Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

M. Fuertes et al.

: Infant Co ping, Still-Face,


European Maternal Behavior,
Psychologist
©2009; and14(4):320–331
2009 Vol. Attachment
Hogrefe Publishing

Infant Coping and Maternal


Interactive Behavior Predict
Attachment in a Portuguese Sample
of Healthy Preterm Infants
Marina Fuertes1,2, Pedro Lopes-dos-Santos2, Marjorie Beeghly1,3, and Ed Tronick1,4
1
Child Development Unit & Harvard Medical School, Boston, MA, USA, 2University of Porto, Porto,
Portugal, 3Wayne State University, Detroit, MI, USA, 4Department of Psychology, University of
Massachusetts Boston & Fielding Graduate University, Santa Barbara, CA, USA

Abstract. In this longitudinal study of a Portuguese sample of healthy preterm infants, the aim was to identify specific, independent
predictors of infant-mother attachment status from a set of variables including maternal education, maternal representations’ of infant
temperament, infant regulatory behavior (coping), and mothers’ interactive behavior in free play. The sample consisted of 48 medically
low-risk preterm infants and their mothers who varied in education. When infants were 1 and 3 months (corrected age), mothers described
their infants’ temperament using a Portuguese temperament scale (Escala de Temperamento do Bebé). At 3 months (corrected age),
infants’ capacity to regulate stress (coping) was evaluated during Tronick’s Face-to-Face Still-Face paradigm (FFSF). At 9 months
(corrected age), mothers’ interactive behaviors were evaluated during free play using the Crittenden’s Child-Adult Relationship Experi-
mental Index (CARE-Index). At 12 months (corrected age), infants’ attachment security was assessed during Ainsworth’s strange situ-
ation. Sixteen (33.3%) infants were classified as securely attached, 17 (35.4%) as insecure-avoidant, and 15 (31.3%) as insecure-resistant.
In bivariate analyses, multiple factors were significantly associated with attachment status. However, in hierarchical regression analyses,
only infant coping and maternal responsiveness were significant predictors of attachment status. These findings suggest that both infant
characteristics identifiable early in the first year, such as coping, and maternal characteristics such as sensitivity influence the process of
attachment formation.

Keywords: infant coping, mother-infant interaction, attachment

Introduction chologically withdrawn (Barnard & Kelly, 1990; Goldberg


& DiVitto, 1995; Muller-Nix et al., 2004). Given that pre-
mature infants are more difficult to handle, we wondered
Compared with the large number of studies on mother-in- whether style of maternal interactive behavior with the in-
fant attachment in the USA, relatively few attachment stud- fant, infants’ ability to regulate stress, and demographic
ies have been carried out with European samples, especial- factors predict attachment variation in this understudied
ly in Mediterranean countries. Moreover, many European Portuguese sample.
attachment studies have not evaluated the effect of associ-
ated risk factors such as premature birth or low levels of
maternal education on attachment formation (van IJzen-
doorn, Goldberg, Kroonenberg, & Frenkel, 1992; van Antecedents of Mother-Infant Attachment
IJzendoorn & Sagi, 1999). In the present longitudinal
study, we addressed this gap in the literature by evaluating Although multiple factors assessed during the infant’s first
specific, independent predictors of attachment in a medi- year of life are linked with infants’ later attachment status,
cally low-risk sample of prematurely born Portuguese in- the specific infant, maternal, dyadic, and contextual vari-
fants and their mothers with varying levels of education. ables that independently predict attachment status remain
Previous research has indicated that preterm infants are less controversial. Classical attachment theorists have postulat-
alert, less attentive, less active, and less responsive than ed that mothers’ sensitivity and warm responsivity to their
infants born at full term (Crnic, Ragozin, Greenberg, Rob- infants are primary factors determining the quality of the
inson, & Basham, 1983; Goldberg & DiVitto, 1995), and infant-mother attachment relationship (Ainsworth, Bell, &
their mothers tend to be more intrusive, controlling or psy- Stayton, 1974; Bowlby, 1969). In support of this hypothe-

European Psychologist 2009; Vol. 14(4):320–331 © 2009 Hogrefe Publishing


DOI 10.1027/1016-9040.14.4.320
M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment 321

sis, investigators have shown that infants who receive more These findings suggest that multiple factors contribute
positive and sensitive responses from their mothers during to early attachment formation. Unfortunately, the partic-
the first year of life are more likely to be classified as se- ular set of variables measured in prior investigations on
curely attached by the end of the first year, compared to attachment antecedents varies from study to study, and
infants receiving less sensitive care (Ainsworth, Blehar, the specific predictors of attachment quality remain un-
Waters, & Wall, 1978; Grossmann, Grossmann, Spangler, determined. Further longitudinal research investigating
Suess, & Unzner, 1985; Isabella, 1993). multiple predictors of attachment is needed, especially in
Others have argued that individual differences in infant understudied cross-cultural samples (van IJzendoorn &
characteristics such as difficult temperament and arousal reg- Sagi, 1999).
ulation underlie variations in infants’ attachment status (e.g.,
Kagan, 1984). Several investigators have shown that infants
who are better able to regulate their behavior and emotions The Present Study
when stressed (Braungart-Rieker, Garwood, Powers, &
Wang, 2001; Cohn, Campbell, & Ross, 1991), and infants The longitudinal analyses in the present study were carried
whose mothers make more positive and stable reports of their out on a Portuguese sample of medically low-risk prema-
temperament (Benoit, Parker, & Zeanah, 1997; Bretherton, turely born infants and their mothers from urban back-
Biringen, Ridgeway, Maslin, & Sherman, 1989; Cox, Hop- grounds who varied in years of completed education. The
kins, & Hans, 2000; Izard, Haynes, Chisholm, & Baak, 1991; goal was to identify specific, independent predictors of in-
Kochanska, 1998; Zeanah, Benoit, Hirshberg, Barton, & Re- fant attachment status from a set of infant, maternal, and
gan, 1994) are more likely to be classified as securely at- demographic variables measured at multiple time points
tached at the end of the first year, compared to infants whose during the infant’s first year of life.
mothers report poorer regulatory behavior in their infants. In Based on findings from prior research, we expected that
addition, individual differences in infants’ emotionality (e.g., several factors including: higher medical or social risk status
emotional responses such as joy, fear, and anger) are associ- (e.g., lower birth weight, lower maternal education), higher
ated with infants’ attachment status (e.g., Kochanska & Coy, maternal ratings of their infants’ difficult temperament, in-
2002). Nonetheless, neither maternal sensitivity (e.g., De fant’s poorer regulatory behavior during the stress of the
Wolff & van IJzendoorn, 1997) nor infant emotional reactiv- Face-to-Face Still-Face paradigm (FFSF), and lower mater-
ity alone adequately explains the findings on attachment for- nal sensitivity during dyadic play would be associated with a
mation (e.g., Goldsmith & Alansky, 1987). higher prevalence of insecure attachment in this sample. To
More recently, several researchers, including some at- identify independent predictors of attachment, hierarchical
tachment theorists, have argued that a transactional, multi- regression analyses were used. Although the literature pro-
contextual approach is needed for predicting attachment vides only limited guidance, our expectation was that both
quality (e.g., Belsky, 2005; Cicchetti & Barnett, 1991; Crit- maternal and infant factors would be independent predictors
tenden, 2000; Grossmann et al., 1985; Rutter, 1991; Sroufe, of attachment status in this Portuguese preterm sample.
1996, van den Boom, 1994; see also Cassidy & Shaver,
1999, for reviews). For instance, it was reported that both
the quality of maternal involvement with the infant and the
infant’s behavior, together with other contextual factors, Method
are significant predictors of attachment. In that study, infant
temperament was only indirectly associated with attach- Participants
ment via its direct associations with the other infant and
maternal variables (Belsky, 2005). Participants were 48 prematurely born infants (60.4%
Infant medical risk factors, such as premature birth, may male; 68.8% first born) and their adult, healthy mothers (M
also contribute to attachment quality, although specific find- age = 29.7 years; SD = 5.7; range: 18–42). Infant birth
ings vary depending on whether or not other risk factors were weight ranged from 1100 to 2500 grams (M = 2095, SD =
present in the sample under investigation. In samples of rel- .47) and their gestation ranged from 31 to 36 weeks (M =
atively healthy infants from well-educated middle-class sam- 33.9 weeks, SD = 1.73). Five minute Apgar scores ranged
ples, premature birth is not typically associated with a higher from 7 to 10 (M = 9.24; SD = .88). Mother-infant dyads
rate of attachment insecurity (e.g., Easterbrooks, 1989; Frodi were recruited in the Neonatal Unit Care of Matosinhos
& Thompson, 1985). However, when samples are character- Hospital, Porto, Portugal.
ized by additional risk factors such as low socioeconomic The participants were primarily Portuguese Caucasian
status (Wille, 1991), low maternal education (Pederson & from urban, middle-class socioeconomic backgrounds
Moran, 1996), comorbid infant health problems (Plunkett, (45.8% were lower middle-class, 18.8% were middle-class,
Meisels, Stiefel, Pasick, & Roloff, 1986), or maternal depres- and 35.4% were upper middle-class). All infants were
sion (Poehlmann & Fiese, 2001), premature birth is associat- healthy and clinically normal at delivery as determined by
ed with a higher prevalence of insecure attachment (see also pediatric examination. No infants had sensory or neuromo-
Belsky, 2005, for a review). tor disabilities, serious illnesses, or congenital anomalies.

