She maintained that if the mother-figure
responded appropriately to the infant's signals across the
first year of life, the infant would develop a working model
of the mother as responsive, reliable, and accessible. She
further emphasized that in addition to protection from
predation, attachment facilitates the use of the mother-figure
as a secure base for exploration of the environment. In other
words, an infant who feels confident of the accessibility of
the caretaker is able to expand the range of the set-goal and
engage in adaptive play, exploration, and problem-solving.
Sroufe and Waters (1977) have elaborated on Ainsworth's
conception of attachment, viewing it as an affective tie
between mother and infant, as well as a behavioral system.
Differences in attachment should theoretically be
distinguishable through observation of qualitative differences
in organization of behavior, rather than discrete behaviors.
As many so-called attachment behaviors are functionally
interchangeable, attachment patterns can only be assessed "by
reference to the organization of attachment behaviors with
respect to the caregiver and in consideration of context"
(Sroufe & Waters, 1977, p. 1188). These authors defined the
term affective bond as the infant's experience of security-
insecurity, which is influenced by a variety of factors such
as context, preceding events, and developmental level. The

4
degree of security which the infant experiences mediates both
proximity-seeking and exploration.
The Strange Situation
Ainsworth and her colleagues (Ainsworth & Wittig, 1969;
Ainsworth, Bell, & Stayton, 1971, 1972; Ainsworth, Blehar,
Waters, & Wall, 1978) have developed a standardized laboratory
paradigm to assess patterns of attachment in one to two year
old infants, called the "strange situation." The procedure
involves the infant, the mother, and a stranger, participating
in eight increasingly stressful episodes in which exploratory
behavior, response to the mother's absence, response to the
stranger, and response to reunion are observed. In episode 1,
the mother and baby are introduced to the experimental room by
an observer, who disappears shortly after this introduction.
During episode 2, the mother is instructed to attempt to
interest the infant in the available toys, and to refrain from
interactions other than responding to the infant's overtures.
Episode 3 involves the first appearance of the stranger, who
sits quietly, then chats with the mother, then becomes more
involved with the child. Finally, the mother leaves the room.
In episode 4, the stranger attempts to engage the baby in
interaction, intervening if the baby becomes distressed at the
mother's absence by distracting him with a toy, talking to
him, or picking him up. The primary behaviors of interest are
the amount of exploration that the baby engages in, as
compared with that occurring when the mother was present, as
well as reaction to separation (e.g., search behaviors,
crying), and response to the stranger. Episode 5 consists of

5
the mother's return to the room and the stranger's departure.
Of particular interest are the baby's response to the mother
upon her return and their subsequent interaction. When the
infant is sufficiently calm, the mother again leaves the room,
leaving the child alone for episode 6. Again, search
behaviors, vocalizations, and crying are noted. In episode 7,
the stranger reenters the room and attempts to comfort and
reengage the infant in play. The baby's response to the
stranger is compared with his interactions with the mother in
previous reunion episodes. Finally, the mother returns,
pausing long enough to allow the baby time for spontaneous
reaction, and the stranger leaves. Each of the episodes lasts
approximately three minutes, with the exception of episode 1,
which lasts approximately 30 seconds. In early research,
trained observers viewed the process of the strange situation
through one-way mirrors and dictated running narratives
throughout, concentrating on the behaviors of the babies and
their interactions with both mother and stranger. The strange
situation was initially developed and standardized on four
samples, consisting of a total of 106 white, middle-class
subjects from the Baltimore area (Ainsworth et al., 1978).
Two observers were used for all but one of the studies, with
both narratives utilized in coding data. Four types of
measures were drawn from the strange situation narratives.
These included frequency counts of specific behaviors, number
of 15 second intervals in which behaviors occurred, scoring of
interactive behavioral variables along a seven point
continuum, and classification of infants into three groups and

6
eight subgroups, according to behavioral similarities and
differences.
Coding in these studies was undertaken by at least two
independent coders who had been carefully trained.
Reliability checks conducted on eight subjects in the second
study revealed reliability coefficients above 0.93 for four
frequency measures. Although frequency counts and interval
estimations were thought to describe attachment behaviors, the
authors felt they did not adequately represent interactional
behavior. Also, since a variety of attachment behaviors are
theoretically interchangeable in terms of functional
significance, frequency counts of single behaviors might
inaccurately reflect the activation of the infant's attachment
system. Furthermore, as the behavior of the mothers could not
be tightly controlled, a system which could be sensitive to
comparisons across situations in spite of individual
differences needed to be devised. Scores of interactive
behavior were designed with these concerns in mind. Six
behavioral variables were described by seven-point scales
which were anchored by behavioral descriptions garnered from
the actual narrative reports of the subjects in samples one
and two. These variables included proximity and contact
seeking, contact maintaining, resistance, avoidance, search,
and distance interaction. Five reliability checks from all of
the samples revealed interscorer reliability coefficients
ranging from 0.75 to 0.97, with a median value of 0.93.
Finally, a classification system was devised by grouping
infants according to their behavioral similarities in as many

7
of the strange situation episodes as possible. Three major
patterns of behavior were classified. Group A infants,
dubbed anxiously-avoidant, were characterized by a conspicuous
avoidance of interaction with their mothers in the reunion
episodes. Although they tended not to seek proximity with
their mothers, they did not demonstrate resistance to contact.
The stranger was treated similarly to the mother, and minimal
distress was evidenced during episodes of separation. Group
B, the securely attached infants, was characterized by active
proximity-seeking, active seeking of interaction with the
mother during reunion, clear preference for the mother over
the stranger, and, frequently, evidence of distress at the
mother's absence. Group C babies, called ambivalently or
resistantly anxious, displayed contact and interaction
resistance, especially during the final reunion episode.
Resistance was generally manifested by striking out or
squirming away from an available adult or by rejecting
proffered toys. They also tended to seek proximity and
maintain contact, giving an ambivalent impression. In
general, these infants appeared to behave in more maladaptive
ways (either angry or passive) than did infants classified
into the other two groups. Tests of interjudge agreement
suggested that experienced judges identified infants highly
reliably according to the overall groups. Identification of
Group C infants appeared to present the most difficulty to the
judges.
Eight subgroups were also formed from the original three.
As these have been used in subsequent research, they will be

8
briefly described. Infants in subgroup A1 displayed the most
conspicuous avoidance of their mothers, in association with
minimal proximity-seeking and contact-maintaining. Subgroup
A2 infants demonstrated a mixed response consisting of both
approach and avoidance. Of the four B subgroups, subgroup B1
babies showed the most distance interaction with their
mothers, and tended not to be distressed during separation.
B2 infants resembled those in group B1 , except for an increase
in proximity-seeking. In subgroup B3, the infants were most
likely to seek physical contact with their mothers, while
demonstrating very little avoidance or resistance. (Subgroup
B3 contained the largest number of infants from the
development sample and was considered by the authors to be
the normative group, reflecting the most harmonious mother-
infant relationship.) Subgroup B4 infants appeared preoccupied
with their mothers throughout the strange situation, seeking
contact and proximity, but in a less effective fashion than
B3 babies. Subgroup Cl infants were distinguishable by their
angry appearance, marked both by strong attempts to gain
contact and conspicuous resistant behaviors. Finally, C2
babies appeared passive, involving themselves in limited
exploration and interaction, and using signalling behaviors to
gain proximity to their mothers. Their resistant behaviors
were not as visible as C1 infants. Although the
classification of infants into subgroups has not been
consistently reliable, Ainsworth et al. have emphasized that
instructing raters to attend to these detailed criteria
improved the overall reliability of the three group

9
classification. Also, the infants in the different subgroups
are hypothesized to have experienced different patterns of
caregiving, such that their attachment differs qualitatively.
In order to test the significance of the multivariate
differences between the three major categories of attachment,
a multiple discriminant function analysis was conducted. This
was also intended to explore whether the variables identified
in the classification instructions were those that best
discriminated the groups. The 73 original variables were
reduced to 22 by eliminating those which did not distinguish
between two groups at the 0.01 level of significance (using
analyses of variance), and those which did not contribute
significantly to 2-group discriminant function analyses (A vs.
B, and B vs. C). One hundred and five subjects were utilized
in the development sample, with two significant discriminant
functions emerging. The discriminant scores successfully
predicted classification of 96% of A infants, 91% of B
infants, and 92% of C infants. A serial cross-validation was
then conducted using unknown subjects from the development
sample. In this procedure, 92% of A infants, 85% of B
infants, and 69% of C infants were correctly classified.
The first discriminant function served to distinguish A
from non-A infants, and the strongest contributing variables
included proximity-seeking and contact-maintaining to the
mother, absence of resistance, interactive behavior with the
stranger, and exploratory behavior and crying. The second
discriminant function, labeled C vs. non-C, correlated highly
with resistance to mother, proximity- and contact-seeking,

10
resistance to the stranger, crying, and reduced exploratory
behavior. The authors concluded that these analyses validated
the utility of their classification system, although the
discrimination of C from non-C infants was most troublesome,
based on the low percentage of correct classification on cross
validation.
The multiple discriminant function analysis described
above suffers from several statistical flaws, and thus the
conclusion drawn by Ainsworth et al. regarding its validation
of their typology is questionable. First, as pointed out by
Connell and Goldsmith (1982), the variables entered into the
analysis were selected on the basis of their previously
identified ability to discriminate between the groups.
Therefore, the discriminant analysis was guaranteed to make
this differentiation and should have correctly classified
close to 100% of the sample. Second, the number of subjects
used in the analysis was considerably smaller than optimal,
considering the total number of variables involved. Given the
small number of subjects utilized, considerable instability of
the results might be expected, and less success predicted if a
replication was performed using a new data set. Finally, the
serial cross-validation procedure used by the authors was
designed to derive, rather than test, the functions, and was
therefore not adequate to assess their value in predicting
group membership for a new sample.
Predictive Validity of the ABC Typology
The long-range utility of the ABC typology depends on its
ability to differentially predict concurrent and future

performance in other arenas of infant competence. Studies
which have been conducted to evaluate the predictive validity
of the ABC classification system have uniformly supported the
more adequate performance of securely attached children on a
variety of social interaction and competence measures. Secure
infants (group B) obtained higher scores than anxious infants
(A and C) on measures of peer competence (Lieberman, 1977;
Pastor, 1981; Waters, Wippman, & Sroufe, 1979), imaginative
play (Matas, Arend, & Sroufe, 1978; Belsky, Garduque, &
Hrncir, 1984), ego resiliency (Arend, Gove, & Sroufe, 1979),
compliance and cooperation (Londerville X Main, 1981),
stranger sociability (Main X Weston, 1981), negotiation of the
environment (Cassidy X Ainsworth, 1983), and toleration of
brief separations at 18 months (Jacobson X Wille, 1984). Two
studies utilized subgroup classifications and found that B12
infants were more independent and interactive during play with
peers than either B3 or B4 children (Easterbrooks X Lamb,
1979), as well as more social with a friendly stranger than
both B34 and C subjects (Thompson X Lamb, 1983). Although
Jacobson, Wille, Tianen, and Aytch (1983) found ambivalent
infants to be more interactive with peers, they interpreted
this as the lack of expected or age-appropriate wariness. In
a longitudinal study, high-risk preschool children who had
maintained stable attachment classifications from 12 to 18
months were independently rated for behavior problems
(Erickson, Sroufe, X Egeland, in press). Group C infants were
rated as being less confident and having poorer social skills

12
than secure children, and Group A infants were more dependent
and less socially skilled. In general, anxiously attached
children as a group were more likely to be rated by observers
and teachers as having a variety of behavior problems.
In spite of the impressive global consistency of this
research, a major question has been raised which remains to be
addressed. In their critique of the strange situation
classification system, Connell and Goldsmith (1982) noted that
many predictive validity studies lumped ambivalent and
avoidant children into one large "insecure" category, when no
behavioral differences emerged between the groups on dependent
measures. In contrast, clear differences between the groups
did emerge in patterns of behavior in the strange situation.
Only 25% of the studies referred to above found differences
along the lines of the ABC classification, and none of these
were consistently interpretable based on theoretical
predictions. On the other hand, several authors have found
differences between subgroups (Easterbrooks & Lamb, 1979;
Thompson & Lamb, 1983). Given the difficulty researchers have
had in finding behavioral differences between the main groups,
these finer distinctions are puzzling. These inconsistencies
may in fact be reflecting inadequacies in the classification
system, rather than a lack of differences in the children.
Alternatively, differences between the three groups which are
clear in early life (e.g., Ainsworth et al., 1978; Belsky,
Rovine, & Taylor, 1984), may be moderated by later
environmental factors such that the ABC classification loses

13
predictive power. One aspect of the system which has been the
subject of some contention is stability of classification over
time. Research addressing this issue will be reviewed in the
following section.
Stability of the ABC Classification System
Much of the predictive validity of the ABC classification
system rests on claims for its demonstrated stability across
time. If infants cannot be consistently identified as
belonging to a particular group, the subsequent prediction
from group membership to their performance on a variety of
developmental tasks has little meaning.
In response to a critique by Masters and Wellman (1974)
suggesting that individual attachment behaviors had not been
found to be stable across time, Ainsworth et al. (1978)
assessed the reliability of the strange situation using a two
week test-retest procedure. They predicted that infants would
remember the first separation experience, and that distressed
infants would demonstrate more distress in the second session,
while undisturbed infants would appear even more comfortable
in the second assessment. In spite of these expected
behavioral changes from first to second assessment,
interactive behaviors were predicted to be significantly
correlated across sessions. Twenty-three white, middle-class
infants were involved in this test-retest study, conducted
when they were 50 and 52 weeks of age.
Overall, attachment behaviors increased significantly in
21 of 23 infants from the first to second session, including

14
proximity-seeking, contact-maintaining, search for mother, and
crying. Avoidance behaviors decreased in the second session
while resistance remained the same across time. In
interactions with the mother, proximity-seeking, contact-
maintaining, crying, searching, and resisting the stranger
were positively correlated across the two sessions, while
avoidance behaviors were not. As expected, frequency measures
of discrete behaviors such as smiling, vocalizing, and
touching did not demonstrate significant stability. The
overall stability of the ABC classification was 57% (kappa =
0.054), with the major changes accounted for by the switch of
all infants classified as Group A in the first session to
Group B in the second session. More specifically, ten of the
23 infants tested changed attachment groups upon retesting.
As Connell and Goldsmith (1982) have noted, the percentage of
correct reclassification expected by chance is 54%, which the
test-retest results only marginally exceeded. Ainsworth et al.
maintained that the differences in classification reflected
the oversensitivity of the system to short-term changes in the
infants' behavioral patterns (induced by the stress of
retesting), rather than reflecting inconsistencies in or
instability of individual differences in the organization of
attachment relationships. This explanation has apparently
been accepted by succeeding researchers, who have concentrated
later efforts on investigating the stability of the system
over longer intervals.

