International Journal of Behavioral
Development
http://jbd.sagepub.com
The quality of maternal secure-base scripts predicts children's secure-base behavior at home in
three sociocultural groups
Brian E. Vaughn, Gabrielle Coppola, Manuela Verissimo, Ligia Monteiro, Antonio José Santos, German Posada, Olga
A. Carbonell, Sandra J. Plata, Harriet S. Waters, Kelly K. Bost, Brent McBride, Nana Shin and Bryan Korth
International Journal of Behavioral Development 2007; 31; 65
DOI: 10.1177/0165025407073574
The online version of this article can be found at:
http://jbd.sagepub.com/cgi/content/abstract/31/1/65
Published by:
http://www.sagepublications.com
On behalf of:
International Society for the Study of Behavioral Development
Additional services and information for International Journal of Behavioral Development can be found at:
Email Alerts: http://jbd.sagepub.com/cgi/alerts
Subscriptions: http://jbd.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations (this article cites 21 articles hosted on the
SAGE Journals Online and HighWire Press platforms):
http://jbd.sagepub.com/cgi/content/abstract/31/1/65#BIBL
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
© 2007 The International Society for the
Study of Behavioural Development
International Journal of Behavioral Development
2007, 31 (1), 65–76
DOI: 10.1177/0165025407073574
http://www.sagepublications.com
The quality of maternal secure-base scripts predicts children’s
secure-base behavior at home in three sociocultural groups
Brian E. Vaughna, Gabrielle Coppolaa, Manuela Verissimob, Ligia Monteirob, Antonio José Santosb,
German Posadac, Olga A. Carbonelld, Sandra J. Platad, Harriet S. Waterse, Kelly K. Bostf,
Brent McBridef, Nana Shinf, and Bryan Kortha
The secure-base phenomenon is central to the Bowlby/Ainsworth theory of attachment and is also
central to the assessment of attachment across the lifespan. The present study tested whether
mothers’ knowledge about the secure-base phenomenon, as assessed using a recently designed wordlist prompt measure for eliciting attachment-relevant stories, would predict their children’s securebase behavior, as assessed by observers in the home and summarized with the Attachment Q-set
(AQS). In each of three sociocultural groups (from Colombia, Portugal, and the US), scores characterizing the quality of maternal secure-base narratives elicited using the word-list prompt procedure
were internally consistent, as indicated by tests of cross-story reliability, and they were positively and
significantly associated with the child’s security score from the AQS for each subsample. The correlation in the combined sample was r(129) = .33, p < .001. Subsequent analyses with the combined
sample evaluated the AQS item-correlates of the secure-base script score. These analyses showed that
mothers whose stories indicate that they have access to and use a positive secure-base script in their
story production have children who treat them as a “secure base” at home. These results suggest that
a core feature of adult attachment models, in each of the three sociocultural groups studied, is access
to a secure-base script. Additional results from the study indicate that cross-language translations of
the maternal narratives can receive valid, reliable scores even when evaluated by non-native speakers.
Keywords: attachment script representation task; secure bare relationships
a Auburn University, USA. b Instituto Superior de Psicologia Aplicada,
Portugal. c Purdue University, USA. d Universidad Javeriana, Bogotá, Colombia.
e State University of New York at Stony Brook, USA. f University of Illinois,
Champaign-Urbana, USA.
giver, the secure-base phenomenon implies sensitive and cooperative attention to the location and state of the attached
person and, on the side of the attached person, it implies the
belief that the caregiver is able and willing to intervene on her
or his behalf if needed and/or called. This has prompted some
attachment theorists to refer to attachments as “secure-base
relationships” (E. Waters & Cummings, 2000). We adopt this
usage in this article.
A second central assumption of attachment theory is that
core features of the child–parent attachment relationship
become abstracted and represented internally (i.e., as “internal
working models” in Bowlby’s terminology) as the child (and
the attachment relationship) matures (Bowlby, 1973). Bowlby
believed that internal working models were relatively open in
the sense that they could be revised in the face of changes in
the attachment relationship in the early years of life, which
might arise as a consequence of changes in the social–
emotional or physical circumstances of the dyad or as the child
acquired more and more sophisticated social and relational
competencies (Bretherton & Munholland, 1999; Sroufe,
Egeland, Carlson, & Collins, 2005). Furthermore, as the child’s
cognitive and linguistic capacities expand, the working model
can assimilate new information about the attachment figure
Correspondence should be sent to Brian E. Vaughn, 203 Spidle Hall,
Human Development and Family Studies, Auburn University,
Auburn, AL 36849–5604, USA; e-mail: vaughbe@auburn.edu
The studies reported here were supported in part by NSF grants
BCS 0126163 and BCS 0126427. Support for the Portuguese data
collection and analysis was provided by FEDER grant
POCTI/1999/PSI/36429. Additional support was provided by a grant
from the Center for Mental Health Promotion, Bayport, NY. The
authors express appreciation to the Director and staff of the Harris
Early Learning Center in Birmingham, Alabama for their support of
this project. We also express our greatest appreciation to Lisa Krzysik
for contributions to the data collection and data organization for this
report. We also thank Arnoldo Aristizabal at the Javeriana University
for his essential support in conducting this research in Bogotá.
Central to the Bowlby/Ainsworth theory of attachment is the
assumption that the attachment figure is a secure base for the
attached individual’s exploration and is a haven of safety for
the attached person in times of stress (Ainsworth, 1967;
Ainsworth, Bell, & Stayton, 1974; Ainsworth, Blehar, Waters,
& Wall, 1978; Ainsworth & Bowlby, 1991; Bowlby, 1973,
1988; E. Waters & Cummings, 2000). The secure-base
phenomenon is evident in attachments across the lifespan,
although for adults the role of secure-base provider and
receiver can cycle between relationship partners, depending on
conditions and need (Crowell et al., 2002); however, during
childhood the adult member of the (healthy) attachment
relationship is the child’s secure base. However else they may
be characterized, attachment relationships always imply the
secure-base phenomenon (Ainsworth et al., 1978; Bowlby,
1988; E. Waters & Cummings, 2000; H. Waters, Rodrigues, &
Ridgeway, 1998; Zimmermann, 2004). On the side of the care-
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
66
VAUGHN ET AL. / MATERNAL SECURE BASE SCRIPTS
and the secure-base relationship through various communicative modalities. Parents frequently take advantage of these
cognitive/linguistic advances as they tell the child in both words
and deeds that he or she is loved and protected, or not (see
Oppenheim & H. Waters, 1995; E. Waters & Cummings, 2000,
for extended discussions).
Presumably, attachment-relevant stories that individuals tell
must reflect, at least in part, the representations of attachment
(i.e., working models) derived from their own experiences in
secure-base relationships (Bowlby, 1973; Bretherton &
Munholland, 1999; Bretherton, Ridgeway, & Cassidy, 1990;
Carlson, Sroufe, & Egeland, 2004). To the extent that parents’
internal representations of attachment inform and influence
parent–child interactions and the secure-base relationships
emergent from those interactions (Karavasilis, Doyle, &
Markiewicz, 2003; Pederson, Gleason, Moran, & Bento, 1998;
Posada, Waters, Crowell, & Lay, 1995; Ward & Carlson, 1995),
we should expect to find that attachment-relevant stories the
parents tell are associated with other indicators of attachment
representation and with the organization of their children’s
attachment behavior. The primary purpose of this article is to
test the expectation that the quality of maternal secure-base
stories and the organization of child secure-base behavior are
related, using a recently constructed instrument for eliciting
maternal secure-base stories (H. Waters & RodriguesDoolabh, 2004) and a widely used, well-understood measure
of child secure-base behavior for preschool children, the
Attachment Q-set (AQS; van IJzendoorn, Vereijken, Bakermans-Kranenburg, & Riksen-Walraven, 2004; Vaughn & E.
Waters, 1990; E. Waters, 1995) as the child measure. The AQS
has been used and validated as an attachment indicator for
preschool-age children (Park & Waters, 1989; Posada et al.,
1995), and van IJzendoorn et al. (2004) suggested that the
measure constituted one of three “gold standard” attachment
measures (with the Strange Situation and the Adult Attachment Interview).
H. Waters and Rodrigues-Doolabh (2004) explicitly
designed their measure to assess the respondent’s awareness
and use of secure-base knowledge in the context of telling
stories about everyday events involving parent–child dyads or
adult couples. They reasoned that adults whose personal
histories were characterized by secure attachments would have
had many opportunities to receive secure-base support from
an attachment figure and, most likely, to have served as the
secure base for a dyad partner (child or adult) as well. H.