© 2009 Hogrefe Publishing European Psychologist 2009; Vol. 14(4):320–331


322 M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment

No parents had any known mental health or drug/alcohol were 1 month old (corrected for gestational age) in a med-
addiction problems. All infants lived with both parents in ical room at the Centro de Medicina Desportiva do Porto
the same household (six couples were unmarried). (Sport Medicine Center of Porto). Each subsequent visit (at
Compared to other samples collected in Western countries, 3, 9 and 12 months of infant age, corrected for gestational
maternal formal education was quite low in our study. Nev- age) took place at the Developmental Psychology Labora-
ertheless, according to the latest available data from the na- tory of the Faculdade de Psicologia e Ciências da Educação
tional census (INE, 2004), values are consistent with indices da Universidade do Porto (i.e., the School of Psychology
for Portuguese women’s education: 20.8% completed up to 4 and Education of the University of Porto). At the 1- and
years (first grade of basic school), 16.7% completed 5 to 6 3-month visits, mothers reported on their infant’s temper-
years (second grade of basic school), 25.0% completed 7 to ament. At the 3-month visit, mother-infant dyads were also
9 years (third grade of basic school), 10.4% completed 10 to videotaped during the FFSF. At the 9-month visit, mother-
12 years (secondary school level), and 27.1% completed infant dyads were videotaped during a dyadic free play ses-
more than 12 years and attained a college/university degree sion, and at the 12-month visit, infants’ quality of attach-
(INE, 2004). ment was assessed using the standardized strange situation
In typical Western countries, maternal education and procedure (Ainsworth et al., 1978).
household income are highly positively correlated (e.g.,
Bugental & Happaney, 2004; Raikes & Thompson, 2005;
van IJzendoorn, Schuengel, & Bakermans-Kranenburg, Measures
1999). However, a unique feature of Portuguese society
is that maternal education and family income are not nec- Medical and Familial Demographic Information
essarily correlated. Indeed, despite their low educational
attainments, most mothers in the present study could be Medical and demographic measures included delivery
characterized as being middle-class. In fact, most moth- method, infant Apgar score, gestational age, birth weight,
ers with lower education had higher income because they sex, parity, family socioeconomic status (SES), mother’s
ran small businesses (e.g., shops) or held relatively well- age, and maternal years of completed education. To gain
paid jobs doing handicraft and skilled hand labor. The statistical power, the measure for maternal years of educa-
dissociation of educational level and income in the pre- tion was coded as a scale variable, ranging from 1 to 5
sent sample make it possible to examine the impact of points. Cases with more than 12 years of education (i.e.,
each variable separately on attachment, as well as wheth- mothers who were attaining or who had completed a col-
er their interaction affects attachment. lege or university degree) received the highest score of 5.
Cases with 10 to 12 years of education (secondary school
level) received a 4. Those who had from 7 to 8 years of
Procedure education (third grade of basic schooling) received a 3,
whereas cases with 5 to 6 years (second grade of basic
Over a 2-year period, a female research assistant contacted schooling were assigned a 2. Lastly, participants who had
potential participant mothers at the Neonatal Care Unit and only up to 4 years of school received the lowest score of 1.
explained the study’s purpose and procedures. To deter-
mine eligibility, mothers were administered a brief inter-
view to collect demographic information. With the moth- Maternal Report of Infant Temperament
er’s consent, information from the infant’s perinatal health
status was abstracted from the hospital medical record. Of At the 1- and 3-month visits, mothers described their in-
60 eligible mother-infant dyads, 54 agreed to participate in fants’ temperament using the Escala do Temperamento do
this longitudinal research and gave their informed consent. Bebé. Mothers were asked to rate nine aspects of their in-
Of these, three infants lost their eligibility for health rea- fant’s current behavior (e.g., “my baby calms down”) using
sons, two families moved away from the area, and one case a 7-point Likert scale (1 = very easy, 7 = very difficult). The
was dropped from the study because of procedural prob- Escala do Temperamento do Bebé was previously evaluat-
lems, resulting in a final sample of 48 mother-infant dyads ed in a sample of 127 healthy infants (60 preterms and 67
with complete data at all longitudinal visits. The study aims full-terms) at 1 and 3 months of age (Lopes dos Santos,
and procedures were approved by the Ethics Commission Fuertes, & Sanches-Ferreira, 2005). Items were factor an-
and People Rights Protection of the Portuguese Health Unit alyzed in each of both applications, yielding a single factor
of Matosinhos Hospital. (Difficult Behavior) that explained more than 60% of the
total variance (60.1% and 65.4%, respectively). Adequate
reliability estimates were found with Cronbach’s α coeffi-
Follow-Up Visit Procedures cients ranging from .922 to .933 and with a test-retest cor-
relation of r = .793. Validity of the Escala do Temperamen-
The 48 mother-infant dyads participated in four laboratory to do Bebé was also successfully checked against other
follow-up visits. The first visit took place when the infants Portuguese, validated measures of infants’ difficult behav-