15
Only two other studies have repeated the strange
situation over a similarly short time interval. These results
will be briefly summarized, although both studies investigated
attachment towards mothers versus fathers, and neither
followed the Ainsworth scoring system faithfully. Willemsen,
Flaherty, Heaton, and Ritchey (1974) contrasted attachment
behaviors in one-year-old infants to mothers versus fathers.
These authors found more attachment behavior was directed to
the parent in the presence of the stranger in session 2, as
well as more distress (particularly crying) being exhibited.
As in the Ainsworth et al. test-retest study, these results
were attributed to the association of the experimental room
with the stress of separation. Lamb (1978) also investigated
patterns of attachment to mothers versus fathers during
testing sessions scheduled one week apart. Results
demonstrated no order effects; i.e., there was not a tendency
for any classification or behavior to be more or less common
in either session.
Connell (1978) tested 55 white, middle-class infants
using the strange situation at 12 and 18 months. He found
significant correlations in measures of contact-maintaining,
proximity- seeking, resistance, and avoidance across the two
sessions. Behavior towards the stranger was less consistent,
with significant correlations emerging in resistance,
avoidance, and proximity seeking. Approximately 81% of the
infants were classified in the same ABC group six months later
(kappa = 0.603), although Connell used techniques which

16
resulted in the elimination of eight subjects in subgroups B1
and B4 from the analyses.
In order to evaluate the construct validity of the
attachment concept, Waters (1978) examined the stability of
individual attachment behaviors, ratings of behavior
categories, and overall group classification in a middle-class
sample. Fifty infants were assessed at ages 12 and 18 months,
utilizing the Ainsworth strange situation paradigm. Time
samples of discrete behaviors and crying were recorded, as
well as behavioral ratings of five "attachment behavior
categories" (i.e., proximity-seeking, contact-maintaining,
proximity and interaction avoiding, contact resisting, and
distance interaction). Infants were also assigned to overall
categories (A, B, or C) as well as subgroups, based on their
strange situation performance. Behaviors were analyzed
separately for preseparation, separation, and reunion
episodes. Waters found that only four of 21 correlations
between discrete behavior scores at 12 and 18 months reached
the level of significance, and suggested that their occurrence
in the strange situation was frequently too low to allow for
adequate reliability. However, when ratings of interactive
behavior categories were considered, 13 of 18 correlations
were significant, including all of the behaviors directed at
the mothers in the reunion episodes. In terms of attachment
classification, 96% of the children maintained the same
attachment pattern over the six month period (kappa = 0.926),
while 30 of 50 subjects were also classified in the same

17
subgroups at both times. Waters interpreted this consistency
as support for the stability of attachment, conceptualized as
patterns of interactive behavior affected by variations in
context.
Vaughn, Egeland, Sroufe, and Waters (1979) investigated
the stability of attachment in a lower-class, urban sample (N
= 100) in order to assess the impact of stress and changing
life circumstances on the mother-infant relationship. As in
the previous study, infants were observed in the strange
situation at 12 and 18 months. In addition, mothers completed
a stressful life-events inventory to document events occurring
between the 12 to 18 month period. Although stability of
classification at the two assessments did reach the level of
significance (kappa = 0.321), considerably more children
changed groups than in the Waters (1978) study (38%, as
compared with 4%). In the evaluation of the relationship
between attachment and life events, avoidant and resistant
infants were grouped together and compared with securely
attached infants. Mothers of securely attached infants at 18
months reported fewer stressful life events than did mothers
of anxious infants. Furthermore, changes from secure to
anxious attachment were associated with higher stressful-
events scores when compared with subjects who maintained their
secure classification over the six month period. Changes in
the opposite direction (from anxious to secure attachment)
were not related to stressful events scores. Vaughn et al.
suggested that these data confirmed that stressful life
circumstances may have seriously disrupted the mother-child

18
relationship, although stable conditions did not appear to be
sufficient for improvements in interaction.
In a larger sample (N = 189) of low-income mother-infant
dyads which included the 100 subjects reported above, Egeland
and Farber (1984) found that 60% of the sample retained the
same attachment classification from 12 to 18 months (kappa =
0.299). In this study, as contrasted with the smaller sample,
virtually no relationship was found between stressful life
events and quality of attachment. The only significant
finding suggested that mothers of girls rated as resistant at
12 months reported higher life stress scores. In addition,
mothers of infants who changed from B to C reported increased
life stress as compared with mothers of stable B babies.
As part of a study investigating differences between
attachment to mothers versus fathers, Main and Weston (1981)
assessed changes in attachment classification at 12 and 20
months. These infants were from predominantly middle-class
homes and were selected for absence of birth complications,
no major separations from the parent within two months of the
assessments (two weeks or more), and 25 hours or less per week
of out-of-home daycare experiences. Fifteen infants were seen
with their mothers at both points, with the remainder seen
with fathers. The authors found that 80% of this sample was
classified into a similar category at both times (kappa =
0.69), although a fourth category called "unclassifiable" was
also utilized. (The unclassifiable category consisted of
infants who demonstrated conflicted or inconsistent behavioral

19
patterns which did not fit readily into one of the other
groups.) This stability was statistically significant, as was
stability of all interactive behavior scores, with the
exception of contact resistance in the mother sample. Only
proximity avoidance with fathers was significantly stable.
Thompson, Lamb, and Estes (1982) addressed the stability
of attachment patterns in an unselected middle-class sample,
maintaining that Waters' (1978) subjects had been selected for
minimal family stress and change. Forty-three infants were
observed in the strange situation at ages 12.5 and 19.5
months, and their mothers completed questionnaires regarding
changes in caregiving and family stability over the
experimental period. Fifty-three percent of the subjects
maintained their original classification status at the second
assessment (kappa = 0.033), and 26% of the children were also
classified into the same subgroup. Insecurely attached
infants at 12.5 months tended to change groups more often than
did securely attached infants. Changes in status were
significantly related to mothers returning to work and the
experience of caregiving by someone other than the mother.
However, changes in classification were bidirectional; i.e.,
infants experiencing these changes were equally likely to
shift from secure to insecure, as from insecure to secure.
Life events other than caregiving circumstances were not
related to attachment, suggesting that the critical variables
were those most likely to affect the ongoing quality (or
quantity) of the mother-infant relationship.

20
Waters (1983) has taken issue with the interpretation
offered by Thompson et al. for the considerable attachment
instability found in their middle-class sample. Maintaining
that the other studies (reviewed above) demonstrated that
attachment stability is a well-replicated finding, he
questioned the validity of these results, as the Thompson et
al. percentage of similar classification (53%) was only
slightly higher than that expected by chance, and was
considerably lower than the rate found by Vaughn et al. in a
disadvantaged sample (62%). Waters denied that his sample had
been specifically selected to reduce the effects of stressful
life circumstances, and therefore results so dissimilar from
the Thompson et al. sample would not have been predicted. He
suggested that the small sample size of the second study and
possible differences in scoring techniques may have resulted
in these disparate findings.
In answer to Waters, Thompson, Lamb, and Estes (1983)
have questioned whether attachment stability can be assumed
from the data presented to date. These authors dismissed the
results of the Connell (1976) and Main and Weston (1981)
studies on the basis of incomparable scoring and
classification techniques. They related their results to
differences in the types of stressful events experienced by
the families studied, rather than the frequency of events, as
assessed by Vaughn et al. Changes in patterns of caregiving
and the work status of the primary attachment figure would be
hypothesized to have impact on quality of attachment, as found

21
in the middle-class sample. Thompson et al. suggested that
middle-class families might have better resources to cope with
such stressors in a positive way, than would disadvantaged
families.
Owen, Easterbrooks, Chase-Lansdale, and Goldberg (1984)
studied quality of attachment at 12 and 20 months to both
mothers and fathers in relation to the mothers' employment
status. Seventy-eight percent of the children showed no
change in attachment classification with mothers (kappa =
0.174), while 62 % of the attachments remained similar with
fathers. Attachment classification was shown to be more
stable in children whose mothers returned to work during their
second year of life, as compared with mothers who maintained
full-time, part-time, or nonemployment over the 20 month
period. The authors suggested that concerns over the effects
of the change in employment status may have motivated the
mothers to maintain continuity in their relationships with
their infants. In addition, the authors cautioned that the
subjects were drawn from a highly educated, middle-class,
white sample, which limits its generalizability. Finally, the
incongruence of these results with those of Thompson et al.,
may relate to the timing of return to work and assessment of
attachment. In other words, a disruption in the relationship
may occur for a short period following change in employment,
which could inflate the overall estimate of attachment
instability.
To summarize, only one study to date has addressed test-
retest stability of the strange situation over a short

22
interval, using the same parent at both sessions (Ainsworth et
al., 1978). Although changes in patterns of behavior due to
the stress of a second assessment were predicted,
classification changes were largely accounted for by the
movement of avoidant infants to the secure category. The
relatively low percentage of infants who maintained similar
group membership suggests that the assessment paradigm may be
more sensitive than expected to temporally-contiguous
stressors involving separation, and less reflective of a
stable, enduring relationship quality. The two other studies
employing similar time intervals yielded contradictory
results: Lamb (1978) found no carryover from Session 1 to 2,
while Willemsen et al. (1974) found an increase in attachment
behaviors and distress from session 1 to 2.
Long-term (six month) test-retest research has produced
mixed results. The most striking feature of this research is
the wide variability in the estimates of stability which were
produced (kappa values ranged from 0.033 to 0.926). Two of
the studies reporting kappa values above 0.6 (Connell, 1976;
Main & Weston, 1981) used unstandardized variations of the
classification methods suggested by Ainsworth and her
colleagues. Vaughn et al. (1979) produced results reflecting
a lower proportion of stability, as predicted for a less
advantaged population. Their findings indicating that anxious
attachment and changes from secure to anxious attachment were
associated with high stress scores support this view.
Unfortunately, these findings were not replicated in their

23
larger study (Egeland & Farber, 1984), in which life stress
was not found to be consistently related to anxious attachment
or changes from secure to anxious attachment.
The Thompson et al. (1982) data also present
interpretative difficulties. Although their speculations
regarding the effects of caregiving circumstances on
attachment seem reasonable, the disparity between their sample
and the similar population studied by Waters is puzzling. As
Waters did not present data on the life circumstances of his
subjects, no firm conclusions can be drawn. Furthermore, Owen
et al. (1984) found a change in employment after the first
year to be related to more stable attachment, which is
inconsistent the Thompson et al. results. In general, these
data do not elucidate orderly conclusions about either the
stability of attachment classifications or their association
with life stress or changes in caregiving arrangements and
maternal work status.
Effects of Family Environmental Variables
Although debate continues regarding the role of
environmental variables in the formation of secure attachment,
several have received considerable attention. Research
investigating these factors, including maternal employment
status, infant temperament, social support, maternal
responsiveness, and life stress, will be addressed in this
section.
Both Belsky and Steinberg (1978) and Ainsworth et al.
(1978) have reviewed the available literature focusing on the
relationship of attachment security and day care. Belsky and

24
Steinberg concluded that very little evidence exists to
suggest that day care results in deleterious effects on the
infant-mother bond. They cautioned, however, that the
majority of studies have investigated high-quality,
university-affiliated centers which may not be representative
of the modal American environment. Similarly, Ainsworth and
her colleagues concluded that the majority of studies
exploring the effects of working versus non-working mothers
produced negative results. These authors have suggested that
the critical variable may be the stability and continuity of
alternative caregiving arrangements, rather than the working
vs. non-working mother dichotomy.
Several studies have addressed this issue since the
publication of these reviews. Chase-Lansdale (1981) studied
single- and dual-wage-earner families, and found a similar
proportion of secure attachments in both groups. Sons of
working mothers did demonstrate fewer secure attachments to
their fathers, as well as fewer instances of secure attachment
to both parents. Owen, Easterbrooks, Chase-Lansdale, and
Goldberg (1983) also evaluated the relationship between
attachment to each parent and maternal employment. No
differences in the proportion of secure attachments were found
between groups of children whose mothers were employed full¬
time, part-time, or not at all. Finally, Schwartz (1983)
studied attachment behavior at 18 months in children who
attended day care on a part-time or full-time basis, and those
who did not attend at all. She found that full-time day care