Waters and Rodrigues-Doolabh (2001) argued that one result
of a lifetime of experience having and being a secure base is
the construction of an abstracted summary of how the securebase phenomenon is experienced and that the structure of this
internal representation would resemble a cognitive script, analogous to scripted knowledge about events at a birthday party
or behavior in a restaurant, that could be primed with relevant
prompts (see also, Bretherton, 1991).
The “script” concept has a long history in cognitive psychology research. In general, knowledge structures abstracted from
recurring everyday life events (in terms of temporal sequences
and expectations about likely outcomes) are considered scripts
(Bretherton, 1991; Fivush & Hudson, 1990; Nelson, 1986;
Nelson & Hudson, 1988; Oppenheim & H. Waters, 1995). H.
Waters and Rodrigues-Doolabh (2001) argued that scripted
knowledge concerning having and/or being a secure base for
another would entail several elements elaborated in roughly the
following sequence: some constructive engagement between
members of an attached dyad; an obstacle to continued
engagement is encountered; a signal that help is needed is
given by one partner, the other partner detects the signal; effective help is offered; the assistance is experienced by the receiver
as comforting; resolution and/or return to constructive engagement in the social or physical environment. Their measure is
designed to prime this script by providing word lists that could
be construed as an outline for the secure-base story. If an adult
has access to the secure-base script and uses it, a story based
on the work prompts would include many or most of the
elements in the aforementioned sequence.
In an initial test of the utility of the word-list prompt
measure, H. Waters and Rodrigues-Doolabh (2001) reported
correlations from .50 to .62 when scores derived from the
narratives elicited by the prompts were correlated with the
Coherence scale score from the Adult Attachment Interview
(AAI) (Hesse, 1999; Main & Goldwyn, 1998) for a group of
mothers. Coherence is considered the central dimension for
classifying individuals as secure on the AAI and their results
suggest that possession of and access to a well-articulated
secure-base script is a part of what it means to think coherently about attachment. More recently, Tini, Corcoran,
Rodrigues-Doolabh, and E. Waters (2003) demonstrated that
maternal secure-base scriptedness was associated with
children’s classifications in the Strange Situation, and
Guttmann-Steinmetz, Elliot, Steiner, and H. Waters (2003)
reported that mothers with higher secure-base script scores
were better able (than mothers with lower scores) to help their
child co-construct stories about attachment relevant content.
Finally (and importantly for the current report) RodriguesDoolabh, Zevallos, Turan, and Green (2003) showed that
mothers from diverse ethnic and cultural groups (including
samples from Peru, Switzerland, Turkey, United Arab
Emirates, Zimbabwe, and the US) produced detailed and
explicit secure-base narratives when presented with the wordprompt lists used with the initial US samples, although certain
of these word-prompt lists were modified, or new lists substituted, in some groups to conform with cultural practices. In
these samples, composite scores based on four secure-base
stories had acceptably high internal consistency values
(Cronbach’s alpha). A second purpose of this article is to add
two non-English speaking sociocultural samples to those
reported on by Rodrigues-Doolabh et al. (2003), from
Colombia and Portugal, and to examine the psychometric
properties of the story score composites in these samples (as
well as assessing the relation between maternal script scores
and child secure-base behavior in each sample).
As suggested above, our sample is multinational and multicultural. It is important to test attachment hypotheses crossculturally for several reasons, perhaps especially because
Bowlby (1982) proposed his attachment model as applicable
to the species as a whole and not just to members of a single
culture. That is to say, Bowlby believed that the behavioral
system supporting co-construction of a secure-base relationship was an adaptation fixed in place through the process of
natural selection and that evidence of secure-base relationships, including mental representations derived from these
relationships, should be observable in any human sociocultural milieu. Although the bulk of evidence reported to
date (Posada et al., 1995; van IJzendoorn & Kroonenberg,
1988; and see van IJzendoorn & Sagi, 1999 for a review) is
consistent with Bowlby’s assumption, aspects of his general
model, and especially his assumptions about consequences of
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2007, 31 (1), 65–76
secure-base relationships, have been criticized as being
parochially Anglo-centric (Harwood & Miller, 1991; Harwood,
Miller, & Irizarry, 1995; Rothbaum, Pott, Azuma, Miyake, &
Weisz, 2000).
Central to the cultural criticisms of Bowby’s theory is the
notion that some cultures, perhaps especially Latin cultures
and certain oriental cultures, value an interdependent (versus
autonomous) interpersonal stance that is inconsistent with
Bowlby’s assumption that secure-base relationships foster selfreliance and independence. In such cultures, these critics
suggest either that parental preferences for children’s behavior
and attributes will lead to a different suite of interaction styles
and behaviors than those supporting attachment security, or,
that security itself will have a unique denotative meaning for
these cultures, corresponding to the preferred type for the
culture. If either of these interpretations is accurate, we might
find significantly different levels of secure-base scriptedness
and/or child secure-base behavior across sociocultural groups,
or, we might find different patterns of correlation between
maternal and child scores in the different groups. Having two
samples from different Latin cultures (i.e., Colombia,
Portugal) allows us to consider the possibility that cultural
differences influence both the form and the function of securebase representations and secure-base behavior.
Methods
Participants
These are convenience samples that were recruited independently for purposes somewhat different than those to which we
put them here and there are several differences between the
samples in terms their demographic indicators and of ages at
which some assessments were made (these are identified
below). Each sample was initially recruited so as to represent
“middle-class” populations from their respective sociocultural
communities and we would not expect that differences we
may identify could be interpreted in terms of education or
income levels. Nevertheless, the samples are embedded in
three distinct sociocultural milieu and may be expected to be
somewhat different with regard to child-rearing values and
goals that could influence their own secure-base knowledge
and their child’s secure-base behavior. It has become common
to characterize differences in cultural practices in terms of the
dimensions of individualism and collectivism (Oyserman,
Coon, & Kemmelmeier, 2002) that subsume parenting practices motivated by a value for autonomy (individualism)
versus interdependence (collectivism) in a given culture.
Oyserman et al. (2002) included samples from both
Colombia and Portugal in their meta-analyses and indicated
that the Portuguese culture was more collectivistic and less
individualistic than US culture, whereas Colombian culture
was higher than the US on both the individualistic and collectivistic dimensions.
Colombian sample. Twenty-five mother–child pairs (15 male
children) from a larger study (total N = 41) on parental
support for secure-base relationships (Posada et al., 2002)
comprised the Colombian sample. Participants from Colombia
came from middle-class neighborhoods in Bogotá, Colombia.
They were contacted originally through a health, housing, and
education provider with whom the families were associated
67
when the children were about 1 year of age (M = 12.4 months,
SD = 3.34 months). When initially enrolled, mothers identified themselves as the infant’s main principal caregiver, their
average age was 31.5 years (SD = 4.72 years), and their
education level ranged from high school to university degree
(7 had a high school degree, 10 had a technical degree, and 8
had a university degree). Eight mothers were homemakers and
the other 17 worked outside the home. All families were intact
and children lived with both parents. Mothers were their
children’s primary caregivers at home. When the children were
approximately 3 years of age, attempts were made to re-contact
the families to recruit them for the maternal word list narrative task. Twenty-seven of the 41 mothers agreed to participate
in this second task. Equipment failed during two interviews,
leaving a final sample of 25. These mothers are representative
of the larger sample in terms of age, education level, and work
status.
Portuguese sample
Fifty-eight mother–child dyads (29 female children) from a
larger study of relations between attachment and peer
competence comprised the Portuguese sample. Families were
recruited from five private childcare centers in a suburb of
Lisbon. All but five mothers were working full-time during the
period of the assessments and all mothers reported being the
primary caregiver at home. Families were predominantly
middle-class by local standards in terms of education levels
and job titles. All of the families were European. All children
were between 30 and 35 months of age at the time of the
assessments (M = 31.4 months, SD = 2.1 months). Mothers’
ages averaged 34 years, SD = 3.52. Mean years of education
was 15.6 years (some university, 60% with first university
degree), SD = 3 years.
US sample
Forty-seven mother–child dyads (25 male children) from a
larger study of social emotional development across the
preschool years comprised the US sample. Families were
recruited from two childcare centers in a major metropolitan
area from a state in the south-eastern region of the US. All
mothers were employed or in school 20 hours/week or more.
Their average age was 35.6 years (SD = 4.4 years). Families
were predominantly middle-class by local standards in terms
of education levels and job titles (over 60% of mothers
reported family incomes greater than $100K/year and only one
reported an annual income less than $25K), however, the
range of occupations was broad and included graduate
students and employees of the local city government as well as
a mix of managerial and professional job titles. Approximately
20% of families were of minority ethnic/racial status but their
occupational range was representative of the sample as a
whole. Over 85% of mothers were college graduates (more
than 50% had obtained professional or other post-graduate
degrees) and all but one reported attending college or obtaining some post-secondary professional training. For this sample,
racial/ethnic status and social class indicators were not
confounded. Families were recruited to the sample when their
children were between 2.0 and 3.0 years of age and all assessments were completed before the child reached 42 months of
age (M = 35.2 months, SD = 4.03 months). In the home,
mothers were primary caregivers for their children.