European Psychologist 2009; Vol. 14(4):320–331 © 2009 Hogrefe Publishing


M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment 323

ior. Since the correlation between maternal ratings at the 1 mouths objects, closes his or her eyes, looks away from the
and 3-month visits was high (r = .766), for the purposes of mother, self touches, squeezes one hand against the other).
the present investigation, scores at each visit were averaged The score on each dimension was not a simple addition of the
to yield a single summary measure of infant difficultness. frequency of the behaviors, but was weighted to index to the
extent to which different behaviors in the dimension co-oc-
curred. One behavior in a dimension was scored as a 1, two
Infant Regulatory (Coping) Behavior behaviors performed together were scored as 3, and 3 or
more behaviors occurring together received 5 points (more
At the 3-month visit, infants’ capacity to regulate stress was than 3 behaviors was a rare situation). Therefore, the range in
evaluated during Tronick’s FFSF (Tronick, Als, Adamson, each second is 0 to 5 points. This weighted system allows
Wise, & Brazelton, 1978). Mother-infant dyads were vid- assessing the relative degree of infants’ responses in a given
eotaped during three successive 3-min episodes: (1) a nor- dimension. For example, simultaneously smiling, looking at
mal face-to-face interaction, in which the mother was in- mother, and trying to reach her is considered a more positive
structed to play with the infant as she typically would at reaction in the positive other-directed PODC dimension than
home, followed by (2) a 3-min still-face episode in which merely looking at mother. In the present study, PODC was
the mother was instructed to look at the baby while keeping not significantly correlated with NODC (Spearman ρ = .068;
a still (neutral) face and to refrain from talking, smiling, or ns) and SDC (Spearman ρ = –.038; ns). However, NODC and
touching the baby, followed by (3) a second 3-min normal SDC were significantly negatively correlated with each other
reunion interaction, in which the mother was instructed to (Spearman ρ = –.358; p < .05).
resume normal play with the infant. The laboratory room Reliability. To evaluate intercoder reliability, an inde-
was equipped with an infant seat mounted on a table, an pendent coder masked to background variables and the
adjustable chair for the mother, and two cameras (with mi- study’s hypotheses scored 15 randomly selected videotapes
crophones). One camera filmed the baby and the other one (30.6% of the sample). The mean Cohen’s κ coefficient for
filmed the mother. Both picture signals were transmitted interrater agreement was 0.87. To evaluate intraobserver
through a split-screen generator into a single camera to ob- reliability (coder drift), 15 randomly selected cases were
tain simultaneous frontal views of the mother and the in- scored again by the first coder after a 6-month interval (κ =
fant. The video recorder and the split-screen generator were 0.89).
located in an adjoining equipment room.

Infant-Mother Play Interaction


Coding of Infant Behavior During the FFSF
At the 9-month visit, infant-mother interactive behavior
The infants’ regulatory (coping) behavior during the three during free play was scored from videotapes using the
episodes of FSFF was scored micro-analytically second- Child-Adult Relationship Experimental Index (CARE In-
by-second from videotapes by a trained coder using the dex) (Crittenden, 2003) by a coder who was trained and
Infant Regulatory Scoring System (IRSS), a micro-analytic certified as reliable by Crittenden. A second trained observ-
and quantitative behavioral system (Tronick & Weinberg, er (blind to study details) independently scored 16 cases
1990). The IRSS codes the infant’s direction of gaze, vo- selected randomly. The mean Cohen’s κ coefficient for in-
calizations, pick-me-up gestures, other gestures, self-com- terrater agreement was high for both maternal (M κ = 0.90)
forting, distancing, and autonomic stress indicators. and infant (M κ = 0.85) scales (overall M κ = 0.87). Mean
In order to better discern individual differences, dis- κs for individual scales were 0.95 for maternal sensitivity,
crete IRSS behaviors were scored on three a priori summary 0.91 for maternal control, 0.84 for maternal unresponsivity,
dimensions describing the infant’s use of different coping 0.93 for infant cooperative behavior, 0.80 for infant com-
styles during the FFSF. Each infant behavior on a second-by- pliance, 0.88 for infant difficultness, and 0.80 for infant
second basis and was assigned to one of the three dimensions: passivity = .80.
positive other-directed coping (PODC) reflects the infants’ Following the instructions in the CARE-Index manual,
tendency to elicit social interaction and exhibit positive affect each dyad was videotaped for an average of 3 min
during the FFSF (i.e., infant looks at the mother’s face, looks (range = 2 to 5 min) during an unstructured free-play sit-
at the mother’s body, reaches toward or touches the adult, uation. The mothers were asked to play with their infant
makes positive vocalizations, or smiles); negative other-di- in a natural manner. No specific instructions were given
rected coping (NODC) reflects the infants’ tendency to ex- about how to play. A standard set of toys was provided
hibit negative affect and distancing behavior during the FFSF for the dyad, arranged on a blanket on the floor of the
(i.e., makes protest vocalizations, cries, arches or pushes play room. The CARE-Index assesses three dimensions
away from the mother, or attempts to escape from the mother of mother’s interactive behavior with the infant (sensi-
by turning away from the mother); and self-directed coping tive, controlling, and unresponsive), and four dimensions
(SDC) reflects the infant’s tendency to engage in SDC behav- of the infant’s interactive behavior with the mother (co-
iors during the FFSF (i.e., infant mouths his or her hands, operative, compulsive-compliant, difficult, and passive).

© 2009 Hogrefe Publishing European Psychologist 2009; Vol. 14(4):320–331


324 M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment

Although the mother and infant are scored independently, infants was evaluated using univariate frequency analyses.
as required by Crittenden (2003), the coder scores each Second, the bivariate association between attachment clas-
partner from the perspective of the other partner (i.e., the sification (secure, insecure-avoidant, insecure-resistant)
coder scores each partner taking into account the behav- and each potential predictor variable was evaluated. The
ior of the other partner). For instance, if the mother potential predictor variables included infant gender, birth-
smiles while the infant is distressed, the mother’s behav- weight, gestational age, Apgar scores, birth order, maternal
ior is considered insensitive whereas, if the mother smiles age, maternal education, SES, maternal representations of
in response to infant’s smile or attentive face, the moth- the infants’ difficult temperament, scores on the three clas-
er’s behavior is considered sensitive. Note that, although sifications of infant regulatory (coping) behavior (PODC,
the term “maternal sensitivity” often refers to an individ- NODC, SDC) during the FFSF at the 3-month visit, and
ual characteristic, in this scale it is operationalized as a the three rated dimensions of mothers’ interactive behavior
dyadic construct (Crittenden, 2003). Therefore, to avoid observed during play at the 9-month visit. α was set at .05.
redundancy in the analyses, we used only the maternal Third, the independent contributions of infant, maternal,
scales to assess variations in the quality of mother-infant and demographic variables to infant attachment status were
interactive behavior in this paper. evaluated using hierarchical regression analyses. Separate
Adequate discriminant validity of the CARE-Index regressions were used to predict each of the three attach-
has been reported. Differential outcomes using the mea- ment patterns (secure, insecure-avoidant, insecure-resis-
sure for middle class, low risk mothers, deaf mothers, low tant). Only variables identified as being significantly asso-
income mothers, mothers with mental retardation, abu- ciated with attachment status in the bivariate analyses were
sive mothers, and neglectful mothers have been reported evaluated as potential predictors in these regressions.
in separate studies (Crittenden & Bonvillian, 1984). Al-
though the CARE-Index does not assess mother-infant at-
tachment directly, the maternal scales have been strongly
linked with attachment outcomes. In samples of typical
mother-infant dyads, maternal sensitivity was related to Results
secure attachment, maternal control to avoidant attach-
ment, and maternal unresponsivity to resistant attach- Distribution of Attachment Classifications
ment (Crittenden, 1988).
Of the 48 infants who participated in this longitudinal
study, 16 (33.3%) infants were classified as securely at-
Attachment tached, 17 (35.4%) as insecure-avoidant, and 15 (31.3%)
as insecure-resistant at 12 months (corrected age).
At the 12-month visit, infants’ attachment behavior with
the mother was videotaped during Ainsworth’s Strange Sit-
uation (Ainsworth et al., 1978). The Strange Situation is a
21-min laboratory paradigm involving a sequence of eight Bivariate Association Between Attachment
episodes designed to place mild but increasing levels of and Potential Predictor Variables
stress on the infant and dyad (i.e., being introduced to an
unfamiliar play room, interacting with an unfamiliar adult Infant Medical and Maternal/Familial
stranger, and brief separations from and reunions with the Demographic Variables
mother). Scoring of infant attachment from the videotapes
of the Strange Situation was accomplished by two trained Contrary to our expectations, attachment status was not sig-
and reliable coders using traditional scoring guidelines de- nificantly associated with infant biologic or medical vari-
veloped by Ainsworth et al. (1978). All cases were scored ables, including gender, birth weight, gestational age, or
independently twice and classified as either secure, inse- Apgar scores. Similarly, attachment status was not signifi-
cure-avoidant, or insecure-resistant. Any disagreements in cantly associated with most maternal or familial demo-
attachment classification between the two coders were re- graphic variables including maternal age and SES. As hy-
solved in conference. Overall intercoder agreement for ma- pothesized, level of maternal education was significantly
jor classification prior to conferencing was over 81%. related to attachment status, F(2, 45) = 4.93, p < .02, ηp2 =
.18. Multiple range testing using the least significant dif-
ference (LSD) test with α set at 0.05 showed that educa-
Analytic Plan tional level was higher among mothers of both insecure-
avoidant (M = 3.50, SD = 1.59) and securely attached in-
Three sets of statistical analyses were carried out to address fants (M = 3.47, SD = 1.33) than mothers of
the goals of this study. First, the distribution of the three insecure-resistant infants (M = 2.13, SD = 1.19). No signif-
attachment classifications (secure, insecure-avoidant, inse- icant difference in education level was found between
cure-resistant) in this Portuguese sample of healthy preterm mothers of securely attached and insecure avoidant infants.