25
children demonstrated more avoidant behavior during the final
reunion episode of the strange situation than did children in
either of the other groups. Attachment classifications and
their relationship to this finding were not reported. To
summarize, the three previous studies, all of which utilized
middle-class samples, did not produce evidence of strong
support for the relationship between quality of attachment and
maternal employment.
In contrast, Vaughn, Gove, and Egeland (1980) found that
regular, daily separations occasioned by out-of-home care in
the first year of life were associated with anxious-avoidant
attachment. These authors hypothesized that physical
inaccessibility of the mother might mirror psychological
inaccessibility in its effects on quality of attachment;
i.e., out-of-home care might contribute to anxious attachment
patterns. Three groups of mother-infant pairs from a
disadvantaged population were selected to meet criteria of
out-of-home care initiated before 12 months of age (n = 34),
out-of-home care begun between 12 and 18 months (n = 18), and
a home-only care control group (n = 52). A variety of
maternal psychological variables, measures of mother-infant
interactions at three and six months, and infant behavioral
observations did not differentiate the three groups. From
assessments with the strange situation at 12 and 18 months,
there emerged a significant relationship between work status
and attachment security, with more anxiously-avoidant infants
appearing in the early return to work group. At 18 months,
security of attachment was related both to early initiation of

26
out-of-home care and family intactness (presence of a male
figure in the household predicted secure attachment). These
findings are in line with those of Schwartz (1983), in
suggesting that full-time maternal employment may result in
avoidant patterns of behavior.
Clearly the study conducted by Vaughn et al. (1980)
implicates social support as an important mediating variable
in the formation of secure attachment. Crockenberg (1981)
studied 48 mother-infant pairs to assess the influence of
infant irritability, mother responsiveness, and social support
on attachment. Babies were administered the NBAS at days five
and ten following birth and evaluated in the strange situation
at 12 months. Home visits at three months were conducted to
observe maternal responsiveness (measured by time to respond
to infant distress signals), as well as to assess sources of
stress and support for the mothers through interviews. The
measure of social support combined both an estimate of number
of support persons available to the mother, as well as her
experience of significant stressors over the first year of
her infant's life. Low social support was found to be
significantly related to anxious attachment, for babies who
were rated as highly irritable. Similarly, low responsiveness
was related to insecure attachment in the high irritable
infant group only.
In further analyses of these data, Crockenberg (1985)
found that maternal responsiveness at 12 months had a greater
effect on irritability in the strange situation than did the

27
behavior of the mothers at three months. She suggested that
mothers of irritable babies tended to grow less responsive
over the first year, particularly under conditions of low
social support.
In line with Crockenberg1s findings on irritability and
attachment, Belsky, Rovine, and Taylor (1984) evaluated the
relationship of reciprocal interaction at one, three, and nine
months to attachment classification at one year. Securely
attached infants experienced moderate levels of stimulation
over their first year of life, while avoidant infants were
overstimulated and resistant infants were understimulated.
Both types of insecure infants were found to be more fussy at
three and nine months, suggesting that infant temperament may
play a role in the development of maladaptive caretaking
patterns. Similarly, Egeland and Farber (1984) found that in
a lower SES sample, infants classified as resistantly attached
at one year had been rated as less alert and responsive during
feeding and play during their first few months of life.
Waters, Vaughn, and Egeland (1980) administered the NBAS
to 100 first-born neonates at age seven and ten days, then
administered the strange situation at one year. Results
indicated that anxious/resistant (Group C) infants scored
lower on orientation, motor maturity, and regulation items at
day seven. These differences had disappeared by day ten. The
authors suggested that these results supported other evidence
linking the more difficult nature of group C children with
difficult interaction and the development of anxious
attachment.

28
Finally, two studies investigated the relationship of
family stress, as measured by the Parenting Stress Index, and
patterns of attachment (Green, 1981; Hamilton, 1981). In
neither of these white, middle-class samples was a significant
relationship between these variables found.
To summarize, the majority of research done to date has
not found that maternal employment negatively affects security
of infant attachment. A major exception is the study by
Vaughn et al. (1980), which demonstrated that early out-of-
home care in a disadvantaged population resulted in a larger
proportion of anxiously attached children. Research
investigating long-term attachment stability (over a six month
period) has suggested that changes in attachment
classification may be associated with lower socioeconomic
status, high life stress, and changes in caregiving
circumstances. As a cautionary note, replication across
studies on these variables has been problematic, even within
research groups working with longitudinal studies. In
addition, some of the results have been directly contradictory
(e.g., Owen et al.'s (1984) finding that mothers who returned
to work in the second year had offspring with more stable
attachments, as contrasted with Thompson et al. (1982) who
found return to work to be associated with changes in
classification).
Aims and Hypotheses of the Study
The present study has two aims. First, given the rather
weak evidence available for the stability of the ABC typology

29
over both short and long time intervals, and the importance of
stability for the predictive validity of the classification
system, the study will address the issue of the contribution
of contextual effects to the stability of attachment over a
two week test-retest period. It is hypothesized that
controlling for context-specific stress will decrease the
occurrence of intensified attachment behavior in session two.
In other words, the stability of the ABC classification is
expected to be significantly higher than that reported by
Ainsworth et al. (1978) for infants seen in two different
contexts. In addition, interactive behavior categories are
expected to demonstrate significant stability across sessions,
in particular, for changed context subjects.
Family environmental variables have not been examined in
other research investigating short-term stability. Factors
which have been hypothesized to play a role in long term (six
to eight month) stability will be examined in relation to the
current sample. These variables will include maternal
employment, continuity of caregiving, and life stress.
Many of these same variables have been implicated in the
quality of mother-infant attachment which develops over the
first year. The second aim of this study is to investigate
the relationship of stressful life events, perceived social
support, infant temperament, maternal temperament, and
stability of caretaking arrangements to attachment
classification.
Given the research on environmental factors which has
been presented, it is predicted that changes in caretaking

30
circumstances will be negatively related to security of
attachment. Measures of social support and maternal
adaptiveness should show positive, though perhaps smaller
associations to secure attachment. The measure of maternal
temperament (including adaptiveness) will be utilized to
approximate the dimension of sensitive responsiveness. High
life stress scores may be related to anxious attachment,
although support for this hypothesis has been less clear.
Finally, irritable infant temperament is expected to be
related to anxious attachment.

METHOD
Subí ects
Thirty mothers and their one-year old infants served as
subjects in this study, although one mother-infant pair
completed only the initial interview and first session. The
mean age for mothers was 28.3 years (range =20 -38). One
mother was separated from her husband; all other couples had
intact marriages. The average family income was approximately
$20,000 (range = $5200 - $60,000). Subjects were recruited
through several sources. These included ads in the
university newspaper and a local advertising flyer, notices
posted on bulletin boards in the teaching hospital and in
local stores, word of mouth, and a list of mothers who had
participated in a previous research project (unrelated to the
current study). Mothers were paid ten dollars for their
participation. The initial testing of the infants was
conducted within three weeks of their first birthday, with the
retest scheduled for two weeks later. The mean age for
infants at first testing was 363.5 days (range = 344 - 390),
and at retest was 379.9 days (range = 359 - 404). Thirteen
boys and seventeen girls participated in the study. Fifty-
three percent of the subjects were first born children (eight
boys and eight girls).
Materials
Interview. A structured interview was utilized to assess

32
demographic variables and the stability of caretaking
arrangements (see Appendix B). This interview was modeled
after one done by Thompson (personal communication, November,
1983) and included questions concerning the number of
babysitters and day care centers utilized, number of changes
in these arrangements, and number of planned events (e.g.,
social engagements, days of employment) which were disrupted
due to problems with caretaking provisions.
Social Support Questionnaire. Social support for the
mothers was assessed using the Social Support Questionnaire
(Sarason, Levine, Basham, & Sarason, 1983). This instrument
was composed of 27 hypothetical situations for which the
subject was asked to list potential support people and rate
satisfaction with that level of support. The questionnaire
yielded both a measure of total number of social supports (N)
and perceived satisfaction with support (S). A test-retest
study conducted over a four week period found correlations of
0.90 for N and 0.83 for S (Sarason et al., 1983).
Life Experiences Survey. Maternal life stress was
measured with a modified version of the Life Experiences
Survey (Sarason, Johnson, & Siegel, 1978), which produced an
estimate of the number of negative and positive life events
occurring over the previous year. This instrument was
selected because of the wide variety of events it lists,
including those concerning employment stressors. Subjects
simply identified events which had occurred in their lives
over the past year, rating them as either negative or
positive. Pearson product-moment correlations computed over a

33
five to six week test-retest period were found to range from
0.19 to 0.53 for positive life change and from 0.56 to 0.88
for negative life change (Sarason et al., 1978).
Infant Behavior Questionnaire. The Infant Behavior
Questionnaire, designed to measure infant temperament, yielded
six dimensions of temperament including activity level,
distress to approach novel stimuli (or fear), distress to
limitations, duration of orienting, soothability, and smiling
and laughter (Rothbart, 1981). The instrument required
frequency judgements on approximately 90 behavioral items
which might have occurred over the previous one to two weeks.
Stability coefficients over the three to 12 month age bracket
ranged from 0.06 to 0.80, with greater stability reported
above six months (Rothbart, 1981 ).
Dimensions of Temperament Survey. Maternal temperament
was measured with the Dimensions of Temperament Survey
(Lerner, Palermo, Spiro, & Nesselroade, 1982). This
instrument consisted of 34 behavioral items which subjects
marked as either true (accurate self-descriptions) or false
(inaccurate). The five temperamental factors identified by
this instrument were activity level, attention
span/distractibility, adaptability, rhythmicity, and
reactivity. Three day test-retest stability ranged from 0.60
to 0.93 (Lerner et al., 1982).
The Strange Situation. As described in the introduction,
the strange situation consisted of eight short episodes, in
which the infant's response to separation, reunion, and

34
overtures by a stranger were noted. The episodes were
designed to gradually increase the stress to which the infant
was exposed, with concomitant increases in attachment
behaviors expected. Strange situation assessments occurred in
two different experimental rooms, one located in Shands
Teaching Hospital and the other in the Department of Clinical
Psychology research trailer. These facilities were located
several miles apart and in very different settings, chosen to
minimize the children's recognition of the assessment
situation. The trailer was located on a quiet drive in a
woodsy setting, isolated from the rest of the university. In
contrast, the room at Shands Teaching Hospital was located
within a large medical center at a busy intersection of the
campus.
Each of the experimental rooms contained two chairs and a
number of toys appropriate for one-year-old infants. Toys at
the trailer included toy phone, foil pie plates, plastic
hammer, milk bottle with shapes, baby doll, musical apple,
felt chicken, plastic shapes, plastic toy rabbit, and toy keys
on a ring. Toys at the hospital consisted of large colored
ball, plastic bucket and shovel, colored plastic blocks, toy
camera music box, chicken rattle, cloth pig, plastic pull
train, cloth doll with rattle, bird squeeze toy, fish rattle,
and soft rubber elephant. The experimental room at the
hospital measured 10 by 12 feet, was carpeted, and contained a
couch and two small end tables in addition to the equipment
described above. The room at the trailer was smaller,
measuring 8 by 6, uncarpeted, and furnished only with a file

35
cabinet.
Procedure
Prior to the first assessment, the subjects were
interviewed by the principal investigator, either in their
homes or at the psychology clinic at Shands Teaching Hospital.
The procedures of the study were explained and informed
consent was obtained (Appendix A). Next, the structured
interview was conducted and mothers were instructed in the
procedures of the four inventories. Finally, the mothers' role
in the strange situation was outlined and each was given a
descriptive handout.
All subjects were tested in two sessions of the Ainsworth
Strange Situation, scheduled at two week intervals. Although
scheduling all children for assessments at the same time of
day proved impossible, an attempt was made to maintain
consistency in time of assessment between the two sessions of
each subject. For half of the subjects, the testing sessions
occurred in two different experimental rooms (at the research
trailer and Shands Teaching Hospital). The remaining subjects
were tested twice in the Clinical Psychology research trailer.
For subjects tested in two different settings, the order of
the experimental rooms was counterbalanced. When the subjects
arrived at the prescribed setting, the procedures of the study
were reviewed with them by the principal investigator or a
research assistant, and the mothers were given a cue sheet to
take into the room with them (Appendix C). Following each
session of the strange situation, mothers were asked (1) if

36
they were surprised by any aspect of their child's behavior,
and (2) whether anything unusual had occurred on the day of
testing which might have influenced the assessment.
Three different female strangers were utilized; the order
of their appearance was varied randomly. The strangers were
recruited from an undergraduate psychology class and they
received course credit for their participation in the study.
They engaged in four training sessions prior to beginning the
experiment. Although these students were familiar with
attachment theory in general, they were unaware of the
specific hypotheses of the study and were blind to the
behavior of each baby in any prior session. (Both the
experimental settings and stranger behavior were reviewed and
approved by Dr. Ross Thompson from the University of Nebraska
following the initial training sessions.)
The infants' behavior in the strange situation was
videotaped from behind one-way mirrors. Measures which were
scored from these records included classification into
groups and subgroups and ratings of proximity and contact¬
seeking, contact-maintaining, avoidance, and resistance during
the two reunion episodes. These four interactive behaviors
have been identified previously as most salient and reliable
in differentiating patterns of attachment (Waters, personal
communication, April, 1985). Descriptive ratings were
provided by Ainsworth et al. (1978), with seven points on each
scale anchored to specific patterns of response.
Two raters (RAI and CRI) were trained by Dr. Everett
Waters in the scoring of the strange situation. Rater CRI was

37
blind to the experimental hypotheses of the study. Training
was undertaken using videotapes from an unrelated study. The
two raters each scored half of the session one and half of the
session two tapes, and were blind to the behavior and
classification of the infants in their other session. The
raters then scored the other half of the videotapes. Tapes
were rerated in groups of four if there was any disagreement
on group or subgroup classification, or if ratings of
interactive behaviors differed by more than two points.
Independent reliability was achieved with Dr. Mary Jo Ward who
scored a random sample of the tapes (N = 15). She was blind
to the hypotheses of the study, as well as to the behavior of
any infant in a previous session. Dr. Ward had been trained
by Dr. Alan Sroufe (Institute of Child Development,
Minneapolis, Minnesota) and had demonstrated consistent
reliability with his ratings.