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
68
VAUGHN ET AL. / MATERNAL SECURE BASE SCRIPTS
Measures and procedures
Word-list prompt measure for secure-base scriptedness. The
measure designed by H. Waters and Rodrigues-Doolabh
(2004) primes the secure-base script using word-list prompts
(six different lists) that form the outline of a story. The stories
elicited by four of the word-prompt outlines are scored on the
basis of the presence (or absence) of the secure-base script and
the richness and detail concerning the relationship between
characters in the story. A single score summarizing both
presence and quality of the secure-base script is given for each
of the attachment-relevant narratives, with the average of these
being the subject’s “scriptedness” score for secure-base knowledge. Low scores (< 4 on the 7-point scale) indicate the general
absence of a secure-base script in the narratives and are
considered to be “insecure” with respect to attachment. The
lowest scores are reserved for stories that both do not include
the secure-base script and introduce unusual content into the
stories (e.g., a child who has been injured soothes the parent
who is upset about the injury). Higher scores (4 or above on
the 7-point scale) indicate the presence of the secure-base
script and are considered to be “secure”. The highest scores
are assigned when the secure-base script is elaborated, shows
evidence of awareness of the partner’s emotional state, reformulates the meaning of the obstacle/conflict in a favorable way,
and/or locates the present interaction in the context of the
ongoing relationship. The remaining two stories are not
relevant to the secure-base script (trip to the park, afternoon
shopping) and are not scored for secure-base content.
Prior to data collection, the word lists and the instructions
to mothers were translated into the relevant language (Spanish,
Portuguese) by fully bilingual translators. A different bilingual
translator back-translated these into English and the resulting
text was compared with the H. Waters and Rodrigues-Doolabh
(2004) original text. Cross-language discrepancies were identified and adjusted until the translations were precisely matched.
Although this method of forward and back-translation of the
text produced equivalence at the level of denotative meaning,
stories elicited in the Colombian sample indicated that the
word list for “Jane and Bob’s camping trip” had a different
meaning than in the other samples. In Colombia, couples
typically do not have the experience of going out to the
wilderness and away from relatives as a recreational activity
and many of the mothers did not understand the premise of
the story. In addition, the mothers from Colombia, more than
in either of the other two samples, tended to include children
in their stories about adult–adult relationships and the resulting stories were judged to be unscoreable by H. Waters and
L. Rodrigues-Doolabh. As a consequence, while we use all
scored stories to estimate the reliability of the composite scores
and cross-rater agreement in each of the subsamples, we only
analyze relations between the maternal scriptedness scores and
child security scores (AQS) for mother–child stories in this
report.
Narrative collection protocols. In each sample, mothers
responded to the six word-prompt outlines (H. Waters &
Rodrigues-Doolabh, 2004), which were presented one at a
time by a research staff member. Each outline consisted of
three columns of four words. Mothers were instructed to read
down each column from left to right to get a sense of the story
outline. The instructions also included an explanation that the
stories would be audiotaped and that the mother could stop a
story and start it over from the beginning if she chose to do
so. Four of the lists were intended to prime secure-base themes
(baby’s morning, doctor’s office, Jane and Bob’s camping trip,
Sue’s accident). Two of the secure-base word-prompt outlines
explicitly concerned mother–child content (e.g., baby’s
morning, doctor’s office) and two were designed to elicit
stories relevant to adult relationships (e.g., Jane and Bob’s
camping trip, Sue’s accident). As mothers were handed the
word-prompt outline, the research staff member identified the
expected story content as “mother–child” or “adult relationship”. Six different orders for the word-prompt outlines were
used, with the convention that the three mother–child wordprompt lists were presented as a cluster and the adult–adult
lists were presented as a cluster (i.e., a mother would respond
to all three stories in one cluster before being presented with
a story from the other cluster). Each list-order was used
approximately equally often in each sample and consecutively
assessed mothers were not given the stories in an identical
order.
For the Colombian sample, maternal narratives were
obtained in the context of a home visit that had been explicitly scheduled to complete this task. For the Portuguese
sample, maternal narratives were collected at the end of the
AQS home visit. When observations were completed, the
mother was asked to accompany one of the home visitors to
another room to complete the narrative task while the other
observer played with the child. In the US sample, the maternal
narrative assessment took place in the context of a laboratory
visit designed to measure qualities of child–parent interaction,
children’s emotion knowledge, and child self-control. After
approximately 25 min of parent–child interaction, the mother
was asked to leave the child with a member of the research staff
and go to an adjacent room to complete the word-prompt task.
In all three samples, most mothers completed this task in
15–20 min. Audio-recordings of the mothers’ stories were
transcribed for later scoring. For the US and Portuguese
samples, two to four persons blind with respect to the home
observation data, including the AQS, read and rated each story
transcript using the 7-point scale designed by H. Waters and
Rodrigues-Doolabh (2004). Stories for a single word-prompt
list (e.g., baby’s morning) were grouped together within each
sample (e.g., all baby’s morning stories for the US sample were
scored in the same session). Different stories were scored on
subsequent days to minimize the possibility that raters might
recognize stories from a single participant.
Colombian stories were translated into English, sent to the
US, and scored by H. Waters and L. Rodrigues-Doolabh at
SUNY Stony Brook. They were blind to all other data for the
participants. Intraclass correlations across stories for the two
raters ranged from .75 to .87. The US narratives were scored
by two native Portuguese speakers who were also fluent in
English and by a fully bilingual Italian/English speaker as well
as by a native English-speaking rater. Three raters had received
an intensive training in decoding the transcripts from H.
Waters. Intraclass correlations between raters ranged from .82
to .93 across all stories and 85% of the codes given by different raters were within one scale-point of codes given by the
other raters. Cronbach’s alpha for the total score (average
across all raters) for each story was > .94. These levels of rater
agreement suggest that raters need not be native English
speakers to master the rules for scoring English language
stories. The Portuguese story transcripts were rated by native
speakers (who also coded the US transcripts). The intraclass
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2007, 31 (1), 65–76
correlations ranged from .68 to .83 with over 90% of scores
being within 1 scale point. Spearman–Brown reliability estimates for individual stories ranged from .82 to .93. One
member of the coding team was blind to all AQS data. The
second was blind to half the sample but had participated in
home visits for the other half of the sample. Scores provided
by the privileged rater were not used in the composite scores
for primary analyses. Scores for the raters were averaged (see
exception noted previously) for each story. To further evaluate
rater agreement in the Portuguese sample, 30 stories from this
sample were translated into English by a professional translator and these transcripts were scored by a native English
speaker. Intraclass correlations for rater-pairs ranged from .52
to .87 and over 90% of the cross-language comparisons agreed
within 1.5 scale points with no pair of scores for a given story
differing by more than 2.5 scale points for any story score.
These analyses suggest that secure-base scriptedness can be
scored reliably from translated stories.
In each sample, a parent–child composite score was also
obtained by averaging the secure-base scriptedness scores for
the two parent–child stories. These composite scores were used
in analyses. A two-factor (Sample ⫻ Child Gender) ANOVA
for the composite score did not yield significant sample or
gender differences and the interaction of sample and child
gender was not significant, indicating that mean differences
among subsamples and between mothers of boys versus
mothers of girls were not meaningfully different. These results
provide partial justification for combining the samples in
subsequent analyses.
Attachment Q-set. The Attachment Q-set (E. Waters, 1995)
consists of 90 items relevant to the child’s use of the caregiver
as a secure base and haven of safety. Informants (observers in
the case of this article) describe the child using the 90 items
by sorting them into a specified distribution after some period
of observation, in terms of how relevant each item is as a
descriptor of the observed child (from very undescriptive of the
child to very descriptive of the child). It is important to recognize
that the observations upon which Q-sort descriptions are based
are not frequency counts of behaviors defined by some
taxonomy, but rather these observations contribute to
observers’ inferences about the general organization of the
child’s behavior with reference to using the caregiver as a
secure base and haven of safety. In this way, Q-items are analogous to items from a questionnaire that parents or teachers
might use to rate temperament, personality, or problem behaviors, with the caveat that instructions for most Q-sorts specify
the distribution of scores to categories on the scale (e.g., a fixed
number of items get scored at each point on the scale) rather
than allowing the respondent to assign a score at any position
along the scale for each item. When multiple observer/raters
are employed to describe a given child, it is possible to estimate
the reliability of the composite Q-sort (aggregate across raters)
using the Spearman–Brown prophecy formula or Cronbach’s
alpha (when more than two observer/raters contribute Q-sort
descriptions) by treating the item-distribution provided by
single raters as “items” and Q-items as “subjects” in a correlation matrix. When the raw correlations between Q-sorters are
reported as indices of rater agreement, they are typically
referred to as “Q-correlations” (see below), but it is the
reliability of the composite Q-description that is typically of
interest to investigators.