European Psychologist 2009; Vol. 14(4):320–331 © 2009 Hogrefe Publishing


M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment 325

Infants’ Difficult Temperament ence between the latter two groups was not significant). In
turn, infants who were classified as insecure-avoidant at 12
As predicted, a significant association was observed be- months used more SDC responses to regulate their stress
tween maternal representations of their infant’s difficult during the FFSF than infants classified as secure or inse-
temperament and infants’ attachment status at 12 months cure-resistant (the difference between the latter groups was
(corrected age), with low effect sizes, F(2, 45) = 4.73, p < not significant).
.02, ηp2 = .17. Results of a LSD posthoc test (α set at 0.05)
indicated that infants classified as securely attached (M =
23.32, SD = 3.71) were perceived as being less difficult, on Maternal Interactive Behavior During Free Play at
average, than infants classified as insecure-avoidant (M = 9 Months and Attachment
27.67, SD = 4.26) or insecure-resistant (M = 28.94, SD =
7.80). The mean values difference between these two last Mothers in this sample exhibited a wide range of scores on
groups was not significant. the sensitivity, control, and unresponsivity scales. Scores
ranged from 2 to 12 on the 14-point maternal sensitivity
scale (M = 6.33; SD = 2.83). An equivalent large range was
Infants’ Regulatory (Coping) Behavior in the FFSF observed for maternal control, with scores varying from 0
to 11 points (M = 4.56; SD = 3.34), and for maternal unre-
Infants’ mean score for PODC was 1.79 (SD = 1.15, sponsivity, with scores varying from 0 to 12 points (M =
range = 0.15 to 4.75), 1.43 of NODC score (SD = 1.12, 3.10; SD = 3.31).
range = 0.06 to 4.23), and 1.29 of SDC score (SD = 0.83; Descriptive statistics for the three measures of maternal
range = 0.00 to 3.67). interactive behavior during mother-infant free play at 9
In Table 1, descriptive statistics for positive other-direct- months (corrected age) are provided in Table 2 by attach-
ed coping, NODC, and SDC weighted scores are presented ment status.
according to infants’ attachment status. As hypothesized, results of one-way ANOVAs showed
As expected, results of one-way analyses of variance that infants’ attachment status at 12 months (corrected age)
(ANOVAs) showed that infants’ attachment status at 12 was associated with all three measures of maternal interac-
months (corrected age) was significantly linked to variations tive behavior during mother-infant free play at 9 months
in their coping style during the FFSF at 3 months (corrected (corrected age), with large to moderate effect sizes: mater-
age), with moderate effect sizes: PODC, F(2, 45) = 10.55, p < nal sensitivity, F(2, 45) = 45.78, p < .001, ηp2 = .67; mater-
.001, ηp2 = .32; NODC, F(2, 45) = 8.49, p < .005, ηp2 = .27; nal control, F(2, 45) = 26.97, p < .001, ηp2 = .55; and ma-
and SDC, F(2, 45) = 12.10, p < .001, ηp2 = .35; at 3 months ternal unresponsivity F(2, 45) = 14.40, p < .001, ηp2 = .39.
(corrected age). Results of LSD posthoc tests (α set at 0.05) indicated
Results of posthoc tests using the LSD procedure (α set that mothers of infants later classified as securely attached
at 0.05) revealed that infants classified as secure at 12 at 12 months had higher sensitivity scores during mother-
months exhibited significantly more positive other-direct- infant play at 9 months (corrected age) than mothers of
ed coping responses during the FFSF at 3 months (correct- insecure-avoidant or insecure-resistant infants (the differ-
ed age) than infants classified as insecure-avoidant or inse- ence in maternal sensitivity scores between the latter two
cure-resistant at 12 months. No significant difference was groups was not significant). Moreover, mothers of infants
found between these latter two groups. On the other hand, classified as insecure-avoidant at 12 months were more
infants classified as insecure-resistant at 12 months exhib- likely to be controlling with their infants during play at 9
ited more NODC responses during the FFSF than infants months than mothers of infants classified as secure or in-
classified as either secure or insecure-avoidant (the differ- secure resistant (no significant difference in controlling be-
havior was found for mothers of infants in these latter two
Table 1. Means and standard deviations for infants’ coping
behavior score during the FFSF* by each second
Table 2. Means and standard deviations for ratings of ma-
at 3 months by attachment classification at 12
ternal interactive behavior during mother-infant
months
free play at 9 months by attachment classification
Infant coping style Attachment classification at 12 months
at 3 months at 12 months
Maternal interactive Attachment classification at 12 months
Insecure- Secure Insecure-
behavior at 9 months Insecure- Secure Insecure-
avoidant resistant
avoidant resistant
(N = 16) (N = 17) (N = 15)
(N = 16) (N = 17) (N = 15)
Positive other-directed coping 1.20 (0.87) 2.65 (1.08) 1.47 (0.89)
Maternal sensitivity 4.88 (1.14) 9.41 (2.34) 4.40 (1.05)
Negative other-directed coping 0.78 (0.70) 1.26 (0.87) 2.21 (1.27)
Maternal control 7.94 (2.26) 2.29 (1.99) 3.53 (2.64)
Self-directed coping 2.09 (0.71) 1.19 (0.67) 0.73 (0.66)
*FFSF = Face-to-Face Still-Face paradigm. Maternal unresponsivity 1.19 (2.56) 2.29 (2.56) 6.07 (2.81)

© 2009 Hogrefe Publishing European Psychologist 2009; Vol. 14(4):320–331


326 M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment

Table 3. Results of multinomial logistic regressions predicting secure, insecure-avoidant, and insecure-resistant attachment
at 12 months
Regression model Pseudo R2
Predicting attachment –2 log likelihood 1
LR* χ²(df) Significance level Cox & Snell Nagelkerke
Secure attachment
1. Maternal sensitivity 34.724 42.18 (2) p < .001 .585 .658
2. Infant positive other-directed coping 56.993 22.27 (4) p < .001 .635 .714
3. Infant difficult temperament 54.97 2.02 (6) ns .650 .732
Insecure-avoidant attachment
1. Maternal control 34.20 33.06 (2) p < .001 .498 .560
2. Infant self-directed coping 61.62 27.42 (4) p < .001 .598 .673
Insecure-resistant attachment
1. Maternal unresponsivity 43.43 21.04 (2) p < .001 .355 .399
2. Infant negative other-directed coping 73.89 30.46 (4) p < .001 .481 .591
3. Maternal education 67.73 6.16 (6) ns .543 .611
1
Intercepts included in the computation of the models. *LR = likelihood ratio.