RESULTS
Reliability
Two independent raters each scored 82% of the videotapes.
Ninety-eight percent agreement was achieved for ABC
classification (Kappa = 0.93). Eighty percent agreement was
reached on subgroup classification. Intraclass correlation
coefficients (Bartko, 1966) were calculated for the
interactive ratings. The values for proximity-seeking,
contact-maintaining, resistance, and avoidance were 0.93,
0.97, 0.86, and 0.90, respectively.
Reliability was also calculated between the final
consensus values used in the analyses and the ratings on a
random sample scored by Dr. Ward. Mean values for these
ratings were 0.80, 0.91, 0.92, and 0.60, respectively.
Stability
The group of infants tested twice in the same setting were
seen both times in the trailer, with no infants seen twice in
the hospital. Therefore, initial analyses were conducted to
ascertain that the effect of a particular setting did not
cloud the analyses of stability. First, a comparison of the
trailer/trailer group (TT, N = 15) and the trailer/hospital
group (TH, N = 8) was made for the distribution of the three
classifications at test and retest. Using chi-square tests,
there were no significant differences. Because the expected

39
cell frequencies were lower than optimal for the chi-square
statistic, comparisons were also made with Fisher's exact
test. For this comparison, the distribution of A and C
infants considered as one group was compared with group B,
with no significant differences found between the treatment
groups. The same two groups (TT/TH) were examined for
differences in number of infants who retained the same
classification over the two sessions. Again, a chi-square
test revealed no significant differences.
As a further check for setting differences, all infants
seen in the trailer at session one (n = 23) were compared to
those seen in the hospital (n = 6). There were no differences
between groups using the Fisher's exact test to examine
distribution of infants into A plus C versus B categories. T-
tests also demonstrated no differences in interactive ratings
during session one between these groups.
Next, the group of infants tested in the order
trailer/hospital (TH) was compared with that tested first in
the hospital, followed by the trailer (HT, N = 6). Chi-square
analyses revealed no differences between these groups in the
distribution of classifications either at the first or second
session. In addition, Fisher's exact tests demonstrated no
differences at test or retest in distribution of subjects into
A plus C versus B categories. Likewise, there were no
differences between these groups in the number of infants who
retained their classification across sessions.
In addition to examining the effects of setting on overall
group membership, the four interactive variables were

40
scrutinized. With one exception, no differences were noted on
t-tests comparing groups TT and TH in terms of proximity¬
seeking, contact-maintaining, resistance, or avoidance in
either reunion episode, during both test and retest.
Avoidance in episode 5 of the retest was significantly higher
for infants in the TT group than the TH group (t = 2.131; p <
0.0489). Given the relatively large number of statistical
tests employed, as well as the lack of support for this
finding in other measures (and theoretically), it seems likely
that this is not a meaningful finding. Difference scores on
each interactive variable calculated across sessions (e.g.,
DavoidS = Avoidance Episode 5, test - Avoidance Episode 5,
retest) were also examined using t-tests. There were no
significant differences.
As with the overall group classification, comparisons
between the TH and HT groups were next made. There were no
differences between these two groups on measures of any of the
four interactive variables in each reunion in both test and
retest sessions. Several significant findings were noted in
difference scores. Babies in the TH group tended to increase
proximity-seeking in episodes 5 and 8 over the two sessions,
as compared with decreases in the HT group (t = -2.8820, p <
0.0138, and t = -2.3039, p < 0.0399, respectively). In
addition, avoidance in episode 8 tended to decrease across
sessions in the TH group, as compared with increases in the HT
group (t = 2.3569, p < 0.0363). All other difference score
comparisons were nonsignificant.

41
In light of the general lack of support for setting
effects, analyses were conducted to compare trailer/trailer
subjects with all changed context subjects (TH and HT). The
distribution of infants for the two groups at test and retest
is described in Table 1. First, the likelihood that a
particular classification was more frequent in one of the
groups at either test or retest was examined. Chi-square
tests were nonsignificant. Fisher's exact tests applied to
the A plus C versus B distribution at both sessions were also
nonsignificant. Similarly, chi-square tests demonstrated no
tendency for retaining classification to be more likely in one
group as compared with the other. As in the previous section,
t-tests were used to compare individual interactive ratings in
both episodes in each session. There were no differences. In
addition, no differences were found on comparisons of any of
the interactive rating difference scores.
As a final test of differences in stability between the
two groups, kappa was computed for each. The values were
-0.1409 and -0.1572, for TT and TH/HT, respectively. As
suggested by Cohen (1960), a two way comparison was made by
evaluating the normal curve deviate, with no significant
differences emerging (z = 0.03, p > 0.05).
As no significant differences in stability between TT and
TH/HT groups were found, all subjects were combined for the
following analyses. First, of the total sample in the first
session, 25 infants (83%) were classified as securely
attached, while three infants were rated as avoidant (10%),

42
TABLE 1
Classification of Infants at Test and Retest
in Same Context/Changed Context Groups
Test A
1
Trailer/Trailer Group
Retest
B
1
10
12
(0.07) (0.80)
2
(0.13)
1
(0.07)
13
(0.86)
1
(0.07)
15
Hospital/Trailer and Trailer/Hospital Group
Retest
B C
Test A
A
0 1
2 10
0 1
2 12
(0.14) (0.86)
0
(0.0)
1
(0.07)
12
(0.86)
1
(0.07)
1 4

43
and two were rated resistant (7%). Although this distribution
is weighted more heavily with secure infants than often found
in the literature, it is comparable to some studies using
middle class samples (e.g., Owen et al., 1984).
The overall stability was 69% (kappa = -0.116). Table 2
describes the distribution of infants at test and retest in
the current study, as compared with Ainsworth et al.'s (1978)
sample. There was no clear trend in direction of change.
Five children changed from secure to anxious attachment (three
B/A and two B/C) and four switched from anxious to secure (two
A/B and two C/B). Twenty-eight percent of the infants
retained the same subgroup classification from the first to
second session. (See Table 3.) Intraclass correlation
coefficients for proximity-seeking, contact-maintaining,
resistance and avoidance were 0.41, 0.56, 0.12, and 0.14,
respectively, for episode 5, and 0.42, 0.57, 0.28, and 0.14,
for episode 8.
Attachment behaviors (proximity-seeking and contact-
maintaining) tended to increase across sessions in episode
five (Dprox5 = -0.7241; Dcont5 = -0.8448) and decrease in
episode eight (Dprox8 = 0.7241; Dcont8 = 0.4828). Avoidance
tended to decrease across sessions in episode five (Davoid5 =
0.5517) and increase in episode eight (Davoid8 = -0.4138).
Means and standard deviations for the interactive ratings are
presented in Table 4. As illustrated, in session one,
proximity-seeking, contact-maintaining, and resistance
increased significantly from episode five to eight, while

44
TABLE 2
Classification of Infants at Test and
for the Current Study and Ainsworth's
Current Study
Retest
Test A
20
3
(0.10)
24 2
(0.83) (0.07)
Test A
B
C
Ainsworth Sample
Retest
ABC
0 7 0
0 12 2
0 1 1
0 20 3
(0.00) (0.87) (0.13)
Retest
Sample
2
(0.07)
25
(0.86)
2
(0.07)
29
7
(0.30)
1 4
(0.61 )
2
(0.09)
23

45
Test
TABLE 3
Subgroup Classification at Test and Retest
A1
A2
B1
Retest
B2 B3
B4
ci
C2
A1_
0
0
i
1
0
0
0
0
2
(0.07)
A2
0
0
0
0
0
0
0
0
0
(0.0)
B1
0
0
1
3
0
0
0
0
4
(0.14)
B2
2
0
2
2
1
0
0
1
8
(0.28 )
B3
1
0
1
0
5
5
1
0
1 3
(0.45)
B4
0
0
0
0
0
0
0
0
0
(0.0)
Cl
0
0
0
0
0
1
0
0
1
(0.03)
C2
0
0
0
0
1
0
0
0
1
(0.03)
3 0 5 6 7 6 1 1
(.10) (.17)(.21 )(.24)(.21 ) (.03)1.03)
29

46
avoidance decreased. In contrast, the only significant change
across episodes in the retest was an increase in contact-
maintaining from five to eight. Combined means for episodes
five and eight for each session, as well as correlations
between these means, can be found in Table 5. Differences
between the means of the two sessions were nonsignificant for
all interactive ratings.
Further paired comparison t-tests were computed for
subjects sorted into two groups according to whether they
changed (n = 9) or retained (n = 20) their classifications
across sessions. Infants who changed classifications showed
increases from episode five to eight (in session one) in
proximity-seeking, contact-maintaining and resistance (p <
0.0327, 0.0112, and 0.0099, respectively). During the retest,
only a decrease in proximity-seeking from episode five to
eight was significant (p < 0.0404). For those infants who
retained their classification, proximity-seeking and contact-
maintaining increased across the episodes in the first session
(p < 0.0024 and p < 0.0009), while avoidance decreased (p <
0.0116). During the second session, the only significant
difference across episodes was an increase in contact-
maintaining (p < 0.0218).
Given the pattern of instability across sessions, the
association between short-term stability of infant-mother
attachment and family environmental and demographic variables
was assessed. For these analyses, subjects were sorted into
those who retained their classification over the two sessions

47
(N =20) versus those infants who changed groups from test to
retest (N = 9). Using t-tests, no significant differences
were found between the groups on the following variables:
social support, life stress, maternal temperament, infant
temperament, caretaking changes, parental age and hours
worked, and separations from mother over the first year of
life. In contrast, infants who retained their classification
were found to come from lower income families (M = $17704)
than those who changed groups (M = $27700; t = 2.1503, p <
0.0403). This difference is contradictory to findings in the
literature suggesting a higher rate of stability in higher
income families, and is likely to be a spurious finding
resulting from the large number of statistical tests employed.
A series of chi-square tests evaluated the association
between stability of classification and categorical variables.
No differences were found based on sex of subject, birth
order, level of parental education, maternal work status, or
changes in caretakers over the first year.
To summarize the previous section, actual setting in which
the strange situation occurred had no effect on classification
in a particular setting or on stability of classification. In
addition, there were no differences in the stability of
classification in the same context group (TT) as compared with
the changed context group. Stability for the sample
considered as a whole was unimpressive (k = -0.166). Finally,
when infants who retained their group status were compared
with those who changed across sessions, no significant

48
TABLE 4
Means and
Standard
Deviations for
Interactive
Ratings
Test
Episode 5
Episode 8
Paired
Comparisons
Mean
SD
Mean
SD
2
Proximity-
Seeking
2.93
1 .98
4.45
1 .49
0.0002
Contact-
Maintaining
2.57
1 .90
4.38
2.25
0.0001
Resistance
1 .43
0.93
1 .90
1 .46
0.0377
Avoidance
2.50
1 .71
1 .57
1 .30
0.0017
Retest
Proximity-
Seekinq
3.66
1 .93
3.72
1.92
0.7636
Contact-
Maintaining
3.43
2.40
3.95
2.32
0.0122
Resistance
1 .38
0.92
1 .59
1.17
0.3392
Avoidance
1 .90
1 .45
1 .97
1 .38
0.7869

TABLE 5
Comparison of Means of Strange Situation Measures
for Sessions One and Two and Correlations Between Sessions
Measure
Session 1
Session 2
Significance
r
Mean
Mean
of Difference
Proximity-
Seeking
3.69
3.69
n.s.
0.54
(0.0024)
Contact-
Maintaining
3.48
3.69
n.s.
0.65
(0.0001)
Resistance
1 . 67
1 . 48
n.s.
0.23
(0.2263)
Avoidance
2.03
1 . 93
n.s.
0.39
(0.0354)

50
differences emerged on demographic or family environmental
variables. One exception to this was family income. Due to
the small sample size, the group of infants which changed
classification could not be sorted according to direction of
change (e.g., secure to anxious versus anxious to secure).
Family Environmental Variables
The second aim of this study was to investigate the
relationship of a variety of family environmental variables
to quality of attachment. For these analyses, only the
classifications from the first session were utilized. The
group of securely attached infants (N = 25) was compared with
those children who were anxiously attached (N = 5), which
included both avoidant (n = 3) and resistant (n = 2) infants.
T-tests were used to compare these groups on measures of
social support, life stress, maternal temperament, infant
temperament, hours that mother and father spent caring for the
baby, caretaking changes, and income. There were no
significant differences except for the maternal temperament
factor of activity. On this variable, mothers of securely
attached infants tended to score lower than did mothers of
anxiously attached infants (t = -2.7282; p < 0.0109). A chi-
square test evaluating the relationship between attachment and
maternal work status also proved to be nonsignificant.