Q-sort data lend themselves to a variety of scoring methods,
69
of which, the most common in developmental research is the
“criterion score” (Block, 1961; E. Waters & Deane, 1985). To
construct a criterion Q-sort, experts with regard to some
relevant dimension (e.g., attachment security in the case of the
AQS) are asked to use the Q-set to describe an individual at a
hypothetical extreme with respect to the dimension (e.g., the
most secure preschool child). After checking Q-correlations
among respondents and ascertaining that the average across
experts has high internal consistency (alpha > .85), the
composite (average item placement for each item in the distribution) becomes the criterion for the dimension. Q-sort
descriptions for empirical cases (e.g., the young children in the
samples reported here), are compared with the hypothetical
criterion and the congruence between the two is calculated as
the Pearson’s product-moment correlation. This is usually
treated as the “score” for the hypothetical construct and can
be analyzed in relation to other constructs or in tests on group
means. When the Q-sort construct of interest (here, attachment security) is found to be associated in theoretically
predicted ways with constructs external to the Q-sort, it is a
common practice to examine the individual Q-item correlates
of that external construct (Block, 1961; Buss, Block, & Block,
1980; Vaughn, Block, & Block, 1988; Vaughn & Martino,
1988) to more descriptively characterize the nature of the
relation between the Q-construct(s) and the external dimension or categorical variable. This practice is analogous to (but
not the equivalent of ) examining factor loadings for scale
items loading on some dimension of temperament or personality and is usually initiated from within the context of discovery rather than the context of justification. The results of such
descriptive analyses are usually heuristic and not considered
prescriptive; that is to say, they do not constitute hypothesis
tests even though the results of significance tests are usually
reported.
AQS observations. Because Colombian infants were about 1
year old, two home visits (total of 4 hours) were completed.
Two observers went on each visit and both independently
described the infant using the AQS. In the Portuguese and US
samples, a 2–3-hour home visit was scheduled for the purpose
of observing the child and mother together. In the Portuguese
sample, two observers went on all home visits and each independently described the child using the AQS after the visit. In
the US sample, one or two observers went on each visit and
the child was described using the AQS immediately afterwards.
In all samples, mothers were asked to go about their daily
routines in as normal a manner as possible while the observers
were present. The observers attempted to remain unobtrusive
for most of the home visit, but they were instructed to respond
naturally to any interactive bids from the child or from the
mother. If the child did not initiate an interaction with the
observer(s) during the first half of the visit, one observer
attempted to engage the child in a playful interaction. If the
child responded positively, the observer played briefly (usually
less than 10 min) with the child before withdrawing. From time
to time, the observers asked mothers about items they could
not observe (e.g., item 10 asks about child’s behavior when
getting ready for a nap; item 85 refers to child’s response to
novel toys or activities) and items they may not have observed
during the visit (e.g., item 26 refers to the child’s reaction to
being left at home with babysitter, father, or grandparent; item
47 refers to the child’s acceptance of loud sounds or being
bounced around in play if the mother smiles and shows that it
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
70
VAUGHN ET AL. / MATERNAL SECURE BASE SCRIPTS
is supposed to be fun). Observers were instructed to query the
mother in the last hour of the observation visit regarding such
items.
For the Colombian sample, visits were conducted by trained
observers who had obtained interobserver reliabilities of at
least .70 in five training tapes prior to going on home visits.
Interobserver reliability for the composite sort was calculated
from the agreement between the Q-sort descriptions and
ranged from .67 to .97. In the Portuguese sample, observers
were trained over a period of several weeks. Both observers
discussed items and jointly completed Q-sorts with the project
coordinator as part of training. Prior to actual data collection,
observers reached acceptable levels of agreement in independent Q-sorts (Q-correlations between .60 and .85). Crossobserver Q-correlations for the study sample ranged from r =
.61 to .89. Q-sort descriptions of the child were completed
within 24 hours of the visit. The Q-sort descriptions for a given
child were averaged across observers. For the US sample, rater
agreement was established (Q-correlations between the raters
between .70 and .80 for the full sort) prior to data collection.
Q-items for the study sample were sorted by consensus when
two observers made the home visit. A single observer
completed most of home visits and AQS descriptions (n = 38).
In each sample, the items were sorted into a rectangular
distribution (9 categories, 10 items in each category) with
items judged undescriptive or not characteristic of the child
sorted into categories 1–3 and items judged descriptive or
more characteristic of the child sorted into categories 7–9.
These Q-sort profiles were summarized using the E. Waters
Security Criterion sort for the AQS (see E. Waters, Vaughn,
Posada, & Kondo-Ikemura, 1995). These scores ranged from
–.31 to .79 (M = .41, SD = .22) for the cases in the combined
sample. Children in the US sample had the lowest average
scores (r = .35) and children in the Colombian sample had the
highest average scores (r = .49). The security score for the
Colombian sample is somewhat higher than reported previously in Colombian samples (i.e., .43 and .46 in Posada et al.,
2002; Posada, Carbonell, Alzate, & Plata, 2004) and the difference between Colombian and US means was not itself significant (z = 1.2, p > .25), suggesting that the samples could be
combined for our primary analyses. E. Waters (1995) also
provides a Dependency Criterion sort for the AQS and we
calculated this score as well. Dependency scores ranged from
–.55 to .49 across the full sample with means ranging from –.11
to –.03. Tests on the sample means did not reveal cross-sample
differences.
Results
The primary goal of this report is to examine the relation
between scriptedness in maternal secure-base stories and the
child’s secure-base behavior at home. By design, we are able to
assess this relation in separate samples from different sociocultural groups and, because the subsamples do not differ in
terms of mean levels for the outcome variables and are comparable with respect to social class indicators, we also present
correlation analyses for the combined sample. A secondary
goal of the study is to add data to the cross-cultural findings
relevant to the psychometric properties (rater agreement,
internal consistency) of the word-list prompt task and to
compare our findings with those reported by RodriguesDoolabh et al. (2003). Finally, the data afford opportunities to
explore relations between the maternal representations of
secure-base knowledge and details of the child’s secure-base
behavior that are available from the individual items in the
AQS. To maintain continuity in the narrative, we first present
results of potential demographic confounds (i.e., income,
maternal age, level of education), then the psychometric
analyses, and finally the substantive analyses relating securebase scriptedness to child secure-base behavior. These last
analyses are presented at the level of the summary score for
security and also at the level of individual Q-set items.
Preliminary analyses
Correlations were calculated for each sample to determine
whether maternal age, education level (three levels: high school
or less, some post-secondary education but without university
degree, university degree or higher), or family income level
were significantly associated with either the maternal securebase scriptedness score or with the child AQS security score.
Neither maternal age nor income level showed significant
associations with the security outcome variables in any of the
three samples. Correlation values for these demographic indicators ranged from –.13 to .15 across the three samples. For
the Portuguese and US subsamples, maternal educational
attainment was not associated with either security indicator,
however, in the Colombian sample, the rank-order correlation
was positive and significant, rSP = .57, p < .01. The rank-order
correlation between maternal education level and AQS security
was not significant, rSP = .26. Education level was retained as
a control variable in subsequent analyses.
Reliability analyses
Rodrigues-Doolabh et al. (2003) reported that cross-story
(within theme) correlations were quite high across a range of
cultural groups. Furthermore, the cross-theme scores (i.e.,
composites of mother–child ⫻ adult–adult themed narratives)
were also highly correlated in all cultural groups. The withinand cross-theme correlations, as well as Cronbach’s alpha for
the total composite (from four stories) values are presented in
Table 1 for each of the three sociocultural groups separately,
then again for the combined sample. Consistent with the
Rodrigues-Doolabh et al. (2003) report, within-theme correlations ranged from .45 to .70 and Cronbach’s alpha estimates
for the composite narrative score were all above .80 in each
sample. Our findings substantially replicate results reported
previously and suggest that the method will be widely applicable in samples of literate participants.
Maternal representations and child behavior
To test the hypothesis that maternal secure-base representations would covary with the organization of children’s securebase behavior, we correlated the maternal and child scores.