groups). In turn, mothers of infants classified as insecure- hierarchical regressions to predict each of the three attach-
resistant at 12 months were more unresponsive to their in- ment patterns (secure, insecure-avoidant, insecure-resis-
fants during play at 9 months, compared to mothers of in- tant). The selection of predictor variables included in each
fants classified as secure or insecure-avoidant (no signifi- regression model was based on the results from the LSD
cant difference in unresponsivity was found for mothers of post hoc tests in the one-way ANOVAs. However, not all
infants in these latter two groups). categories of variables qualified for retention in each re-
gression. Thus, the regression predicting secure attachment
status included three predictors: maternal sensitivity, infant
Hierarchical Regression Analyses Results PODC, and infant difficult temperament. The regression
predicting insecure-avoidant attachment included two pre-
Building on the results of the bivariate analyses, multino- dictors: maternal control and infant SDC. The regression
mial logistic regression was used to test for independent predicting insecure-resistant attachment included three
predictors of attachment style. This statistical technique is predictors: maternal unresponsivity, infant NODC, and ma-
an extension of the maximum likelihood-based estimations ternal education. These variables differentiated securely at-
(MLE) procedure where the dependent variable is categor- tached infants from insecure-avoidant and insecure-resis-
ical and has more than two nonordered categories (Hosmer tant infants without distinguishing the latter groups. How-
& Lemeshow, 2000). ever, in each analysis the retained predictor variables were
Only those variables identified in the bivariate analyses entered in the same order: (1) maternal interactive behavior
as being significantly associated with infant attachment sta- at 9 months, (2) infant coping style at 3 months, (3) infant
tus using an α of 0.05 or less were retained in the regression difficult temperament, and (4) maternal education. Because
analyses. Eight variables met these criteria: maternal level the impact of any given measure in a MLE-based model is
of education, maternal report of their infant’s difficult tem- not constant across values and cannot be interpreted inde-
perament, the three summary measures of the infant’s ca- pendently of other variable scores, parameters were esti-
pacity to regulate stress (coping) during the FFSF at 3 mated through rescaled independent variables with values
months, and the three ratings of mother interactive behav- changing between zero and one. Regression results are pre-
ior during free play at 9 months. All predictor variables, sented in Table 3.
however, could not be entered at the same time in a single In the regression predicting secure attachment, maternal
multinomial regression model for two reasons. First, the sensitivity (first step), infant PODC (second step), and infant
obtained scores of the CARE-Index coding system for the difficult temperament (third step) were evaluated as possible
three dimensions of maternal interactive behavior (sensi- predictors. Maternal sensitivity, with a likelihood ratio χ² of
tivity, control, and unresponsiveness) are linked measures. 42.18 with 2 degrees of freedom was statistically significant,
Considering them together in the same model would violate and PODC with a likelihood ratio χ² of 22.27 resulted in a
a critical assumption underlying multinomial logistic re- significant increment in R2. Both were retained as predictors.
gression. Second, the relatively small size of our sample Infant difficult temperament with a likelihood ratio of 2.02
resulted in a decrease of the ratio of cases to independent did not lead to a significant increment in R2.
variables, which would lower power. In the regression predicting insecure avoidant attach-
To address these issues, we carried out three separate ment, maternal control (first step) and infant SDC (second

European Psychologist 2009; Vol. 14(4):320–331 © 2009 Hogrefe Publishing


M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment 327

step) were evaluated as predictors. Both factors were re- (Soares & Dias, 2007), attachment distributions were sim-
tained as significant independent predictors of insecure- ilar to those reported in international meta-analytic studies
avoidant attachment. Maternal control was statistically sig- and contrast with those in the present study (van IJzen-
nificant with a likelihood ratio χ² of 33.06 with 2 degrees doorn & Sagi, 1999).
of freedom. Infant SDC had a likelihood ratio χ² of 27.42 Within the EU, mother-infant attachment has been stud-
with 4 degrees of freedom and contributed a statistically ied primarily in Western European countries such as Ger-
significant increment in R2. many and Great Britain. Overall, approximately 66% of
In the regression predicting insecure-resistant attach- infants in those countries exhibit secure attachment, which
ment, maternal unresponsivity (first step), NODC (second is roughly comparable to that reported for middle-class
step), and maternal education (third step) were evaluated samples in the US (van IJzendoorn & Sagi, 1999). Yet the
as predictors. Maternal unresponsivity and infant NODC percentage of secure attachment varies from country to
were retained as significant predictors. Maternal unrespon- country within Western Europe. For instance, the percent-
sivity had a likelihood ratio χ² of 21.04 with 2 degrees free- age of securely attached infants ranges from 57% in Ger-
dom, and infant NODC had a likelihood ratio χ² of 30.46 man samples to 75% secure in Great Britain samples (van
with 4 degrees of freedom, leading to a significant gain in IJzendoorn & Sagi, 1999). This variation is consistent with
R2. Maternal education, with a likelihood ratio of 6.16, did general findings from cross-cultural research, which indi-
not lead to a significant increment in R2. cate that intracultural differences are higher than intercul-
tural differences (van IJzendoorn & Kroonenberg, 1988).
A unique aspect of our sample is the large range in moth-
ers’ years of completed education, which varied from 4 and
Discussion 19 years. Nearly 63% of the mothers in this sample had not
attained a high school-level education. This is in striking
The analyses in this study were based on longitudinal data contrast to the higher level of education attained by moth-
collected in a Portuguese urban sample of medically low- ers in most attachment studies in other Western European
risk prematurely born infants and their healthy adult moth- countries and in the US (van IJzendoorn & Kroonenberg,
ers who were generally middle class but varied in years of 1988; van IJzendoorn & Sagi, 1999), including low income
completed education. Our aim was to identify specific in- samples. We speculate that this low level of maternal edu-
fant, maternal, and demographic variables measured during cation may have contributed to the lower prevalence of se-
the first year of life that independently predicted infants’ cure attachment in the present sample of preterm infants.
attachment status at 12 months (corrected age) in this un- Prior studies on attachment formation in preterm samples
derstudied sample. indicate a greater prevalence of insecure attachment only
when other risk conditions are present, such as low SES
(Pederson & Moran, 1996; Wille, 1991). It is likely that the
Distribution of Attachment Classifications consequences of premature birth combined with additional
risk factors such as low maternal education put infants and
In contrast to findings reported for healthy low-risk preterm caregivers in “double jeopardy” for forming insecure at-
infants in better educated, middle-class samples in the US tachments. Indeed, our results indicate a significant link
(e.g., Easterbrooks, 1989; Frodi & Thompson, 1985), there between maternal education and the insecure-resistant at-
was an unexpectedly high prevalence of insecure attach- tachment pattern.
ment in this sample of prematurely born Portuguese infants.
Thirty-two of the 48 (66.6%) infants were classified as in-
securely attached (35.4% avoidant, 31.3% resistant) at 12
months (corrected age). Factors Linked to Mother-Infant Attachment
The reason for the low prevalence of secure attachment
is unclear, especially given the paucity of normative data
in Bivariate Analyses
for mother-infant attachment in the Portuguese culture. Re-
sults from a few attachment studies with healthy, middle- Results of bivariate analyses indicated that both maternal
class Portuguese samples using the Strange Situation indi- and infant behavioral variables measured during the first
cate that the prevalence of secure attachment varied be- year of life were associated with infants’ attachment status
tween 45% and 50% (Costa, 2000; Soares et al., 1995; at 12 months, but (contrary to our expectations) infant bi-
Soares, Silva, Cunha, Costa, & Lopes dos Santos, 1998). ologic variables (e.g., gestational age, birth weight) were
This is lower than that reported in international meta-ana- not. This pattern of findings may reflect the relatively low
lytic attachment research (van IJzenoorn & Kroonenburg, medical-risk status of the infants in the current sample. Of
1988), but higher than that found in the present study. In the demographic variables evaluated in this study (e.g.,
other attachment research with Portuguese samples using SES, maternal age), only level of maternal education was
different methodology, i.e., the Attachment Q-Sort significantly related to attachment. Specifically, mothers of
(Vaughn et al., 2007) or the Adult Attachment Interview infants classified as insecure-resistant had completed sig-