DISCUSSION
The present study investigated the hypothesis that the
stability of attachment classification would be increased if
an infant's recognition of the assessment context was
eliminated. In order to evaluate this assumption, half of the
subjects were tested in two very different settings, presented
in counterbalanced order, with the remainder seen twice in the
same context. Results demonstrated that context effects had
no bearing on classification stability. The coefficients of
agreement (kappa) for the two groups (same context and changed
context) were similarly unimpressive, and nonsignificantly
different. This finding was supported by the lack of
differences between the groups in interactive ratings and
changes in those ratings across time. Although the relatively
small sample size and the large proportion of securely
attached infants found in the study preclude taking a
definitive stand on the meaning of these findings, it should
be noted that none of the statistical comparisons were even
close to significantly different. The lack of differences
between these groups suggests that if memory for the strange
situation is affecting classification stability, it is likely
to be recognition of the entire sequence of events rather than
of the physical environment alone that affects the children's
behavior in the second assessment.

52
As the recognition of context hypothesis received no
support, subjects were collapsed into one group in order to
examine stability in the sample as a whole. In this study,
there was a higher percentage of infants who retained their
classification across sessions as compared with the Ainsworth
et al. (1978) sample (69% and 57%, respectively). However, a
comparison of the kappa values computed for these two samples
revealed that when the amount of agreement expected by chance
was considered, stability in the current study was even lower
than in Ainsworth's results.
Temporal instability would not necessarily undermine the
validity of the classification system, if changes in the
quality of attachment were either the predictable result of
anxiety engendered in the first session spilling over into the
second, or if the changes could be related predictably to
environmental changes. Ainsworth and her colleagues were able
to develop reasonable explanations for their findings, based
on the pattern of changes they observed. First, the bulk of
changes in classification were accounted for by avoidant
infants appearing secure in the second session. In addition,
proximity-seeking and contact-maintaining were significantly
higher in session 2, while avoidance decreased from time one
to time two. The authors stated: "it is clear that the
effect of the repetition of the strange situation after a
brief two-week interval was to increase distress and the
intensity of attachment behavior, while at the same time the
negative behaviors of avoidance and resistance decreased or

53
remained at about the same level of intensity" (Ainsworth et
al., 1978, p.222).
The present study does not support these conclusions. Of
the nine infants who changed classification, four moved from
anxiously to securely attached, and the remainder changed from
secure to anxious. Also in contrast to the Ainsworth study,
changes in interactive behaviors across sessions were not
consistently in the expected directions. In general, the
explanation posed by other researchers to explain short-term
instability lacks validity in relation to these results: the
hypothesis that the anxiety engendered by encountering a
second "strange situation" results in increased attachment
behaviors and a general progression to secure attachment is
not supported.
As Ainsworth and her colleagues suggested, it seems
unlikely that the quality of the mother-infant attachment
would change dramatically over a two week period, unless there
were gross changes in the factors which influence the mother-
infant interaction. Post-experimental questions administered
after each assessment in the current study provided anecdotal
evidence to the effect that infants who changed
classifications had not experienced any noteworthy disruptions
or trauma in the two week interim, relative to those infants
who remained in their original group. Mothers occasionally
reported illness in the child or an unrestful nap, but these
comments were equally distributed across groups and time of
testing.

54
In a further attempt to understand the pattern of
instability which emerged in the present study, family
environmental variables which have been implicated in long¬
term stability were examined. It was not expected that gross
changes in the factors measured would have occurred between
sessions, and in fact, the questionnaires were completed prior
to both assessments. However, based on six to eight month
test-retest studies, it seemed that certain variables such as
infant difficulty, life stress, or caregiving continuity might
affect short-term stability. Clearly, there was no evidence
to support these hypothesized relationships.
To summarize, the current study did not support the
contention that short-term instability in the ABC
classification system was related to stress-inducing effects
of returning twice to the same assessment context.
Furthermore, instability in this sample as a whole was greater
than that found in the Ainsworth sample. As Lamb (1978) had
reported with a mother/father test-retest study, there was no
predictable pattern in the direction of changes which
occurred. Infants were as likely to change from secure to
anxious as they were from anxious to secure. Finally, none of
the personality or environmental variables which were
investigated differentiated between infants who changed versus
retained classification. As mentioned previously, due to the
small number of subjects, the changed classification group
could not be analyzed according to direction of change.

55
The conclusion which emerges from these results is that
classification of attachment quality over a two week period
lacks stability. In contrast to Ainsworth's findings, but in
line with Lamb's, this instability appeared random in
direction, rather than following a trend to normative or
secure attachment.
This study does have limitations which constrain
overgeneralization of results. The inclusion of a treatment
group comprised of infants seen twice in the hospital setting
would have eliminated the need for analyses of specific
setting effects. In addition, the relatively small sample
size and the securely weighted distribution of infants
presented difficulties for statistical analysis.
Overriding these concerns was the fact that none of the
statistical comparisons were even marginally significant.
Converging evidence from both classification status and
interactive ratings strongly supported the finding of lack of
stability. Further research to document these results should
include a larger sample, both to increase the power of
statistical evaluation and avoid the problems raised by the
ABC sample distribution. There is little reason to believe,
however, that the presence of more anxiously attached infants
in the first session would improve stability estimates.
Both the Ainsworth data and results of long-term studies
(e.g., Thompson et al., 1983) suggest that anxiously attached
children are less likely to retain their classification in a
retest than are secure infants. Finally, future research

56
might also investigate stability at an interval longer than
two weeks. As a cautionary note, however, as the time between
test and retest increases, the risk of- confounds from
developmental changes and alterations in environmental
circumstances increases as well.
Unfortunately, very little can be concluded regarding the
second aim of this study. In this case, the large percentage
of securely attached infants weighed against the probability
of identifying differences between groups on family
environmental variables.
Keeping in mind that the sample size and group
distribution limit firm conclusions, what might these results
mean for the ABC classification system? One speculation would
be that the ABC taxonomy is not the most parsimonious and
accurate model for describing differences in quality of
attachment. Conceivably., differences within each group are
larger than stylistic differences between groups. More
specifically, infants in different subgroups may have very
different styles of responding to increases in anxiety, which
influence patterns of temporal stability. Although Ainsworth
et al. did not report subgroup membership, their most notable
finding was the movement of seven infants from avoidantly to
securely attached. In the present study, only two infants
were classified as avoidant in the first session, and both
moved to a secure classification (B1 or B2) at time two. In
terms of the B subgroups, several tentative directional trends
can be identified. B3 infants tended either to remain in that
group or move to B4.
B2 infants tended to remain stable or

57
change to a B1 or A1 classification. B1 infants tended to
look like B2's in their second session.
Lamb, Thompson, Gardner, Charnov, and Estes (1984) have
suggested that cluster analysis does not provide strong
support for the ABC classification system. In applying
cluster analysis to five samples which varied in age tested
and cultural background, Gardner, Lamb, and Thompson
(1985) found that boundaries between groups generally
did not fall along the A/B and B/C distinctions. They
reported the C group to appear extremely heterogeneous and
that differences between some B subgroups were often greater
than differences between A and B subgroups. More
specifically, "distal" B's (B1 and B2) frequently clustered
with A infants, while "proximal" B's (B3 and B4) were often
grouped with C's. In addition, the fit of the ABC
classification and the empirically derived clusters were found
to vary according to the age and cultural background of the
sample, with the closest fit shown in an American, one-year-
old population. The authors conclude by recommending the
consideration of continuous measures comprised of (1) a
composite summing proximity-seeking and contact-maintaining
and subtracting avoidance, and (2) the resistance scale.
The general trend of the movement between subgroups in the
present study is largely consistent with Gardner et al.'s
findings using cluster analysis. In other words, a change in
classification from B1 or B2 to group A (or vice versa) may
actually be reflective of behavioral similarity and continuity

58
among those groups, rather than suggesting temporal
instability of individual differences, as the current
classification system would dictate.
An examination of the overall pattern of interactive
rating change in the present study demonstrated that there
were significant changes in all four ratings from episode five
to eight in the first session. Two weeks later, only contact-
maintaining showed a significant increase. By sorting the
infants according to stability of classification, an attempt
was made to evaluate whether certain infants had habituated to
the stress of the strange situation, while others had not.
The two groups showed reasonably similar patterns of change in
proximity-seeking and contact-maintaining in session one,
though differing on the other two variables. Differences in
the pattern of change across episodes continued in the retest.
In a larger sample which would allow investigation of
direction of classification change, analyses along these lines
might clarify which types of infants show predictable
behavioral reactions to the stress of separation.
The final section of this paper focuses on practical and
sociological observations of the ABC classification system.
As reviewed in the introduction, the strange situation and its
ensuing ABC taxonomy are plagued with a number of
problems. Questions of both short-term and long-term
stability remain unanswered. Results of the present study
illustrate the questions surrounding short-term temporal
stability. Efforts which have been directed to understanding
six to ten month stability have produced inconsistent results.

59
First, there has been wide variability reported across
studies. Second, the differences found by various researchers
are not clearly explicable on the basis of sample
characteristics (e.g., note the wide discrepancies in
estimates of stability found in similar middle-class samples).
Finally, attempts to link stability/instability to external
variables have not been definitive: while Vaughn et al.
(1980) found changes from secure to anxious attachment to be
associated with life stress, Egeland and Farber (1984) were
unable to replicate this in a larger sample. Likewise,
Thompson et al. (1982) reported bidirectional changes in
attachment status to be related to changes in mothers'
employment status and caregiving disruption, while Owen et al.
(1984) found increased stability in infants whose mothers
returned to work during the second year.
In spite of problems such as those cited above, the
strange situation is a highly popular and visible metholology.
A brief review of the abstracts from the Society for Research
in Child Development catalog (biennial meeting, 1985)
demonstrates that of the 32 articles discussing attachment, 31
or 97% utilized this methodology. The only author who did not
use the ABC taxonomy was Everett Waters, who is in the process
of developing a Q-sort instrument for assessing attachment.
A cursory look at the major figures producing attachment
research reveals that they fall roughly into those associated
with Ainsworth through training or collaboration (e.g.,
Sroufe, Waters, Vaughn, Main, Egeland) and those not (e.g.,

60
Lamb, Thompson, Belsky, Crockenberg). The authors who have
been most critical of the methodology while continuing to be
extremely involved in the area are Michael Lamb and his
students. In general, critiques of the taxonomy tend to be
ignored or to generate considerable heat (e.g., Thompson et
al., 1982; Waters, 1983; Thompson et al., 1983). It seems
noteworthy that the major review paper produced by Lamb et al.
(1984) which has a format specifically designed for peer
commentary, stimulated no response from any of the Ainsworth
related group.
Frequently, the focus of the controversy over this
methodology comes to rest on the scoring system for the
strange situation videotapes. In several discussion sections,
results which contradict the ABC model were deemphasized or
dismissed on the basis of inadequate scoring (e.g., Waters,
1983; Ainsworth et al., 1978). Practically speaking, learning
the scoring system is a cumbersome, time-consuming, and
expensive task for researchers who are not collaborating with
a major figure. Clearly, working only from the written
material provided in the Ainsworth et al. book is inadequate
for establishing reliability. Waters (1978) made reference to
a set of standardized videotapes which would be available for
training and reliability. In fact, these were never
developed, presumably because of disagreements among the
prospective authors (Waters, personal communication, 1985).
Given how firmly entrenched the ABC system has become in the
attachment literature, the standardization of the methodology
is critical. Resolution of the controversy surrounding the

61
taxonomy rests on the provision of standardized materials
which would allow the assumption of uniform ratings across
research laboratories and theoretical camps.