Results are presented in Table 2. The results are consistent with
our hypothesis. The composite maternal narrative score (i.e.,
average of the two mother–child secure-base stories) was positively and significantly correlated with the AQS security score
in each sample (range of rs = .35 to .50, ps < .05 in each
analysis) and in the full sample, r(129) = .37, p < .001.
Between group z-tests indicated that no pair of correlations
differed significantly. Because we had obtained a significant
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
71
INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2007, 31 (1), 65–76
Table 1
Internal consistency of narrative scores in the full sample and for each sub-sample separately
Colombia
Mother/child stories (n = 25)
Adult/adult stories
Secure-base composite scores (n = 25)
Portugal
Mother/child stories (n = 58)
Adult/adult stories (n = 58)
Secure-base composite scores
United States
Mother/child stories (n = 47)
Adult/adult stories (n = 47)
Secure-base composite scores
Total sample
Mother/child stories (N = 129)
Adult/adult stories (N = 105)
Secure-base composite scores (N = 105)
r = .70***
Not calculated
Spearman–Brown = .81
Spearman–Brown = .81a
r = .63***
r = .43***
Spearman–Brown = .76
Spearman–Brown = .62
α = .82
r = .63***
r = .59***
Spearman–Brown
Spearman–Brown
α = .86
Reliability
Spearman–Brown
Spearman–Brown
α = .83
Item correlations
r = .63***
r = .52***
= .78
= .74
= .77
= .68
a Reliability
estimate based only on scores for two mother–child stories in Colombian sample.
*** = p < .001.
association between maternal education level and the maternal
narrative scriptedness score, we computed the correlation
again controlling for education level. This did not appreciably
change the magnitude of association for the full sample, rpartial
= .38, and effects for the US and Portuguese samples were
similarly not affected. In the Colombian sample, the correlation was somewhat reduced, r = .43, but remained significant,
p < .05.
Additional analyses tested possible relations between the
narrative scriptedness score and child Dependency scores from
the AQS. The association between these variables did not reach
significance for the full sample, r(129) = .16, ns, and the
relation was not significant in either the US or the Portuguese
samples. However, in the Colombian sample, maternal narrative scriptedness was significantly associated with Dependency
from the AQS, r = .42, p < .05. Perhaps interestingly, the AQS
Security and Dependency scores from this sample were also
significantly associated for the Colombian sample, r = .56, p <
.01, but this pattern of association was not obtained for either
the US or the Portuguese samples, rs = .20 and –.02, ns, for
the US and Portuguese samples, respectively. A partial correlation analysis controlling for AQS Security reduced the
relation between AQS Dependency and the maternal narrative
scriptedness score to .19. It appears that the relation between
Table 2
Correlations between maternal narrative scores and child
secure-base behavior at home
Colombia
Mother/child
Portugal
Mother/child
United States
Mother/child
Total sample
Mother/child
stories (n = 25)
r = .50*
stories (n = 58)
r = .35**
stories (n = 47)
r = .38**
stories (N = 130)
r = .37***
Note. Attachment composite score uses only the mother/child stories
in Colombian sample.
*p < .05; **p < .01; ***p < .001.
AQS Dependency and the narrative scriptedness score is due
to their joint relation with AQS Security.
To consider the relation between scriptedness in maternal
representations of secure-base knowledge and their children’s
secure-base behavior in more detail, we examined the individual AQS item correlates of the narrative scores. For these
analyses, we combined the subsamples and correlated each of
the 90 AQS items with the composite derived from the two
mother–child stories. A total of 25 AQS items (over 27% of
the 90-item pool) proved to have a significant association with
the maternal scriptedness score (Table 3) and 20 of these items
were also significant correlates of the AQS Security criterion
score (marked with an asterisk in the AQS Security correlate
column of Table 3). Significant correlations were signed in the
same direction for the AQS security and maternal scriptedness
scores. Although these 25 item-correlates do not exhaust the
secure-base content of the AQS (67 Q-sort items were significantly correlated with the security criterion), they suggest that
the secure-base scriptedness score for the mother is, in fact,
attracting this content from the AQS. A mother who used the
secure-base script to shape her narratives tended to have
children who tended to pay attention and understand her, to
readily share with her, and to enjoy physical contact with her.
They tended to be less demanding or impatient, less likely to
act like the mother would interfere with their activities, less
likely to fuss or get angry when the mother did not do what
they wanted right away, and more likely to turn to her after
finishing with an activity or if in need of assistance. We also
identified five item correlates of the narrative scores that were
not also significant correlates of security. Mothers with higher
total narrative scores had children who tended to be better
coordinated physically and more curious with new toys, to
make an effort to keep track of mother’s location in the house
and to go to her when bored, and who tended to be more
positively oriented to unfamiliar adults. Although there is
diversity of content in these item correlates, our results suggest
that it is primarily child secure-base content in the AQS
that is predictable from the scriptedness of maternal securebase stories. That is, mothers with well-scripted secure-base
knowledge have children who treat them as a secure base for
exploration at home.
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
72
VAUGHN ET AL. / MATERNAL SECURE BASE SCRIPTS
Table 3
AQS item correlates of maternal mother–child scriptedness score
AQS item
AQS(18)
AQS(41)
AQS(32)
AQS(53)
AQS(21)
AQS(46)
AQS(19)
AQS(1)
AQS(43)
AQS(11)
AQS(83)
AQS(40)
AQS(37)
AQS(23)
AQS(2)
AQS(30)
AQS(52)
AQS(35)
AQS(69)
AQS(59)
AQS(54)
AQS(72)
AQS(79)
AQS(38)
AQS(74)
Child follows mother’s suggestions readily, even when they are clearly suggestions rather than
orders.
When mother says to follow her, child does so.
When mother says “no” or punishes him, child stops misbehaving. Doesn’t have to be told twice.
Child puts his arms around mother or puts his hand on her shoulder when she picks him up.
Child keeps track of mother’s location when he plays around the house.
Child walks and runs around without bumping, dropping, or stumbling.
When mother tells child to bring or give her something he obeys.
Child readily shares with mother or lets her hold things if she asks to
Child stays closer to mother or returns to her more often then the simple task of keeping track
of her requires
Child often hugs or cuddles against mother without her asking or inviting him.
When child is bored, he goes to mother looking for something to do
Child examines new objects or toys in great detail. Tries to use them in different ways or to take
them apart
Child is very active. Always moving around. Prefers active games to quiet ones
When mother sits with other family members, or is affectionate with them, child tries to get
mom’s affection for himself
When child returns to mother after playing, he is sometimes fussy for no clear reason
Child easily becomes angry with toys
Child has trouble handling small objects or putting small things together
Child is independent with mother. Prefers to play on his own, leaves mother easily when he
wants to play
Rarely asks mother for help
When child finishes with an activity or toy, he generally finds something else to do without
returning to mother
Child acts like he expects mother to interfere with his activities when she is simply trying to help
him with something
If visitors laugh at or approve of something the child does, he repeats it again and again.
Child is easily angry at mother
Child is demanding and impatient with mother. Fusses and persists unless she does what he
wants right away
When mother doesn’t do what child wants right away, he behaves as if mom were not going to
do it at all
Correlation with
scriptedness
AQS Security
correlate
.28***
.28***
.24**
.23**
.22*
.22*
.20*
.20*
*
*
*
*
*
.19*
.17*
.17*
*
*
*
–.17*
–.17*
*
–.18*
–.18*
–.19*
–.21*
*
*
*
*
–.21*
–.21*
*
*
–.22*
*
–.24**
–.24**
–.25**
*
–.25**
*
–.25**
*
*
* p < .05; **p < .01; ***p < .001.
Discussion
At the outset of this article, we emphasized the centrality of
the secure-base concept in the Bowlby/Ainsworth theory of
attachment. When the caregiving member of an attached dyad
provides support for the attached member’s exploration of the
local and distant physical, cognitive, and social environments,
by monitoring, assisting, or intervening on the behalf of the
other, he or she is acting as a secure base. When the caregiving member welcomes, nurtures, or comforts that attached
member upon return to proximity after exploration or other
circumstance involving separation, she or he is acting as a
secure base. When the attached member knows that the caregiving member of the dyad is able and willing to provide this
support for exploration and succor upon reunion, he or she can
more confidently maintain engagement during exploration and
will be more enthusiastic at the prospect of reunion. For young
children, it is this apparent confidence in exploration and the
satisfaction of the child at reunion with the attachment figure
that supports the inference that she or he believes that the
parent is a secure base and that the child is, indeed, secure in
her or his attachment with the caregiver. The nature of secure-
base support and the manner in which such support is
provided, as well as the child’s expressed need for support, may
fluctuate over developmental periods but the secure-base
phenomenon is a feature of attachment relationships at every
age (Sroufe et al., 2005). So critical was this idea, that Bowlby
titled his last volume of published essays on attachment, “A
Secure Base”.