© 2009 Hogrefe Publishing European Psychologist 2009; Vol. 14(4):320–331


328 M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment

nificantly fewer years of education than mothers of infants coping behavior (PODC, NODC, SDC) during the FFSF
from the other two attachment groups. and specific attachment patterns (secure, insecure-avoid-
As expected, mothers of securely attached infants per- ant, and insecure resistant). To our knowledge, our study is
ceived their infants as being less difficult in temperament the first to sort infants’ reactions during each of the three
at 1 and 3 months (corrected age), compared to mothers of episodes of the FFSF at 3 months into these distinct sum-
insecurely attached infants. These findings corroborate the mary styles of behavior. These styles were identified from
results of previous studies in which mothers of securely individual behaviors derived from the IRSS, a microana-
attached infants were more likely to give positive descrip- lytic and quantitative behavioral system with more than 20
tions of their infants’ temperament than mothers of inse- discrete infant behaviors, scored in each episode of the
curely attached infants (Benoit et al., 1997; Bretherton et FFSF. Thus, it was possible to examine whether different
al., 1989; Cox et al., 2000; Izard et al., 1991; Kochanska, styles of infant coping during the FFSF were effective pre-
1998; Zeanah et al., 1994). Vaughn and Bost (1999) and dictors of attachment style.
others have suggested that this association may be strongly Consistent with an extensive literature relating maternal
influenced by the quality of the mother’s interactions with sensitivity to the infant during the first year of life and in-
their infant, in a manner similar to Crittenden’s argument fant attachment at 12 months (e.g., Ainsworth et al., 1978;
that maternal sensitivity is actually a dyadic measure (Crit- Isabella, 1993), we found that mothers of infants later clas-
tenden, 2003). sified as secure were more sensitive with their infants dur-
Similarly, variations in infants’ coping behavior during ing mother-infant play at 9 months (corrected age), than
the FFSF at 3 months were linked with infants’ attachment mothers of infants later classified as insecure-attached. We
status at 12 months. Infants classified as having secure at- extended this research by showing associations between
tachment at 12 months exhibited more PODC responses specific maternal behaviors during play at 9 months and
during the FFSF at 3 months, compared to infants classified attachment patterns at 12 months: Mothers of infants later
as having insecure attachment. In turn, infants classified as classified as insecure-avoidant were more controlling dur-
insecure-resistant at 12 months exhibited more NODC re- ing mother-infant play at 9 months than mothers of infants
sponses during the FFSF than either infants classified as later classified as secure or insecure-resistant. In turn,
having secure or insecure-avoidant attachment. In turn, in- mothers of infants later classified as insecure-resistant were
fants classified as insecure-avoidant at 12 months engaged more unresponsive to their infants than mothers of infants
in more SDC behaviors during the FFSF at 3 months, com- later classified as secure or insecure-avoidant. These find-
pared to infants with the other two attachment patterns. ings corroborate similar findings reported in the attachment
Although understudied, a few studies also have reported literature (see Ainsworth et al., 1978; Cassidy & Shaver,
links between variations in infant affective regulation dur- 1999). It is unclear, however, why there were no significant
ing FFSF and later attachment status. However, specific group differences in maternal controlling behavior between
results vary. Cohn et al. (1991) found that 6-month-old in- mothers of secure vs. insecure-resistant infants, or in ma-
fants who made positive elicitations to the mother during ternal unresponsivity between mothers of secure and inse-
the still-face episode of the FFSF were more likely to be cure-avoidant infants. This lack of significant group differ-
securely attached at 12 months. However, no relation was ences may reflect, in part, the CARE-Index scoring system.
reported between other types of infant coping behavior and In fact, this scale challenges the coder to distribute 14
insecure attachment patterns. Braungart-Rieker et al. points across three categories: maternal sensitivity, mater-
(2001) found that the amount that infants exhibited nega- nal control, and maternal unresponsivity. Thus, mothers
tive arousal during the still-face episode of the FFSF at 4 with high scores on a specific scale have necessarily low
months of age was related with level of distress exhibited points on the two other scales. For example, mothers of
during the Strange Situation in later infancy, rather than securely attached infants with high scores on maternal sen-
with secure vs. insecure attachment status per se. Specifi- sitivity also have a low score on control. At the same time,
cally, infants who cried less during the FFSF were more mothers of resistant-attached infants, by scoring high on
likely to be classified as having either the B1 or B2 patterns unresponsivity, have low scores on control as well. There-
of secure attachment, or as insecure-avoidant. Each of these fore, the differences in control behavior between these two
attachment patterns are associated with less infant distress groups of mothers are not high enough to be significant.
during the Strange Situation. Conversely, infants who cried
more during the FFSF were later classified as having either
the B3 or B4 patterns of secure attachment or insecure-re-
sistant attachment. In another study, variations in discrete
Specific Predictors of Attachment in
infant behaviors observed during the FFSF, such as infant Multivariate Analyses
gaze, self-touch, or facial expressions differentiated infants
later classified as having secure vs. avoidant attachment The contributions of maternal interactive behavior, infant
(Koulomzin et al., 2002). coping behavior, maternal representation of infants’ tem-
Our study adds to this research and is unique in finding perament, and maternal education to attachment status
a significant association between specific styles of infant were evaluated using multinomial logistic regression anal-

European Psychologist 2009; Vol. 14(4):320–331 © 2009 Hogrefe Publishing


M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment 329

yses. As hypothesized, both maternal interactive behavior also indicate that infant styles of coping with stressors dur-
with the infant during mother-infant free play at 9 months ing early infancy also plays an important role. In the present
and infants’ regulatory (coping) behavior during the FFSF study, both factors were significant independent predictors
at 3 months were independent predictors of specific attach- of later attachment styles. In a recent paper, Tronick (2006)
ment patterns at 12 months. Specifically, secure attachment has hypothesized that the parent-infant dyad’s ability to re-
was predicted by a higher level of maternal sensitivity to pair interactive errors, a measure related to and more spe-
the infant during play and infants’ greater use of a PODC cific than maternal sensitivity, is linked to the development
style during the FFSF. In contrast, insecure-avoidant at- of infants’ capacities to cope with stress, and that, recipro-
tachment was predicted by a higher level of maternal con- cally, mothers of infants with greater coping capacities may
trolling (intrusive) behavior during play and infants’ ten- find it easier to engage in sensitive parenting.
dency to exhibit a SDC style during the FFSF. In turn, in- Our results may also be helpful to professionals who
secure-resistant attachment was predicted by a higher level provide intervention services to young babies and may con-
of maternal unresponsive behavior with the infant during tribute to the development of more accurate evidence-
play and infants’ tendency to exhibit a NODC style during based practice models. In particular, our findings suggest
the FFSF. that careful consideration of both infants’ coping with
Classical attachment theorists have highlighted maternal stress and the quality of mothers’ interactive behavior with
sensitivity, and temperament theorists have highlighted in- their infant should be considered in designing more effec-
fant reactivity, as central factors predicting attachment se- tive interventions. Our findings suggest that such interven-
curity. Our findings suggest that both of these factors play tions need to be specific with respect to both the forms of
an important role in attachment formation this low-risk pre- coping exhibited by the infant and the quality of parental
term sample. Our findings are also consistent with findings interactive behavior exhibited by the parent. Both factors
from an international meta-analytic review that showed were significantly and independently associated with infant
that maternal sensitivity alone explains only a relatively attachment formation in this study.
modest amount of the variance in attachment status (e.g.,
De Wolff & van IJzendoorn, 1997).
Several limitations of our study must be acknowledged. Acknowledgments
The relatively small sample size prevented us from evalu-
ating potential indirect as well as direct effects of maternal, The first author was supported by a postdoctoral fellowship
infant, and demographic variables on infants’ attachment from Fundação para a Ciência e Tecnologia (SFRH/BPD/
formation. Additional longitudinal research with larger 26726/2006) and a postdoctoral fellowship from the Child
samples is needed to investigate how multiple factors from Development Unit, Children’s Hospital, Harvard Medical
different levels of analysis – and the complex interplay School, Boston. The researchers would like to thank the
among them – may contribute to individual differences in families that participated in this study. We are also grateful
infant attachment status. Indeed, we still do not know how to Joana Barreiros, Sandra Carneiro, Raquel Corval, Carla
variations in maternal sensitivity to preterm infants early in Guedes, and Gustavo Mendes, who helped carry out the
life contributes to infants’ ability to cope with stressors dur- procedures; Sandra Carneiro, who independently scored
ing the FFSF in early infancy, and vice versa, and how the FFSF data, and Hélia Soares who coded the CARE-In-
biologic and demographic risk factors affect these dynam- dex for reliability.
ic, transacting variables.
Another limitation is that the specific characteristics of
this sample (e.g., an urban Portuguese sample of healthy
preterm infants and their mothers varying in level of edu-
cation) may have contributed to the higher-than-expected References
rate of insecure attachment in this sample and preclude our
generalizing the results to other samples. Further studies on Ainsworth, M.D., Bell, S., & Stayton, D. (1974). Infant/mother
larger samples of Portuguese preterm infants varying in attachment and social development: “Socialization” as a prod-
level of biologic risk are needed. On the other hand, the uct of reciprocal responsiveness to signals. In P.M. Richards
(Ed.), Integration of a child into a social world (pp. 99–135).
sample’s unique characteristics may also be viewed as a
Cambridge, UK: Cambridge University Press.
major strength. Our study is the first study to evaluate lon-
Ainsworth, M.D., Blehar, M., Waters, E., & Wall, S. (1978). Pat-
gitudinally the specific, independent predictors of attach- terns of attachment: A psychological study of the strange sit-
ment in this understudied Mediterranean culture. uation. Hillsdale, NJ: Erlbaum.
In sum, this research indicates that, although several fac- Barnard, K.E., & Kelly, J.F. (1990). Assessment of parent-child
tors are associated with attachment in bivariate analyses, interaction. In S.J. Meisels & J.P. Shonkoff (Eds.), Handbook
only a few were retained as significant predictors of attach- of early childhood intervention (pp. 278–302). New York:
ment status in multivariate analyses. Our findings corrob- Cambridge University.
orate prior attachment research in highlighting the impor- Belsky, J. (2005). Attachment theory and research in ecological
tance of maternal sensitivity in attachment formation, but perspective. In K. Grossmann, K. Grossmann, & E. Waters