APPENDIX A
INFORMED CONSENT
Informed consent to Participate in Research
J. Hillis Miller Health Center
University of Florida
Gainesville, Florida 32610
You are being asked to volunteer as a participant in a
research study. This form is designed to provide you with
information about this study and to answer any of your
questions.
TITLE: Mother-infant attachment: Stability and relationship
to family environmental variables
PROJECT DIRECTOR: Roberta A. Isleib, M.S. 392-2944
PURPOSE: The purpose of the present study is to understand
the effects of variables such as stressful life events, social
support, mother and child temperament, and caretaking
arrangements on infant social behavior and the mother-child
relationship. The study will also investigate the stability
of the classification system which is currently used to
describe infant behavior.
PROCEDURES: Each participant will be briefly visited in their
home, where the study will be explained and questionnaires
concerning life stress, social support networks, temperament,
and caretaking arrangements will be completed. In addition,
each mother-infant pair will visit the laboratory two times,
at two weeks apart, in order to participate in a 20 minute
session in which the baby's reaction to the mother, a friendly

63
stranger, and being briefly left alone will be filmed.
POTENTIAL RISKS OR DISCOMFORTS: The only potential risk to
subjects for participation in this study would be the
relatively mild distress experienced by some infants during
separation from their mothers for a three minute period.
Mothers will be allowed to watch their infants from behind a
mirror during this time and may terminate the study at any
point, if desired.
POTENTIAL BENEFITS TO YOU OR TO OTHERS: There are no clear
benefits to be gained by experimental subjects. The primary
benefits of the research involve increasing our understanding
of the factors involved in infant social behavior, and
indirectly, to improve the quality of preventative mental
health care in this important area.
GENERAL CONDITIONS: I understand that I will receive ten
dollars for my participation in this study. I understand that
I will not be charged additional expenses for my participation
in this study. I understand that I am free to withdraw my/my
child's consent and discontinue participation in this research
project at any time without this decision affecting my child's
medical care. In the event of my/my child's sustaining a
physical injury _ which is proximately caused by this
experiment, no professional medical care received at the J.
Hillis Miller Health Center exclusive of hospital expenses
will be provided me without charge. It is understood that no
form of compensation exists other than those described above.
All data obtained from this research will remain
confidential. The University of Florida will protect the

64
confidentiality of this document and your records from this
research to the extent provided by law.
I have fully explained to the
nature and purpose of the above-described procedure and the
benefits and risks that are involved in its performance. I
have answered and will answer all questions to the best of
my ability. I may be contacted at telephone # .
Signature of person obtaining consent Date
I have been fully informed of the above-described procedure
with its possible benefits and risks and I have received a
copy of this description. I have given permission for my/my
child's participation in this study.
Signature of relative, parent, or guardian Date
Witness to signature
Date

APPENDIX B
STRUCTURED INTERVIEW
Baby's birthdate Sex M F
Mother's age Father's age
Ages of other children living in the home
Names and relationships of others living in the home
Level of education: Mother Father
Occupation: Mother Father
Describe any changes in the father's employment status since
the baby's birth (such as promotion, job change, laid off)
Mother returned to work since the birth? y n
If so, on what date? Occupation?
How many hours weekly? Still working?
If not, stopped working when?
Who regularly (7 hours or more daily, 4 or more days per week)
cares for the baby when the baby is awake? mother
father relative babysitter or neighbor
day care center two share equally
On the average, how many hours weekly does the mother care for
the baby when the baby is awake? Father?
On the average, how much time weekly is the baby with a
caregiver who is not the parent?
Have you changed babysitters/day care centers in the past year
for any reason? y n # times Reason

66
Has either parent missed work in the past year due to
caretaking problems? y n # days mother # days
father Reasons
Have you cancelled any social engagements because of
caretaking problems? y n # times Reason
Has the child ever been separated from the mother for a period
of 24 hours or longer (such as for a vacation,
hospitalization, etc.)? y n If so, date
Reason Length of separation
Caretaker during this time
Does your child regularly have contact with other infants or
young children? y n Circumstances
Total family income

APPENDIX C
INSTRUCTIONS TO PARENTS
Duration
of Episode
30 seconds
3 minutes
3 minutes
3 minutes
(or less)
3 minutes
3 minutes
(or less)
Description of Episode
1. MOTHER, BABY, AND EXPERIMENTER
Experimenter will accompany you and your
child to the playroom, and show you your
seat.
2. MOTHER AND BABY
Try to interest your baby in toy play,
then return to your chair. Respond
naturally and appropriately to your
child's initiatives, but try not to
initiate interaction on your own.
3. MOTHER, BABY, AND STRANGER
First minute: stranger sits quietly
Second minute: stranger chats with
mother
Third minute: stranger begins to play
with baby
4. BABY AND STRANGER
Knock on window signals your departure.
Walk quietly to the observation room
door and leave the room, closing the
door behind you.
5. MOTHER AND BABY
Talk loudly before opening the door so
baby knows you're coming. Pause in
doorway to allow baby to greet you.
Comfort and soothe your child, if
necessary. When baby is ready, try to
reinterest him/her in toy play, and
return to your chair.
6. BABY ALONE
Knock on window signals your departure.
Walk quietly to the observation room
door, say "good-bye" to your baby, and
leave the room, closing the door.

68
3 minutes
(or less)
3 minutes
7. BABY AND STRANGER
Stranger will try to comfort baby, if
needed.
8. MOTHER AND BABY
Talk loudly before opening the door so
baby knows you're coming. Pause in
doorway to allow baby to greet you.
Comfort and soothe your child, if
needed. When baby is ready, try to
reinterest him/her in toy play, and
return to your chair.
Remain in the room; we will return
shortly to end the session.

REFERENCES
Ainsworth, M.D. (1964). Patterns of attachment behavior shown
by the infant in interaction with his mother. Merrill-Palmer
Quarterly, 10, 51-58.
Ainsworth, M.D. (1973). The development of infant-mother
attachment. In Betty Caldwell & H. Ricciuti (eds.), Review of
child development research (Vol. 3). Chicago: University of
Chicago Press, 1-94.
Ainsworth, M.D. (1979a). Attachment as related to mother-
infant interaction. In J.S. Rosenblatt, R.A. Hinde, C.Beer, &
M.C. Busnel (Eds.), Advances in the study of behavior, vol. 9.
New York: Academic Press, 1-51.
Ainsworth, M.D. (1979b).
Psychologist, 34, 932-937.
Infant-mother attachment. American
Ainsworth, M.D., Bell, S.M., & Stayton, D.J. (1971).
Individual differences in strange situation behavior of one-
year-olds. In H.R. Schaffer (Ed.), The origins of human
social relations. London: Academic Press, 17-57.
Ainsworth, M.D., Bell, S.M., & Stayton, D.J. (1972).
Individual differences in the development of some attachment
behaviors. Merri11-Palmer Quarterly, 1 8 , 1 23-1 43.
Ainsworth, M.D., Blehar, M.C., Waters, E., & Wall, S. (1978).
Patterns of attachment: A psychological study of the strange
situation. Hillsdale, N.J.: Lawrence Erlbaum Associates,
Inc.
Ainsworth, M.D., & Wittig, B. A. ( 1969 ). Attachment and
exploratory behavior of one-year-olds in a strange situation.
In B.M. Foss (Ed.), Determinants of infant behaviour IV.
London: Methuen, 111-153.
Arend, R., Gove, F.L., & Sroufe, L.A. (1979). Continuity of
individual adaptation from infancy to kindergarten: A
predictive study of ego-resiliency and curiosity in
preschoolers. Child Development, 50, 950-959.
Bartko, J. (1966). The intraclass correlation coefficient as
a measure of reliability. Psychological Reports, 19, 3-11.

70
Belsky, J., Garduque, L., & Hrncir, E. (1984). Assessing
performance, competence, and executive capacity in infant
play: Relations to home environment and attachment.
Developmental Psychology. 20, 406-417.
Belsky, J., Rovine, M., & Taylor, D. (1984). The Pennsylvania
infant and family development project, III: The origins of
individual differences in infant-mother attachment: Maternal
and child contributions. Child Development, 55, 718-728.
Belsky, J., & Steinberg, L.D. (1978). The effects of day
care: A critical review. Child Development, 49, 929-949.
Bowlby, J. (1958). The nature of a child's tie to his mother.
International Journal of Psychoanalysis, 39, 350-373.
Bowlby, J. ( 1 969 ). Attachment and loss, Volume 1:
Attachment. New York: Basic Books.
Bowlby, J. ( 1973 ). Attachment and loss. Volume 2:
Separation. New York: Basic Books.
Bowlby, J. (1977). The making and breaking of affectional
bonds. I. Aetiology and psychopathology in the light of
attachment theory. British Journal of Psychiatry, 130, 201-
210.
Cairns, R.B. (1966). Attachment behavior of mammals.
Psychological Review, 73, 409-426.
Cassidy, J., & Ainsworth, M.D. (1983). The toddler's ability
to negotiate the environment: Its relation to quality of play
and quality of attachment. Presented at the Society for
Research in Child Development, Detroit.
Chase-Lansdale, P.L. (1981). Maternal employment in a family
context: Effects on infant-mother and infant-father
attachments. Dissertation Abstracts International, 42, 2562-A.
Cohen, J. (1960). A coefficient of agreement for nominal
scales. Educational and Psychological Measurement, 1_, 37-46.
Connell, D.B. (1978). Individual differences in attachment:
Long-term stability and relationships to early language
development. Dissertation Abstracts International, 38,
3954-B.
Connell, J.P., & Goldsmith, H. (1982). A structural modeling
approach to the study of attachment and strange situation
behaviors. In R. Emde & R. Harmon (Eds.), The development of
attachment and affiliative systems. New York: Plenum Press,
213-243.

71
Crockenberg, Susan B. (1981). Infant irritability, mother
responsiveness, and social support influences on the security
of infant-mother attachment. Child Development, 52, 857-865.
Crockenberg, S. (1985). Predicting infant attachment from
early and current maternal behavior. Paper presented to the
Society for Research in Child Development, Toronto.
Easterbrooks, A., & Lamb, M. (1979). The relationship between
quality of infant-mother attachment and infant competence in
initial encounters with peers. Child Development, 50, 380-
387.
Egeland, B., & Farber, A. (1984). Infant-mother attachment:
Factors related to its development and changes over time.
Child Development, 55, 753-771.
Erickson, M., Sroufe, L., & Egeland, B. (in press). The
relationship between quality of attachment and behavior
problems in preschool in a high-risk sample. In I. Bretherton
& E. Waters (Eds.), Child development monographs. Chicago:
University of Chicago Press.
Gardner, W., Lamb, M. , & Thompson, R. (1985). On the
quantitative consistency of the strange situation
classification system. Unpublished manuscript.
Gewirtz, J.L. (1972). Attachment, dependence, and a
distinction in terms of stimulus control. In J.L. Gewirtz
(Ed.), Attachment and dependency. Washington, D.C.: Winston.
Green, J.H. (1981). The relationship of stress to infant-
mother attachment. Dissertation Abstracts International, 41,
3557-B. "
Hamilton, E. (1981). The relationship of maternal patterns of
stress, coping, and support to quality of early infant-mother
attachment. Dissertation Abstracts International, 42, 1172-A.
Hoffman, H.S., & DePaulo, P. (1977). Behavioral control by an
imprinting stimulus. American Scientist, 65, 58-66.
Jacobson, L., & Wille, D. (1984). Influence of attachment and
separation experience on separation distress at 18 months.
Developmental Psychology, 20, 477-484.
Jacobson, J.L., Wille, D.E., Tianen, R.L., & Aytch, D.M.
(1983). The influence of infant-mother attachment on toddler
sociability. Paper presented at the Society for Research in
Child Development, Detroit.
Lamb, M.E. (1978). Qualitative aspects of mother- and father-
infant attachments. Infant Behavior and Development, 1_, 265-
275.

72
Lamb, M., Thompson, R., Gardner, W., Charnov, E., & Estes, D.
(1984). Security of infantile attachment as assessed in the
"strange situation": Its study and biological interpretation.
The Behavioral and Brain Sciences. T_, 127-171.
Lerner, L.M., Palermo, M., Spiro III, A., & Nesselroade, J.R.
(1982). Assessing the dimensions of temperamental
individuality across the life span: The dimensions of
temperament survey (DOTS). Child Development, 53. 149-159.
Lieberman, A.F. (1977). Competence with a peer: Relations
with attachment and peer experience. Child Development, 48,
1277-1287.
Londerville, S., & Main, M. (1981). Security of attachment,
compliance, and maternal training methods in the second year
of life. Developmental Psychology, 2» 289-299.
Main, M., & Weston, D.R. (1981). The quality of the toddler's
relationship to mother and to father: Relationship to
conflict behavior and the readiness to establish new
relationships. Child Development, 52, 932-940.
Masters, J.C., & Wellman, H.M. (1974). The study of human
infant attachment: A procedural critique. Psychological
Bulletin, 81_, 218-237.
Matas, L., Arend, R., & Sroufe, L.A. (1978). Continuity of
adaptation in the second year: The relationship between
quality of attachment and later competence. Child
Development. 49, 547-556.
Owen, M., Easterbrooks, A., Chase-Lansdale, L., & Goldberg, W.
(1983). Infancy to toddlerhood: Effects of maternal
employment on infant-mother and infant-father attachments.
Paper presented to the Society for Research in Child
Development, Detroit.
Owen, M., Easterbrooks, M., Chase-Lansdale, L., & Goldberg, W.
(1984). The relationship between maternal employment status
and the stability of attachments to mother and to father.
Child Development, 55, 1894-1901.
Pastor, . D.L. (1981). The quality of mother-infant attachment
and its relationship to toddlers' initial sociability with
peers. Developmental Psychology, 17, 326-335.
Rajecki, D., Lamb, M., & Obmascher, P. (1978). Toward a
general theory of infantile attachment: A comparative review
of aspects of the social bond. Behavioral and Brain Sciences,
3, 417-464.
Robertson, J. (1952). Film: A two-year-old goes to the
hospitalâ–  London: Tavistock Child Development Research Unit.