The second critical assumption of attachment theory that
we highlighted was the notion that the attached individual coconstructs, with the assistance of the attachment figure, an
internal working model of the attachment relationship itself,
as well as correlated models of the self and of the caregiver.
Because the secure-base phenomenon figures so prominently
in the formation and maintenance of attachment relationships,
we expect to find information about having a secure base and
(possibly) being a secure base for another embedded in the
working models co-constructed by the attachment dyad
(Bowlby, 1973, 1982, 1988), at least for those individuals with
secure attachment histories (see Oppenheim & H. Waters,
1995; E. Waters & Cummings, 2000 for related discussions).
Although some of the more widely used protocols for assessing attachment beyond infancy and childhood do not explic-
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2007, 31 (1), 65–76
itly dimensionalize secure-base information (George & West,
2001; Main & Goldwyn, 1984, 1998), these protocols imply
the presence of secure-base phenomena when they classify
cases as secure versus insecure. By way of contrast, the H.
Waters and Rodrigues-Doolabh (2004) word-list prompt
measure was designed explicitly to assess secure-base knowledge and the individual’s access/use of that knowledge, when
confronted with a story-telling task. H. Waters and RodriguesDoolabh (2004) assumed that internal working models of
secure-base knowledge would have a formal structure similar
to other domains of knowledge built on repeated experiences
in a circumscribed set of parameters; that is to say, they
believed that this knowledge would be abstracted and generalized in the form of a secure-base script. Their measure and
scoring procedures provide means to quantify individual differences regarding the presence and use of secure-base scripts
adults in the construction of stories about routine and emergency events for attached dyads.
H. Waters and associates have examined both the generality
and the implications of secure-base scripts for adults. In several
studies (Rodrigues-Doolabh et al., 2003), findings suggest that
scripted knowledge of secure-base relationships are found in
wide range of societies, many of which differ markedly from
North American society. Tini et al. (2003) provided important
data regarding the convergent validity of the script measure by
demonstrating a significant relation between maternal securebase script scores and infant classifications n the Strange Situation. We organized the data from the three samples included
here to extend findings reported by Rodrigues-Doolabh et al.
(2003) and Tini et al. (2003) by adding new cultural groups
(from Colombia and from Portugal) to the list of countries in
which secure-base scriptedness was present and by showing
that an assessment of child attachment security other than the
Strange Situation is associated with maternal secure-base
scriptedness scores.
With respect to extending the range of sociocultural groups
for which reliable assessments could be obtained, we found
that correlations within and across types of stories were similar
to those reported by Rodrigues-Doolabh et al. (2003) and that
the reliability of a composite made up of both mother–child
and adult–adult stories was satisfactory (α > .8) in the US and
Portuguese samples. The Spearman–Brown prophecy formula
suggests that reliability of the composite will also be above .8
in Colombian samples. These findings suggest that the wordprompt story method of obtaining secure-base narratives will
be useful in cross-cultural studies of the secure-base phenomenon and secure-base relationships. However, we caution
potential consumers for this measure that the current set of
word-prompt lists does not necessarily constitute a universal
kit that can be applied in any sociocultural group. As
Rodrigues-Doolabh et al. (2003) reported, if a particular story
theme does not apply in the cultural context (as the camping
trip story did not in our Colombian sample), for a sociocultural group, it will be necessary to design a word prompt list that
is relevant to the particular culture. This sort of solution was
applied successfully in the United Arab Emirates sample
reported by Rodrigues-Doolabh et al. (2003). We also
discovered that considerable pilot testing was required in
presenting the task to mothers. In the Colombian sample, the
translated instructions (even though they had been back-translated to be equivalent to the English text) apparently prompted
mothers to tell stories focused on children rather than on
adults in the adult–adult themed word-prompt lists. Instruc-
73
tions for the adult–adult word-prompt lists now include an
explicit statement that the adults in the story are a couple
(married or romantic partners) and that the story is supposed
to be about them. In our ongoing research using the word-list
prompt measure, this small adjustment to instructions has all
but eliminated adult-themed stories including children.
A second caution concerns the typical length of stories in a
given sociocultural group. Although secure-base scriptedness
scores are not necessarily a function of story length, this is both
because some very long stories can be bereft of secure-base
content and because some relatively short stories can be scored
at “4” or above. For example, in the US sample, the correlation between story length and scriptedness was r = .21, ns, but
adding a story length variable in a regression changed neither
the beta value for scriptedness nor the total R2 for the AQS
security score. Nevertheless, it is generally true that shorter
stories (i.e., < 75 words) will receive scores of “3” or less
because they do not establish the relationship between story
protagonists and do not provide any relevant detail about the
meaning of events in the story or the relevance of the story
events to the protagonists. Likewise, stories receiving the
highest secure-base scriptedness scores provide considerable
detail about the relationship between dyad members and
demonstrate that the caregiving partner understands the
emotional and motivational states of the attached partner. To
provide this detail, these stories will necessarily be relatively
long (i.e., > 250 words).
We have found that presenting the word-prompt lists as
outlines for a story and asking the respondent to tell a “full”
or “complete” story based on the words in the outline is generally sufficient to prompt a range of story lengths (resulting in
a full range of scores along the 7-point scale developed by H.
Waters & Rodrigues-Doolabh, 2004). Nevertheless, it will be
important for any research team to pilot test the instructions
and review stories after 10–15 pilot participants have been
tested to assure that both longer and shorter stories are represented. If 80% or more of the stories are brief (i.e., < 75
words), it will not be possible to get the expected range of
scores for a given sample and some revision to the instructions
will be required for that particular sociocultural group. Revised
instructions should also be pilot tested until it is clear that
respondents are providing the full range of story lengths and
scores. While it is anticipated that some (< 25%) of stories in
a sample of 25 or more adults will be short (and generally will
receive scores lower than 4, indicating that the secure-base
script was not present in the story), we do not expect to find
more than 50% of a nonclinical sample with insecure attachment representations. Thus, pilot testing to establish the best
instruction and the most appropriate interpersonal technique
for creating an environment conducive to mothers producing
their best stories will be beneficial for most research investigations.
It is also important to note that all respondents evaluated to
date have been literate (even well educated) and could comprehend the word-list prompts easily. It may be that the low correlations between scriptedness scores and education level that we
report here arise because we recruited primarily middle-class
participants with relatively high education levels to these
studies. Recall also that in the Colombian sample, the group
with the greatest heterogeneity for education levels, mothers
with a university degree had higher scriptedness scores than
did those less education. This did not, however, severely attenuate the relation between scriptedness and AQS Security
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
74
VAUGHN ET AL. / MATERNAL SECURE BASE SCRIPTS
scores for the Colombian sample. Finally, it is always preferable to elicit narratives in the respondent’s native language.
Pilot testing using foreign visitors/immigrants in the US
(Posada, 2004, unpublished data) suggested that some ESL
participants, had difficulties providing stories that could be
scored using the H. Waters and Rodrigues-Doolabh (2004)
coding rules.
With respect to the major goal for the study, we found that
the secure-base scriptedness scores derived from the maternal
narratives were positively and significantly associated with the
child security scores from the AQS. This was true in each of
the three sociocultural groups and the finding was reproduced
in the full sample. The fact that we obtained predicted associations across measures holds in the face of sample differences
(different cultures, ethnic/racial makeup of groups, education
levels) suggests that the basic finding is robust. We interpret
our results as support for important premises arising from the
Bowlby/Ainsworth attachment theory regarding relations
between maternal internal working models and child attachment behavior. Although there was a range of correlations
across samples (rs = .35–.50), no correlation value differed
significantly from any other, when tested across samples. Our
findings are comparable with results of other studies linking
maternal AAI security classifications to AQS security scores
(Posada, Waters, et al., 1995; Tarabulsy et al., 2005). The
results suggest that the word-prompt narrative procedure has
utility for assessing script-like representations of secure-base
knowledge for adults and that individual differences in scriptedness scores map onto children’s secure-base behavior in a
predictable manner. We also found that, in general, the
maternal scriptedness scores were unrelated to the Dependency dimension scored from the AQS. At the level of individual subsamples, nonsignificant associations were found in the
two sociocultural samples for which the AQS had been
completed at the end of toddlerhood, however, for the sample
observed as infants (i.e., Colombia), Dependency was significantly associated with both Security and with maternal scriptedness scores. Such a finding is consistent with a discussion of
security and dependency by E. Waters and Deane (1985), who
noted that these two constructs were both conceptually and
empirically related for infants but not for preschool age
children (see also, Sroufe et al., 2005 for a related discussion).