© 2009 Hogrefe Publishing European Psychologist 2009; Vol. 14(4):320–331


330 M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment

(Eds.), Attachment from infancy to adulthood: The major lon- in the strange situation: Effects of prematurity and of quality
gitudinal studies (pp. 71–97). New York: Guilford. of attachment. Child Development, 56, 1280–1290.
Benoit, D., Parker, K.C., & Zeanah, C.H. (1997). Mothers’ rep- Goldberg, S., & DiVitto, B. (1995). Parenting children born pre-
resentations of their infants assessed prenatally: Stability and term. In M. Bronstein (Ed.), Handbook of parenting (Vol. 1:
association with infants’ attachment classifications. Journal of Children and parenting, pp. 209–231). Mahwah, NJ: Erlbaum
Child Psychology and Psychiatry, 38, 307–313. Goldsmith, H.H., & Alansky, J.A. (1987). Maternal and infant
Bowlby, J. (1969). Attachment and loss (Vol. I). London: Penguin. temperamental predictors of attachment – A meta-analytic re-
Braungart-Rieker, J.M., Garwood, M.M., Powers, B.P., & Wang, view. Journal of Consulting and Clinical Psychology, 55,
X. (2001). Parental sensitivity, infant affect, and affect regula- 805–816.
tion: Predictors of later attachment. Child Development, 72(1), Grossmann, K., Grossmann, K.E., Spangler, G., Suess, G., &
252–270. Unzner, L. (1985). Maternal sensitivity and newborns’ orien-
Bretherton, I., Biringen, Z., Ridgeway, D., Maslin, C., & Sher- tation responses as related to quality of attachment in northern
man, M. (1989). Attachment – The parental perspective. Infant Germany. Monographs of the Society for Research in Child
Mental Health Journal, 10, 203–221. Development, 50(1–2), 233–256.
Bugental, D.B., & Happaney, K. (2004). Predicting infant mal- Hosmer, D.W., & Lemeshow, S. (2000). Applied logistic regres-
treatment in low-income families: The interactive effects of sion. New York: Wiley.
maternal attributions and child status at birth. Developmental INE. (2004). Estatísticas Demográficas [Demographic statistics].
Psychology, 40, 234–243. Lisbon: Instituto Nacional de Estatistica.
Cassidy, J., & Shaver, P.R. (Eds.). (1999). Handbook of attach- Isabella, R.A. (1993). Origins of attachment: Maternal interactive
ment: Theory, research, and clinical applications. New York: behavior across the first year. Child Development, 64,
Guilford. 605–621.
Cicchetti, D., & Barnett, D. (1991). Attachment organization in Izard, C.E., Haynes, O.M., Chisholm, G., & Baak, K. (1991).
maltreated preschoolers. Development and Psychopathology, Emotional determinants of infant-mother attachment. Child
3, 397–411. Development, 62, 906–917.
Cohn, J.F., Campbell, S.B., & Ross, S. (1991). Infant responses Kagan, J. (1984). The nature of the child. Cambridge, MA: Har-
in the still-face paradigm at 6 months predicts avoidant and vard University Press.
secure attachment at 12 months. Development and Psychopa- Kochanska, G. (1998). Mother-child relationship, child fearful-
thology, 3, 367–376. ness, and emerging attachment: A short-term longitudinal
Costa, A. (2000). Qualidade da integração sensorial e organi- study. Developmental Psychology, 34, 480–490.
zação dos comportamentos de vinculação [Quality of sensory Kochanska, G., & Coy, K.C. (2002). Child emotionality and ma-
integration and attachment behavior]. Porto: University of Por- ternal responsiveness as predictors of reunion behaviors in the
to. strange situation: Links mediated and unmediated by separa-
Cox, S.M., Hopkins, J., & Hans, S.L. (2000). Attachment in pre- tion distress. Child Development, 73, 228–240.
term infants and their mothers: Neonatal risk status and mater- Koulomzin, M., Beebe, B., Anderson, S., Jaffe, J., Feldstein, S.,
nal representations. Infant Mental Health Journal, 21, & Crown, C. (2002). Infant gaze, head, face, and self-touch at
464–480. 4 months differentiate secure vs. avoidant attachment at 1 year:
Crittenden, P.M. (1988). Relationships at risk. In J. Belsky & T. A microanalytic approach. Attachment and Human Develop-
Nezworski (Eds.), The clinical implications of attachment. ment, 4(1), 3–24.
Hillsdale, NJ: Erlbaum. Lopes dos Santos, P., Fuertes, M., & Sanches-Ferreira, M. (2005).
Crittenden, P.M. (2000). A dynamic-maturational exploration of A percepção materna dos atributos temperamentais do bebé:
the meaning of security and adaptation: Empirical, cultural, Características psicométricas de um questionário e seu valor
and theoretical considerations. In P.M.C.A. Claussen (Ed.), prognóstico relativamente à qualidade da vinculação [Mater-
The organization of attachment relationships: Maturation, cul- nal perception of babies’ temperaments: Scale validation and
ture, and context (pp. 358–384). New York: Cambridge Uni- attachment prediction]. In J. Bairrão (Ed.), Desenvolvimento:
versity Press. Contextos familiares e educativos [Development: Family and
Crittenden, P.M. (2003). CARE-Index manual. Unpublished educational contexts] (pp. 142–170). Porto: Livpsic.
manuscript, Family Relations Institute, Miami. Muller-Nix, C., Forcada-Guex, M., Pierrehumbert, B., Jaunin, L.,
Crittenden, P.M., & Bonvillian, J.D. (1984). The relationship be- Borghini, A., & Ansermet, F. (2004). Prematurity, maternal
tween maternal risk status and maternal sensitivity. American stress, and mother-child interactions. Early human Develop-
Journal of Orthopsychiatry, 54, 250–262. ment, 79, 145–158.
Crnic, K., Ragozin, S., Greenberg, M., Robinson, M., & Basham, Pederson, D.R., & Moran, G. (1996). Expressions of the attach-
R. (1983). Social interaction and development competence of ment relationship outside of the strange situation. Child De-
preterm and full-term during the first year of life. Child Devel- velopment, 67, 915–927.
opment, 54, 1199–1210. Plunkett, J.W., Meisels, S.J., Stiefel, G.S., Pasick, P.L., & Roloff,
De Wolff, M.S., & van IJzendoorn, M.H. (1997). Sensitivity and D.W. (1986). Patterns of attachment among preterm infants of
attachment: A meta-analysis on parental antecedents of infant varying biological risk. Journal of the American Academy of
attachment. Child Development, 68, 571–591. Child and Adolescent Psychiatry, 25, 794–800.
Easterbrooks, M.A. (1989). Quality of attachment to mother and Poehlmann, J., & Fiese, B.H. (2001). The interaction of maternal
to father – Effects of perinatal risk status. Child Development, and infant vulnerabilities on developing attachment relation-
60, 825–830. ships. Development and Psychopathology, 13(1), 1–11.
Frodi, A., & Thompson, R. (1985). Infants’ affective responses Raikes, H.A., & Thompson, R.A. (2005). Links between risk and