73
Rothbart, M.K. (1981). Measurement of temperament in infancy.
Child Development. 52, 569-578.
Sarason, I., Johnson, J., & Siegel, J. (1978). Assessing the
impact of life changes: Development of the Life Experiences
Survey. Journal of Consulting and Clinical Psychology, 46,
932-946.
Sarason, I., Levine, H., Basham, R., & Sarason, B. (1983).
Assessing social support: The Social Support Questionnaire.
Journal of Personality and Social Psychology, 44, 127-139.
Schwartz, P. (1983). Length of day-care attendance and
attachment behavior in eighteen-month-old infants. Child
Development, 54, 1073-1078.
Sroufe, L.A., & Waters, E. (1977). Attachment as an
organizational construct. Child Development, 48, 1184-1199.
Thompson, R., & Lamb, M. (1983). Security of attachment and
stranger sociability in infancy. Developmental Psychology,
J_9, 184-191.
Thompson, R., Lamb, M., & Estes, D. (1982). Stability of
infant-mother attachment and its relationship to changing life
circumstances in an unselected middle-class sample. Child
Development. 53, 144-148.
Thompson, R., Lamb, M., & Estes, D. (1983). Harmonizing
discordant notes: A reply to Waters. Child Development, 54,
521-524.
Vaughn, B., Egeland, B., Sroufe, L.A., & Waters, E. (1979).
Individual differences in infant-mother attachment at twelve
and eighteen months: Stability and change in families under
stress. Child Development, 50. 971-975.
Vaughn, B., Gove, F., & Egeland, B. (1980). The relationship
between out-of-home care and the guality of infant-mother
attachment in an economically disadvantaged population. Child
Development, 51, 1203-1214.
Waters, E. (1978). The reliability and stability of
individual differences in infant-mother attachment. Child
Development, 49, 483-494.
Waters, E. (1983). The stability of individual differences in
infant attachment: Comments on the Thompson, Lamb, and Estes
contribution. Child Development, 54, 516-520.
Waters, E., Vaughn, B., & Egeland, B. (1980). Individual
differences in infant-mother attachment relationships at age
one: Antecedents in neonatal behavior in an urban,
economically disadvantaged sample. Child Development, 51,
208-216.

74
Waters, E., Wippman, J., & Sroufe, L.A. (1979). Attachment,
positive affect, and competence in the peer group: Two
studies in construct validation. -Child Development, 50, 821 —
829.
Willemsen, E., Flaherty, D., Heaton, C., & Ritchey, G. (1974).
Attachment behavior of one-year-olds as a function of mother
vs. father, sex of child, session and toys. Genetic
Psychology Monographs, 90, 305-324.

BIOGRAPHICAL SKETCH
Roberta Ann Isleib was born on January 14, 1953, in
Hackensack, New Jersey. She was the second of four children
born to Janet Burdette and Charles Robert Isleib. She and her
family adapted to several cross country moves by spending time
together: reading, camping, traveling. She entered Princeton
University in 1971 with plans to prepare for medical school,
but amended these in favor of majoring in French literature
and art history. After graduation and a year of travel and
work, she attended the University of Tennessee and obtained a
master's degree in vocational rehabilitation counseling. The
following two years were spent in the foothills of the Great
Smoky Mountains as a rural rehabilitation counselor. In
1980, she moved to Florida to enter the doctoral program in
clinical psychology. She recently completed her internship at
Yale University, working as a primary clinician on an
adolescent inpatient unit. Currently, Ms. Isleib is employed
at the Yale University Health Plan, Division of Mental
Hygiene, as a postgraduate fellow.
75

I certify that I have read this study and that in ray
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Suzanne
Ü-S.
Suzáhne Bépnett Johnson, Chairman
Associate Professor of Clinical
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Jaitoés H. Johnson
Associate Professor of Clinical
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
i
Andrew Bradlyn
Assistant Professor o 1
Clinical
Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Patricia Miller
Associate Professor of Psychology
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Randy Carter
Associate Professor of Statistics

This dissertation was submitted to the Graduate Faculty of the
College of Health Related Professions and to the Graduate
School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
December, 1985
Dean, College of Health Related
Professions
Q.
Dean, Graduate School

UNIVERSITY OF FLORIDA
IIIIII 'inte M Val A
3 1262 08564 7924



61
taxonomy rests on the provision of standardized materials
which would allow the assumption of uniform ratings across
research laboratories and theoretical camps.


TABLE 5
Comparison of Means of Strange Situation Measures
for Sessions One and Two and Correlations Between Sessions
Measure
Session 1
Session 2
Significance
r
Mean
Mean
of Difference
Proximity-
Seeking
3.69
3.69
n.s.
0.54
(0.0024)
Contact-
Maintaining
3.48
3.69
n.s.
0.65
(0.0001)
Resistance
1 .67
1 .48
n.s.
0.23
(0.2263)
Avoidance
2.03
1.93
n.s.
0.39
(0.0354)


16
resulted in the elimination of eight subjects in subgroups B1
and B4 from the analyses.
In order to evaluate the construct validity of the
attachment concept, Waters (1978) examined the stability of
individual attachment behaviors, ratings of behavior
categories, and overall group classification in a middle-class
sample. Fifty infants were assessed at ages 12 and 18 months,
utilizing the Ainsworth strange situation paradigm. Time
samples of discrete behaviors and crying were recorded, as
well as behavioral ratings of five "attachment behavior
categories" (i.e., proximity-seeking, contact-maintaining,
proximity and interaction avoiding, contact resisting, and
distance interaction). Infants were also assigned to overall
categories (A, B, or C) as well as subgroups, based on their
strange situation performance. Behaviors were analyzed
separately for preseparation, separation, and reunion
episodes. Waters found that only four of 21 correlations
between discrete behavior scores at 12 and 18 months reached
the level of significance, and suggested that their occurrence
in the strange situation was frequently too low to allow for
adequate reliability. However, when ratings of interactive
behavior categories were considered, 13 of 18 correlations
were significant, including all of the behaviors directed at
the mothers in the reunion episodes. In terms of attachment
classification, 96% of the children maintained the same
attachment pattern over the six month period (kappa = 0.926),
while 30 of 50 subjects were also classified in the same


initial training period. Special thanks are due to Penny
Phelps, Joy Colle, and Katie Kato, who played the difficult
role of the stranger. In addition, Rae Hendlin generously
filled in for one session on a last minute basis, and Roman
Urbana served as an experimenter throughout the study.
Scoring the videotapes proved to be an enormous task,
made possible initially by the training provided by Everett
Waters. Thanks are also due to Mary Jo Ward for her
reliability checks, and most certainly, to Susan Grajek,
without whom the data analysis would have seemed overwhelming.
Don Quinlan kindly allowed me access to his computer, such
that typing and editing the manuscript became manageable.
My most earnest thanks go to my father, C. Robert Isleib,
who painstakingly coded a seemingly endless number of
videotapes, and tolerated my irritability with equanimity in
times when the project lagged. This dissertation is dedicated
to him, not only for his invaluable assistance during this
year, but also in recognition of the unselfish support and
encouragement he has given me in many ways over the past 32
years.
Roberta A. Isleib
New Haven, CT.
September, 1985
iv


41
In light of the general lack of support for setting
effects, analyses were conducted to compare trailer/trailer
subjects with all changed context subjects (TH and HT). Th
distribution of infants for the two groups at test and retest
is described in Table 1. First, the likelihood that a
particular classification was more frequent in one of the
groups at either test or retest was examined. Chi-square
tests were nonsignificant. Fisher's exact tests applied to
the A plus C versus B distribution at both sessions were also
nonsignificant. Similarly, chi-square tests demonstrated no
tendency for retaining classification to be more likely in one
group as compared with the other. As in the previous section,
t-tests were used to compare individual interactive ratings in
both episodes in each session. There were no differences. In
addition, no differences were found on comparisons of any of
the interactive rating difference scores.
As a final test of differences in stability between the
two groups, kappa was computed for each. The values were
-0.1409 and -0.1572, for TT and TH/HT, respectively. As
suggested by Cohen (1960), a two way comparison was made by
evaluating the normal curve deviate, with no significant
differences emerging (z = 0.03, p > 0.05).
As no significant differences in stability between TT and
TH/HT groups were found, all subjects were combined for the
following analyses. First, of the total sample in the first
session, 25 infants (83%) were classified as securely
attached, while three infants were rated as avoidant (10%),


BIOGRAPHICAL SKETCH
Roberta Ann Isleib was born on January 14, 1953, in
Hackensack, New Jersey. She was the second of four children
born to Janet Burdette and Charles Robert Isleib. She and her
family adapted to several cross country moves by spending time
together: reading, camping, traveling. She entered Princeton
University in 1971 with plans to prepare for medical school,
but amended these in favor of majoring in French literature
and art history. After graduation and a year of travel and
work, she attended the University of Tennessee and obtained a
master's degree in vocational rehabilitation counseling. The
following two years were spent in the foothills of the Great
Smoky Mountains as a rural rehabilitation counselor. In
1980, she moved to Florida to enter the doctoral program in
clinical psychology. She recently completed her internship at
Yale University, working as a primary clinician on an
adolescent inpatient unit. Currently, Ms. Isleib is employed
at the Yale University Health Plan, Division of Mental
Hygiene, as a postgraduate fellow.
75


59
First, there has been wide variability reported across
studies. Second, the differences found by various researchers
are not clearly explicable on the basis of sample
characteristics (e.g., note the wide discrepancies in
estimates of stability found in similar middle-class samples).
Finally, attempts to link stability/instability to external
variables have not been definitive: while Vaughn et al.
(1980) found changes from secure to anxious attachment to be
associated with life stress, Egeland and Farber (1984) were
unable to replicate this in a larger sample. Likewise,
Thompson et al. (1982) reported bidirectional changes in
attachment status to be related to changes in mothers'
employment status and caregiving disruption, while Owen et al.
(1984) found increased stability in infants whose mothers
returned to work during the second year.
In spite of problems such as those cited above, the
strange situation is a highly popular and visible metholology.
A brief review of the abstracts from the Society for Research
in Child Development catalog (biennial meeting, 1985)
demonstrates that of the 32 articles discussing attachment, 31
or 97% utilized this methodology. The only author who did not
use the ABC taxonomy was Everett Waters, who is in the process
of developing a Q-sort instrument for assessing attachment.
A cursory look at the major figures producing attachment
research reveals that they fall roughly into those associated
with Ainsworth through training or collaboration (e.g.,
Sroufe, Waters, Vaughn, Main, Egeland) and those not (e.g.,


70
Belsky, J., Garduque, L., & Hrncir, E. (1984). Assessing
performance, competence, and executive capacity in infant
play: Relations to home environment and attachment.
Developmental Psychology, 20, 406-417.
Belsky, J., Rovine, M., & Taylor, D. (1984). The Pennsylvania
infant and family development project, III: The origins of
individual differences in infant-mother attachment: Maternal
and child contributions. Child Development, 55, 718-728.
Belsky, J., & Steinberg, L.D. (1978). The effects of day
care: A critical review. Child Development, 49, 929-949.
Bowlby, J. (1958). The nature of a child's tie to his mother.
International Journal of Psychoanalysis, 39, 350-373.
Bowlby, J, ( 1 969 ). Attachment and loss. Volume 1 :
Attachment. New York: Basic Books.
Bowlby, J. ( 1973 ). Attachment and loss. Volume 2:
Separation. New York; Basic Books.
Bowlby, J. (1977). The making and breaking of affectional
bonds. I. Aetiology and psychopathology in the light of
attachment theory. British Journal of Psychiatry, 130, 201-
210,
Cairns, R.B. (1966). Attachment behavior of mammals.
Psychological Review, 73, 409-426.
Cassidy, J., & Ainsworth, M.D. (1983). The toddler's ability
to negotiate the environment: Its relation to quality of play
and quality of attachment. Presented at the Society for
Research in Child Development, Detroit.
Chase-Lansdale, P.L. (1981). Maternal employment in a family
context: Effects on infant-mother and infant-father
attachments. Dissertation Abstracts International, 42, 2562-A.
Cohen, J. (1960). A coefficient of agreement for nominal
scales. Educational and Psychological Measurement, 1_, 37-46.
Connell, D.B. (1978). Individual differences in attachment:
Long-term stability and relationships to early language
development. Dissertation Abstracts International, 38,
3954-B.
Connell, J.P., & Goldsmith, H. (1982). A structural modeling
approach to the study of attachment and strange situation
behaviors. In R. Emde & R. Harmon (Eds.), The development of
attachment and affiliative systems. New York: Plenum Press,
213-243.