We also examined the individual AQS item correlates of
maternal scriptedness scores derived from the maternal narratives so as to describe the relation between child secure-base
behavior and maternal secure-base scriptedness in greater
detail. We found 25 of the 90 AQS items to be significantly
associated with maternal narrative scores and 20 of these items
were also significant correlates of AQS security. The positive
AQS item correlates (e.g., willing to share, hugs and cuddles
with mother, follows suggestions readily, obeys when told to
bring something) suggest a warm and mutually satisfying
relationship between the child and mother. Similarly, the
negative AQS correlates (e.g., child is demanding or impatient
with mother, acts like he expects mother to interfere, is easily
angered by mother) suggest a relatively more conflicted
mother–child relationship. These item-level correlates suggest
that mothers who have ready access to a secure-base script
interact with their children in a manner that fosters a secure
relationship. Although finding these child-relevant correlates of
the maternal scriptedness score is consistent with our broader
theoretical position regarding secure-base relationships, we
caution that our correlation data are intended to be broadly
representative of the sorts of relations expectable for these
measures and we do not mean to imply that these items necessarily constitute “the” secure-base correlates of maternal
secure-base scriptedness. Studies with larger sample sizes and
with observations made across more occasions by more raters
will likely find additional significant associations between the
Q-sort items and maternal scriptedness.
Our data also make a modest methodological point. From
time to time, we have received (negative) feedback about
problems that arise from etic approaches to understanding
people in sociocultural groups different from the investigators.
It is not our intention to enter into the debate about the relative
value of emic versus etic approaches here; nevertheless, our
data suggest that the measures used in this study transcend (at
least some) sociocultural boundaries. In all three sociocultural
groups we find a significant association between mothers’
secure-base scriptedness and child secure-base behavior.
Furthermore, our rater agreement data suggest that one need
not be a native speaker of English to accurately score narratives produced by native English speakers. In addition, and
importantly, narratives elicited from non-English speakers
(Portuguese and Spanish, in our case) can receive valid
interpretations and produce reliable and valid scores when
translated into English and scored by monolingual English
speakers. Portuguese and Italian native speakers were able to
decode and score English language narratives with a high
degree of fidelity to an English speaker’s scores. The English
speaker achieved a substantial degree of agreement (from
translated protocols) with Portuguese protocols scored from
the native language transcripts. Finally, stories translated from
Spanish to English were scored by two observers (blind to all
other study data), who were not Spanish speakers, and these
scores were significantly correlated with AQS scores that had
been obtained two or more years prior to the narrative assessments. These findings should encourage researchers in nonEnglish speaking countries that data gathered using the etic
procedure can produce valid data about secure-base scriptedness, even if it is necessary to translate the original narratives
into English for expert scoring. We are presently preparing a
multinational study to test this speculative hypothesis.
We believe that our results constitute evidence for the crosscultural validity of the Bowlby/Ainsworth attachment theory
and of the utility of the AQS and the word-list prompt assessments. Of course, identifying a theoretically consistent link
between maternal secure-base representation and child securebase behavior in different sociocultural groups does not mean
that the behavioral mechanism underlying that link has been
specified. Just what it is that mothers do (and how they do it)
with their children, that leads to the convergence between
maternal representation and child behavior remains unexplored here. Traditionally, this link has been identified as
“maternal sensitivity to child communicative signals”
(Ainsworth et al., 1978) but evidence for sensitivity as a mediating mechanism is equivocal, with some investigators failing
to find evidence for mediation (Atkinson et al., 2005) and
others reporting that sensitivity mediates between maternal
state of mind and attachment security, but only when controlling for relevant demographic and personality variables
(Tarabulsy et al., 2005). It is possible that mothers in different
sociocultural groups, who endorse different child-rearing
values, goals, and practices may link maternal representations
to child behavior using different transactional means. To test
for such possibilities, it will be necessary to study secure-base
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
INTERNATIONAL JOURNAL OF BEHAVIORAL DEVELOPMENT, 2007, 31 (1), 65–76
phenomena across several different cultural groups and to
include comprehensive assessments of parent–child interaction
and parenting values, in addition to measures of secure-base
organization. Having demonstrated the possibilities of the
word-prompt and Q-sort measures here, we are more confident to initiate larger and more comprehensive studies that will
include measures of possible cultural differences that might
inform our understandings of both similarities and differences
across groups from different sociocultural milieu.
References
Ainsworth, M.D.S. (1967). Infancy in Uganda: Infant care and the growth of love.
Baltimore, MD: Johns Hopkins University Press.
Ainsworth, M.D.S., Bell, S.M., & Stayton, D.J. (1974). Infant–mother attachment and social development: Socialization as a product of reciprocal responsiveness to signals. In M.J.M. Richards (Ed.), The integration of a child into a
social world (pp. 99–135). London: Cambridge University Press.
Ainsworth, M.D.S., Blehar, M.C., Waters, E., & Wall, S. (1978). Patterns of
attachment: A psychological study of the Strange Situation. Hillsdale, NJ:
Erlbaum.
Ainsworth, M.D.S., & Bowlby, J. (1991). An ethological approach to personality development. American Psychologist, 46, 333–341.
Atkinson, L., Goldberg, S., Raval, V., Pederson, D., Benoit, D., Moran, G., et
al. (2005). On the relation between maternal state of mind and sensitivity in
the prediction of infant attachment security. Developmental Psychology, 41,
42–53.
Bowlby, J. (1973). Attachment and loss:Vol. 2. Separation, anxiety, and anger. New
York: Basic Books.
Bowlby, J. (1982). Attachment and loss:Vol. 1. Attachment (2nd ed.). New York:
Basic Books.
Bowlby, J. (1988). A secure-base: Clinical applications of attachment theory. London:
Tavistock.
Bretherton, I. (1991). Pouring new wine into old bottles: The social self as
internal working model. In M. Gunnar & L.A. Sroufe (Eds.), Minnesota
symposia on child psychology: Vol. 23. Self-processes in development (pp. 1–41).
Hillsdale, NJ: Erlbaum.
Bretherton, I., & Munholland, K.A. (1999). Internal working models in attachment relationships: A construct revisited. In J. Cassidy & P.R. Shaver (Eds.),
Handbook of attachment:Theory, research, and clinical applications (pp. 89–111).
New York: Guilford Press.
Bretherton, I., Ridgeway, D., & Cassidy, J. (1990). Assessing internal working
models of the attachment relationship: An attachment story-completion task
for 3-year-olds. In M. Greenberg, D. Cicchetti, & E.M. Cummings (Eds.),
Attachment during the preschool years: Theory, research, and intervention
(pp. 272–308). Chicago: University of Chicago Press.
Buss, D.M., Block, J.H., & Block, J. (1980). Preschool activity level: Personality correlates and developmental implications. Child Development, 51,
401–408.
Carlson, E.A., Sroufe, L.A., & Egeland, B. (2004). The construction of experience: A longitudinal study of representation and behavior. Child Development,
75, 66–83.
Crowell, J.A., Treboux, D., Gao, Y., Pan, H., Fyffe, C., & Waters, E. (2002).
Secure-base behavior in adulthood: Measurement, links to adult attachment
representations, and relations to couples’ communications skills and selfreports. Developmental Psychology, 38, 679–702.
De Wolff, M.S., & van IJzendoorn, M.H. (1997). Sensitivity and attachment: A
meta-analysis on parental antecedents of infant attachment. Child Development, 68, 571–591.
Fivush, R., & Hudson, J.A. (Eds.). (1990). Knowing and remembering in young
children. Cambridge: Cambridge University Press.
George C., & West, M. (2001). The development and preliminary validation of
a new measure of adult attachment: The Adult Attachment Projective. Attachment and Human Development, 3, 30–61.
Guttmann-Steinmetz, S., Elliot, M., Steiner, M.C., & Waters, H.S. (2003). Coconstructing script-like representations of early secure base experience. In H.
Waters & E. Waters (Chairs), Script-like representations of secure base experience:
Evidence of cross-age, cross-cultural, and behavioral links. Poster symposium
presented at the Biennial Meetings of the Society for Research in Child
Development, Tampa, FL.
Harwood, R.L., & Miller, J.G. (1991). Perceptions of attachment behavior: A
comparison of Anglo and Puerto-Rican mothers. Merrill-Palmer Quarterly –
Journal of Developmental Psychology, 37, 583–599.
75
Harwood, R.L., Miller, J.G., & Irizarry, N.L. (1995). Culture and attachment:
Perceptions of the child in context. Culture and human development. New York:
Guilford Press.
Hesse, E. (1999). The Adult Attachment Interview: Historical and current
perspectives. In J. Cassidy & P.R. Shaver (Eds.), Handbook of attachment:
Theory, research, and clinical implications (pp. 395–433). New York: Guilford
Press.