European Psychologist 2009; Vol. 14(4):320–331 © 2009 Hogrefe Publishing


M. Fuertes et al.: Infant Coping, Still-Face, Maternal Behavior, and Attachment 331

attachment security: Models of influence. Journal of Applied base scripts predicts children’s secure-base behavior at home
Developmental Psychology, 26, 440–455. in three sociocultural groups. International Journal of Behav-
Rutter, M. (1991). Protective factors – Independent or interactive. ioral Development, 31(1), 65–76.
Journal of the American Academy of Child and Adolescent Wille, D.E. (1991). Relation of preterm birth with quality of moth-
Psychiatry, 30(1), 151–152. er-infant attachment at one year. Infant Behavior and Devel-
Soares, I., & Dias, P. (2007). Apego y psicotalogía en jóvenes y opment, 14, 227–240.
adultos: Contribuciones recientes de la investigacion [Attach- Zeanah, C.H., Benoit, D., Hirshberg, L., Barton, M.L., & Regan,
ment and psychopathology in adults: Recent contributions C. (1994). Mothers’ representations of their infants are con-
from research]. Internacional Journal of Clinical and Health cordant with infant attachment classifications. Developmental
Psychology, 7(1), 177–195. Issues in Psychiatry and Psychology, 1, 1–14.
Soares, I., Lopes dos Santos, P., Jorgenlen, I., Henriques, M., Silva,
A., Figueiredo, B. et al. (1995). Organização comportamental da
vinculação na Situação Estranha: Avaliação da estabilidade
através do método de teste-reteste [Attachment organization in
strange situations: Stability study. In L.A.I. Ribeiro (Ed.), About the authors
Avaliação psicológica: Formas e contextos, 3, 247–258.
Soares, I., Silva, M.C., Cunha, J.P., Costa, O., & Lopes dos San- Marina Fuertes, Ph.D., is a Research Associate in the Child De-
tos, P. (1998). Attachment organization and heart-rate variabil- velopment Unit at Harvard Medical School (Boston, MA, USA)
ity in the strange situation, Aquém e Além do Cérebro [Behind and chief of an early intervention team (Gabinete de Apoio ao
and beyond the brain] (pp. 285–291). Porto: Fundação Bial. Desenvolvimento Infantil e à Família – GADIF) in Lisbon, Por-
Sroufe, L.A. (1996). Emotional development: The organization tugal. Her scientific interests include mother-infant attach-
of emotional life in the early years. Cambridge: Cambridge ment, risk and resilience processes in child development, and
Studies in Social and Emotional World. evolutionary psychology.
Tronick, E. (2006). The stress of normal development and inter-
action leads to the development of resilience and variation. Pedro Lopes dos Santos, Ph.D., is a psychologist teaching as As-
Annals New York Academy of Sciences, 1094, 83–104. sociate Professor in the Psychology Department of Oporto Uni-
Tronick, E., Als, H., Adamson, L., Wise, S., & Brazelton, T.B. versity, Portugal. His scientific interests include mother-infant at-
(1978). Infants response to entrapment between contradictory tachment, infant special needs, and theoretical developmental
messages in face-to-face interaction. Journal of the American models.
Academy of Child and Adolescent Psychiatry, 17(1), 1–13.
Tronick, E.Z., & Weinberg., M.K. (1990). The Infant Regulatory Marjorie Beeghly, Ph.D., is Associate Professor in Psychology at
Scoring System (IRSS). Unpublished manuscript, Boston, Wayne State University (Detroit, MI, USA), Lecturer in Pediatrics
Childrens Hospital & Harvard Medical School. at Harvard Medical School, and Senior Research Associate at
van den Boom, D. (1994). The influence of temperament and moth- Children’s Hospital, both Boston, MA, USA. Her research eval-
ering on attachment and exploration: An experimental manipu- uates how biologic and social risk and resilience factors alter tra-
lation of sensitive responsiveness among lower-class mothers jectories of communicative and socioemotional development in
with irritable infants. Child Development, 65, 1457–1477. groups of children at risk for developmental problems because of
van IJzendoorn, M.H., Goldberg, S., Kroonenberg, P.M., & Frenkel, stressors such as premature birth, poverty, intrauterine drug ex-
O.J. (1992). The relative effects of maternal and child problems posure, developmental disorders, violence exposure, child mal-
on the quality of attachment: A meta-analysis of attachment in treatment, or maternal psychopathology.
clinical samples. Child Development, 63, 840–858.
van IJzendoorn, M.H., & Kroonenberg, P.M. (1988). Cross-cul- Ed Tronick, Ph.D., is Distinguished University Professor of Psy-
tural patterns of attachment: A meta-analysis of the strange chology at the University of Massachusetts, Lecturer in Pediatrics
situation. Child Development, 59, 147–156. at Harvard Medical School, Director of the Child Development
van IJzendoorn, M.H., & Sagi, A. (1999). Cross-cultural patterns Unit at Children’s Hospital, Boston, MA, USA, and on the Faculty
of attachment: Universal and contextual dimensions. In J. Cas- at Fielding Graduate University, Santa Barbara, CA, USA. His
sidy & P. Shaver (Eds.), The handbook of attachment: Theory, prime research focus is on the social emotional develop-
research, and clinical applications (pp. 713–734). New York: ment of infants in the context of parent-infant interaction in the
Guilford. US and other cultural communities
van IJzendoorn, M.H., Schuengel, C., & Bakermans-Kranenburg,
M.J. (1999). Disorganized attachment in early childhood:
Meta-analysis of precursors, concomitants, and sequelae. De-
velopment and Psychopathology, 11, 225–249. Marina Fuertes
Vaughn, B.E. & Bost, K.K. (1999). Attachment and temperament:
Redundant, independent, or interacting influences on interper- Child Development Unit, Children’s Hospital
sonal adaptation and personality development? In J. Cassidy 1295 Boylston Street, Suite 320
& P.R. Shaver (Eds.), Handbook of attachment: Theory, re- Boston, MA 02215
search, and clinical applications (pp. 198–225). New York: USA
Guilford. Tel. +1 617 355-3545
Vaughn, B.E., Coppola, G., Verissimo, M., Monteiro, L., Santos, Fax +1 617 730-0074
A.J., Posada, G. et al. (2007). The quality of maternal secure- E-mail marina.fuertes@childrens.harvard.edu

© 2009 Hogrefe Publishing European Psychologist 2009; Vol. 14(4):320–331

You might also like