48
TABLE 4
Means and Standard Deviations for Interactive Ratings
Test
Episode 5
Episode
! 8
Paired
Comparisons
Mean
SD
Mean
SD
p
Proximity-
Seeking
2.93
1.98
4.45
1.49
0.0002
Contact-
Maintaining
2.57
1.90
4.38
2.25
0.0001
Resistance
1 .43
0.93
1.90
1 .46
0.0377
Avoidance
2.50
1 .71
1.57
1.30
0.0017
Proximity-
Seeking
3.66
1.93
Retest
3.72
1.92
0.7636
Contact-
Maintaining
3.43
2.40
3.95
2.32
0.0122
Resistance
1.38
0.92
1 .59
1.17
0.3392
Avoidance
1.90
1 .45
1 .97
1.38
0.7869


43
and two were rated resistant (7%). Although this distribution
is weighted more heavily with secure infants than often found
in the literature, it is comparable to some studies using
middle class samples (e.g., Owen et al., 1984).
The overall stability was 69% (kappa = -0.116). Table 2
describes the distribution of infants at test and retest in
the current study, as compared with Ainsworth et al.'s (1978)
sample. There was no clear trend in direction of change.
Five children changed from secure to anxious attachment (three
B/A and two B/C) and four switched from anxious to secure (two
A/B and two C/B). Twenty-eight percent of the infants
retained the same subgroup classification from the first to
second session. (See Table 3.) Intraclass correlation
coefficients for proximity-seeking, contact-maintaining,
resistance and avoidance were 0.41, 0.56, 0.12, and 0.14,
respectively, for episode 5, and 0.42, 0.57, 0.28, and 0.14,
for episode 8.
Attachment behaviors (proximity-seeking and contact-
maintaining) tended to increase across sessions in episode
five (DproxS = -0.7241; DcontS = -0.8448) and decrease in
episode eight (Dprox8 = 0.7241; Dcont8 = 0.4828). Avoidance
tended to decrease across sessions in episode five (Davoid5 =
0.5517) and increase in episode eight (Davoid8 -0.4138).
Means and standard deviations for the interactive ratings are
presented in Table 4. As illustrated, in session one,
proximity-seeking, contact-maintaining, and resistance
increased significantly from episode five to eight, while


27
behavior of the mothers at three months. She suggested that
mothers of irritable babies tended to grow less responsive
over the first year, particularly under conditions of low
social support.
In line with Crockenberg's findings on irritability and
attachment, Belsky, Rovine, and Taylor (1984) evaluated the
relationship of reciprocal interaction at one, three, and nine
months to attachment classification at one year. Securely
attached infants experienced moderate levels of stimulation
over their first year of life, while avoidant infants were
overstimulated and resistant infants were understimulated.
Both types of insecure infants were found to be more fussy at
three and nine months, suggesting that infant temperament may
play a role in the development of maladaptive caretaking
patterns. Similarly, Egeland and Farber (1984) found that in
a lower SES sample, infants classified as resistantly attached
at one year had been rated as less alert and responsive during
feeding and play during their first few months of life.
Waters, Vaughn, and Egeland (1980) administered the NBAS
to 100 first-born neonates at age seven and ten days, then
administered the strange situation at one year. Results
indicated that anxious/resistant (Group C) infants scored
lower on orientation, motor maturity, and regulation items at
day seven. These differences had disappeared by day ten. The
authors suggested that these results supported other evidence
linking the more difficult nature of group C children with
difficult interaction and the development of anxious
attachment.


ACKNOWLEDGMENTS
The planning and execution of this manuscript have been
extremely complex and demanding. It could not have been
completed without the assistance of many people along the way.
Thanks go first to Suzanne Johnson, Roger Blashfield, Jim
Johnson, Pat Miller, Drew Bradlyn, and Randy Carter for their
supportive and constructive comments during the designing
stages of the study. Suzanne also served as an anchor
throughout my internship year, encouraging me to continue to
make progress in spite of various setbacks and the pull to
procrastinate.
Identifying and recruiting subjects was a major hurdle,
for which the input of Jim Siwi, Keith Berg, and Janice Benton
is gratefully acknowledged. Thanks also go to the many
friends who told others about my research and recruited
subjects by word of mouth. Prior to data collection, Jerry
Rudd contributed skill and labor while installing a one-way
mirror and Dan McNeil served as a cheerful video equipment
consultant.
Ross Thompson kindly reviewed several strange situation
videotapes and made helpful suggestions about the physical
environment, stranger behavior, and timing of the episodes.
Fi Kieffer and her son Christopher, and Dawn Bowers and her
daughter Martina agreed to serve as guinea pigs during the
iii


2
characterized by active protest, despair or hopelessness, and
finally, detachment and apathy* As Ainsworth has been
instrumental in translating Bowlby*s ideas into research, as
well as contributing to their elaboration and clarification
(e*g., Ainsworth, 1964; 1973; 1979a), the theory has become
known as the Bowlby-Ainsworth ethological model of attachment.
The Bowlby-Ainsworth model views attachment as a species
general behavior system whose function is to protect the
relatively helpless human infant from predation. Infants are
born with a biological predisposition to seek contact with
adults, which is manifested early on by signalling behaviors
such as crying, smiling, and calling, in addition to searching
and following behaviors as locomotion develops. Adults are
equipped with complementary predispositions to protect
infants* Bowlby originally suggested that the development of
a specific attachment was mediated by the amount of
interaction occurring between the infant and attachment
figure. As a cognitive appreciation for person permanence was
assumed to be a necessary prerequisite, the development of
attachment was not expected to occur until the second half of
the first year of life.
Drawing on control systems theory, Bowlby incorporated
the notion of a set-goal into his attachment model. The set-
goal, defined as the amount of distance from the attachment
figure which the infant will tolerate, is mediated by a
variety of internal (e.g., illness, hunger) and external
(e.g., stress, separation) factors. When the set-goal is
exceeded, the infant should respond with the activation of


44
TABLE 2
Classification of Infants at Test and
for the Current Study and Ainsworth's
Current Study
Test A
B
C
Retest
ABC
0 2 0
3 20 2
0 2 0
3 24 2
(0.10) (0.83) (0.07)
Test A
B
C
Ainsworth Sample
Retest
B C
7 0
A
0
0
0
0
(0.00)
12
1
20
(0.87)
2
1
3
(0.13)
Retest
Sample
2
(0.07)
25
(0.86)
2
(0.07)
29
7
(0.30)
1 4
(0.61 )
2
(0.09)
23


15
Only two other studies have repeated the strange
situation over a similarly short time interval. These results
will be briefly summarized, although both studies investigated
attachment towards mothers versus fathers, and neither
followed the Ainsworth scoring system faithfully. Willemsen,
Flaherty, Heaton, and Ritchey (1974) contrasted attachment
behaviors in one-year-old infants to mothers versus fathers.
These authors found more attachment behavior was directed to
the parent in the presence of the stranger in session 2, as
well as more distress (particularly crying) being exhibited.
As in the Ainsworth et al. test-retest study, these results
were attributed to the association of the experimental room
with the stress of separation. Lamb (1978) also investigated
patterns of attachment to mothers versus fathers during
testing sessions scheduled one week apart. Results
demonstrated no order effects; i.e., there was not a tendency
for any classification or behavior to be more or less common
in either session.
Connell (1978) tested 55 white, middle-class infants
using the strange situation at 12 and 18 months. He found
significant correlations in measures of contact-maintaining,
proximity- seeking, resistance, and avoidance across the two
sessions. Behavior towards the stranger was less consistent,
with significant correlations emerging in resistance,
avoidance, and proximity seeking. Approximately 81% of the
infants were classified in the same ABC group six months later
(kappa = 0.603), although Connell used techniques which


APPENDIX C
INSTRUCTIONS TO PARENTS
Duration
of Episode
30 seconds
3 minutes
3 minutes
3 minutes
(or less)
3 minutes
3 minutes
(or less)
Description of Episode
1. MOTHER, BABY, AND EXPERIMENTER
Experimenter will accompany you and your
child to the playroom, and show you your
seat*
2. MOTHER AND BABY
Try to interest your baby in toy play,
then return to your chair. Respond
naturally and appropriately to your
child's initiatives, but try not to
initiate interaction on your own.
3. MOTHER, BABY, AND STRANGER
First minute: stranger sits quietly
Second minute: stranger chats with
mother
Third minute: stranger begins to play
with baby
4. BABY AND STRANGER
Knock on window signals your departure.
Walk quietly to the observation room
door and leave the room, closing the
door behind you.
5. MOTHER AND BABY
Talk loudly before opening the door so
baby knows you're coming. Pause in
doorway to allow baby to greet you.
Comfort and soothe your child, if
necessary. When baby is ready, try to
reinterest him/her in toy play, and
return to your chair.
6. BABY ALONE
Knock on window signals your departure.
Walk quietly to the observation room
door, say "good-bye" to your baby, and
leave the room, closing the door.


28
Finally, two studies investigated the relationship of
family stress, as measured by the Parenting Stress Index, and
patterns of attachment (Green, 1981; Hamilton, 1981). In
neither of these white, middle-class samples was a significant
relationship between these variables found.
To summarize, the majority of research done to date has
not found that maternal employment negatively affects security
of infant attachment. A major exception is the study by
Vaughn et al. (1980), which demonstrated that early out-of-
home care in a disadvantaged population resulted in a larger
proportion of anxiously attached children. Research
investigating long-term attachment stability (over a six month
period) has suggested that changes in attachment
classification may be associated with lower socioeconomic
status, high life stress, and changes in caregiving
circumstances. As a cautionary note, replication across
studies on these variables has been problematic, even within
research groups working with longitudinal studies. In
addition, some of the results have been directly contradictory
(e.g., Owen et al.'s (1984) finding that mothers who returned
to work in the second year had offspring with more stable
attachments, as contrasted with Thompson et al. (1982) who
found return to work to be associated with changes in
classification).
Aims and Hypotheses of the Study
The present study has two aims. First, given the rather
weak evidence available for the stability of the ABC typology


47
(N =20) versus those infants who changed groups from test to
retest (N = 9). Using t-tests, no significant differences
were found between the groups on the following variables:
social support, life stress, maternal temperament, infant
temperament, caretaking changes, parental age and hours
worked, and separations from mother over the first year of
life. In contrast, infants who retained their classification
were found to come from lower income families (M = $17704)
than those who changed groups (M = $27700; t = 2.1503, p <
0.0403). This difference is contradictory to findings in the
literature suggesting a higher rate of stability in higher
income families, and is likely to be a spurious finding
resulting from the large number of statistical tests employed.
A series of chi-square tests evaluated the association
between stability of classification and categorical variables.
No differences were found based on sex of subject, birth
order, level of parental education, maternal work status, or
changes in caretakers over the first year.
To summarize the previous section, actual setting in which
the strange situation occurred had no effect on classification
in a particular setting or on stability of classification. In
addition, there were no differences in the stability of
classification in the same context group (TT) as compared with
the changed context group. Stability for the sample
considered as a whole was unimpressive (k = -0.166). Finally,
when infants who retained their group status were compared
with those who changed across sessions, no significant


74
Waters, E., Wippman, J., & Sroufe, L.A. (1979). Attachment,
positive affect, and competence in the peer group: Two
studies in construct validation. -Child Development, 50, 821-
829.
Willemsen, E., Flaherty, D., Heaton, C., & Ritchey, G. (1974).
Attachment behavior of one-year-olds as a function of mother
vs. father, sex of child, session and toys. Genetic
Psychology Monographs, 90, 305-324.


64
confidentiality of this document and your records from this
research to the extent provided by law.
I have fully explained to the
nature and purpose of the above-described procedure and the
benefits and risks that are involved in its performance. I
have answered and will answer all questions to the best of
my ability. I may be contacted at telephone # .
Signature of person obtaining consent Date
I have been fully informed of the above-described procedure
with its possible benefits and risks and I have received a
copy of this description. I have given permission for my/my
child's participation in this study.
Signature of relative, parent, or guardian Date
Witness to signature
Date


TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS iii
LIST OF TABLES vi
ABSTRACT vii
INTRODUCTION 1
The Bowlby-Ainsworth Theory of Attachment 1
The Strange Situation 4
Predictive Validity of the ABC Typology 11
Stability of the ABC Classification System 13
Effects of Family Environmental Variables 24
Aims and Hypotheses of the Study 29
METHOD 31
Subj acts 31
Materials... 31
Procedure 35
RESULTS 38
Reliability 38
Stability 38
Family Environmental Variables 50
DISCUSSION 51
APPENDIX A INFORMED CONSENT 62
APPENDIX B STRUCTURED INTERVIEW 65
APPENDIX C INSTRUCTIONS TO PARENTS 67
REFERENCES 69
BIOGRAPHICAL SKETCH 75
v


50
differences emerged on demographic or family environmental
variables. One exception to this was family income. Due to
the small sample size, the group of infants which changed
classification could not be sorted according to direction of
change (e.g., secure to anxious versus anxious to secure).
Family Environmental Variables
The second aim of this study was to investigate the
relationship of a variety of family environmental variables
to quality of attachment. For these analyses, only the
classifications from the first session were utilized. The
group of securely attached infants (N = 25) was compared with
those children who were anxiously attached (N = 5), which
included both avoidant (n = 3) and resistant (n = 2) infants.
T-tests were used to compare these groups on measures of
social support, life stress, maternal temperament, infant
temperament, hours that mother and father spent caring for the
baby, caretaking changes, and income. There were no
significant differences except for the maternal temperament
factor of activity. On this variable, mothers of securely
attached infants tended to score lower than did mothers of
anxiously attached infants (t = -2.7282; p < 0.0109). A chi-
square test evaluating the relationship between attachment and
maternal work status also proved to be nonsignificant.