Karavasilis, L., Doyle, A., & Markiewicz, D. (2003). Associations between
parenting style and attachment orientation in middle childhood and early
adolescence. International Journal of Behavioral Development, 27, 153–164.
Main, M., & Goldwyn, R. (1984). Adult attachment scoring and classification
system. Unpublished manuscript, University of California at Berkeley.
Main, M., & Goldwyn, R. (1998). Adult attachment scoring and classification
system. Unpublished manuscript, University of California at Berkeley.
Nelson, K. (1986). Event knowledge: Structure and function in development. Hillsdale, NJ: Erlbaum.
Nelson, K., & Hudson, J. (1988). Scripts and memory: Functional relationships
in development. In F.E. Weinert & M. Perlmutter (Eds.), Memory development:
Universal changes and individual differences (pp. 87–105). Hillsdale, NJ:
Erlbaum.
Oppenheim, D., & Waters, H.S. (1995). Narrative processes and attachment
representations: Issues of development and assessment. In E. Waters, B.E.
Vaughn, G. Posada, & K. Kondo-Ikemura (Eds.), Caregiving, cultural, and
cognitive perspectives on secure-base behavior and working models: New
growing points of attachment theory and research. Monographs of the Society
for Research in Child Development, 60(2–3), Serial No. 244.
Oyserman, D., Coon, H.M., & Kemmelmeier, M. (2002). Rethinking individualism and collectivism: Evaluation of theoretical assumptions and metaanalyses. Psychological Bulletin, 128, 3–72.
Park, K.A., & Waters, E. (1989). Security of attachment and preschool friendships. Child Development, 60, 1076–1081.
Pederson, D.R., Gleason, K.E., Moran, G., & Bento, S. (1998). Maternal attachment representations, maternal sensitivity, and the infant–mother attachment
relationship. Developmental Psychology, 34, 925–933.
Posada, G., Carbonell, O.A., Alzate, G., & Plata, S.J. (2004). Through Colombian lenses: Ethnographic and conventional analyses of maternal care and
their associations with secure-base behavior. Developmental Psychology, 40,
508–518.
Posada, G., Gao, Y., Wu, F., Posada, R., et al., (1995). The secure-base phenomenon across cultures: Children’s behavior, mothers’ preferences, and experts’
concepts. In E. Waters, B.E. Vaughn, G. Posada, & K. Kondo-Ikemura (Eds.),
Caregiving, cultural, and cognitive perspectives on secure-base behavior and
working models: New growing points of attachment theory and research,
Monographs of the Society for Research in Child Development, 60(2–3, Serial No.
244).
Posada, G., Jacobs, A., Carbonell, O.A., Alzate, G., Bustamante, M.R., &
Arenas, A. (1999). Maternal care and attachment security in ordinary and
emergency contexts. Developmental Psychology, 35, 1379–1388.
Posada, G., Jacobs, A., Richmond, M., Carbonell, O.A., Alzate, G., Bustamante,
M.R., & Quiceno, J. (2002). Maternal care giving and infant security in two
cultures. Developmental Psychology, 38, 67–78.
Posada, G., Waters, E., Crowell, J.A., & Lay, K-L. (1995). Is it easier to use a
secure mother as a secure base? Attachment Q-sort correlates of the adult
attachment interview. In E. Waters, B.E. Vaughn, G. Posada, & K. KondoIkemura (Eds.), Caregiving, cultural, and cognitive perspectives on securebase behavior and working models: New growing points of attachment theory
and research, Monographs of the Society for Research n Child Development,
60(2–3, Serial No. 244).
Rodrigues-Doolabh, L., Zevallos, A., Turan, B., & Green, K. (2003). Attachment scripts across cultures: Further evidence for a universal secure-base
script. In H. Waters & E. Waters (Chairs), Script-like representations of secure
base experience: Evidence of cross-age, cross-cultural, and behavioral links. Poster
symposium presented at the Biennial Meetings of the Society for Research in
Child Development, Tampa, FL, March.
Rothbaum, F., Pott, M., Azuma, H., Miyake, K., & Weisz, J. (2000). Trade-offs
in the study of culture and development: Theories, methods, and values. Child
Development, 71, 1159–1161.
Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005). The development of the person:The Minnesota study of risk and adaptation from birth to adulthood. New York: Guilford Press.
Tarabulsy, G.M., Bernier, A., Provost, M.A., Maranda, H., Larose, S., Moss,
E., Larose, M., & Tessier, R. (2005). Another look inside the gap: Ecological
contributions to the transmission or attachment in a sample of adolescent
mother–infant dyads. Developmental Psychology, 41, 212–224.
Tini, M., Corcoran, D., Rodrigues-Doolabh, L., &. Waters, E. (2003). Maternal
attachment scripts and infant secure-base behavior. In H. Waters & E. Waters
(Chairs), Script-like representations of secure base experience: Evidence of cross-age,
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.
76
VAUGHN ET AL. / MATERNAL SECURE BASE SCRIPTS
cross-cultural, and behavioral links. Poster symposium presented at the Biennial
Meetings of the Society for Research in Child Development, Tampa, FL.
van IJzendoorn, M.H., & Kroonenberg, P.M. (1988). Cross-cultural patterns of
attachment: A meta-analysis of the strange situation. Child Development, 59,
147–156.
van IJzendoorn, M.H., & Sagi, A. (1999). Cross-cultural patterns of attachment:
Universal and contextual dimension. In J.A. Cassidy & P. Shaver (Eds.),
Handbook of attachment: Theory, research, and clinical implications
(pp. 713–734). New York: Guilford Press.
van IJzendoorn, M.H., Vereijken, C.M.J.L., Bakermans-Kranenburg, M.J., &
Riksen-Walraven, J.M. (2004). Assessing attachment security with the attachment Q-sort: Meta-analytic evidence for the validity of the observer AQS.
Child Development, 75, 1188–1213.
Vaughn, B.E., Block, J.H., & Block, J. (1988). Parental agreement on childrearing during early childhood and the psychological characteristics of
adolescents. Child Development, 59, 1020–1033.
Vaughn, B.E., & Martino, D. (1988). Age related Q-sort correlates of visual
regard in groups of preschool children. Developmental Psychology, 24,
589–594.
Vaughn, B.E., & Waters, E. (1990). Attachment behavior at home and in the
laboratory: Q-sort observations and strange situation classifications of oneyear-olds. Child Development, 61, 1965–1973.
Ward, M.J., & Carlson, E. (1995). Associations among adult-attachment representations, maternal sensitivity, and infant–mother attachment in a sample of
adolescent mothers. Child Development, 69, 69–79.
Waters, E. (1995). The attachment behavior Q-set (Version 3.0). In E. Waters,
B.E. Vaughn, G. Posada, & K. Kodo-Ikemura (Eds.), Caregiving, cultural,
and cognitive perspectives on secure-base behavior and working models: New
growing points of attachment theory and research. Monographs of the Society
for Research in Child Development, 60 (Serial No. 244).
Waters, E., & Cummings, E.M. (2000). A secure base from which to explore
close relationships. Child Development, 71, 164–172.
Waters, E., & Deane, K. (1985). Defining and assessing individual differences
in attachment relationships: Q-methodology and the organization of behavior
in infancy and early childhood. In I. Bretherton & E. Waters (Eds.), Monographs of the Society for Research in Child Development, 50.
Waters, E., Vaughn, B.E., Posada, G., & Kondo-Ikemura, K. (1995). Caregiving, cultural, and cognitive perspectives on secure-base behavior and working
models: New growing points of attachment research. Monographs of the Society
for Research in Child Development, 60(Serial No. 244).
Waters, H.S., & Rodrigues-Doolabh, L. (2001). Are attachment scripts the
building blocks of attachment representations? Narrative assessment of representations and the AAI. In H.Waters & E.Waters (Chairs), Narrative Measures
of Attachment for Adults. Poster symposium presented at the Biennial Meetings
of the Society for Research in Child Development, Minneapolis, MN.
Waters, H.S., Rodrigues, L.M., & Ridgeway, D. (1998). Cognitive underpinnings of narrative attachment assessment. Journal of Experimental Child
Psychology, 71, 211–234.
Waters, H.S., & Rodrigues-Doolabh, L. (2004). Manual for decoding secure base
narratives. Unpublished manuscript, State University of New York at Stony
Brook.
Zimmermann, P. (2004). Attachment representations and characteristics of
friendship relations during adolescence. Journal of Experimental Child Psychology, 88, 83–101.
Downloaded from http://jbd.sagepub.com at AUBURN UNIV on February 1, 2007
© 2007 International Society for the Study of Behavioral Development. All rights reserved. Not for commercial use or unauthorized distribution.