Infant attachment: Contributions of infant temperament and maternal characteristics

Preview:

Citation preview

Infant attachment: Contributions of infant temperamentand maternal characteristics

Sarah C. Mangelsdorfa,*, Jean L. McHaleb, Marissa Dienerc,Lauren Heim Goldsteind, Lisa Lehne

aDepartment of Psychology, University of Illinois, 603 E. Daniel, Champaign, IL 61820, USAbShaftsbury, VT, USA

cUniversity of Utah, USAdUniversity of California-Berkeley, USA

eUniversity of Minnesota, USA

Received 2 August 2000; received in revised form 21 September 2000; accepted 8 February 2001

Abstract

This study examined the joint contributions of maternal characteristics and infant characteristics toquality of attachment. When infants were 8 months, one hundred and two mothers and their infantscompleted a videotaped interaction and infants completed a laboratory assessment of temperament.Mothers completed personality and infant temperament questionnaires. At 12 months, infant motherattachment quality was measured in the Strange Situation. In a discriminant function analysis usingboth child and maternal characteristics, seventy-eight percentage of infants were correctly classifiedas secure, resistant, or avoidant. Insecurely attached infants were higher on activity and distress tonovelty and had mothers who were lower on Constraint than securely attached infants. Infantsclassified as avoidant were lower on positive affect and higher on fearfulness and had mothers loweron positive affectivity than infants classified as resistant. The results of this study point to theimportance of examining both parent and child characteristics in the prediction of attachment. © 2000Elsevier Science Inc. All rights reserved.

Keywords:Attachment; Temperament; Maternal personality

* Corresponding author. Tel.:11-217-244-0672; fax:11-217-244-5876.E-mail address:smangels@s.psych.uiuc.edu (S.C. Mangelsdorf).

Infant Behavior & Development 23 (2000) 175–196

0163-6383/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved.PII: S0163-6383(01)00035-2

1. Introduction

In recent years there has been considerable controversy regarding the antecedents ofattachment security in infancy. Traditional attachment theory accounts (e.g., Ainsworth et al.,1978) have focused on maternal characteristics, especially sensitivity (e.g., Bretherton, 1985;de Wolff & van Ijzendoorn, 1997), as the primary determinant of attachment security. Incontrast, some temperament researchers (e.g., Buss & Plomin, 1986; Kagan, 1982) haveargued that infants’ responses to the Strange Situation procedure have more to do withendogenous characteristics of the infants than to the sensitivity of maternal care. In response,attachment theorists such as Sroufe (1985) have suggested three possible explanations of theassociations between temperament and attachment. The first explanation is that temperamentand attachment are orthogonal constructs (temperamental variation may influence aspects ofbehavior, but not the overall organization or security of attachment). The second explanationis that temperament may determine what type of insecurity (avoidance or resistance) that thechild manifests, and the third explanation is that relationship history totally transformsconstitutional temperament variation such that it makes little or no contribution to the qualityof attachment.

Belsky and Rovine (1987) examined Sroufe’s hypothesis that temperament would influ-ence thetypeof insecurity that infants manifest, rather than security or insecurity overall.They found significant differences in temperament measures during the first year for infantswho were later classified as A1-B2 (low reactive) and B3-C2 (high reactive) in the StrangeSituation. However, temperamental differences in these reactivity groupings have not beendetected in some other investigations (e.g., Mangelsdorf et al., 1990), though they have beenin others (e.g., Fox, Kimmerly & Schafer, 1991; Susman-Stillman et al., 1996). For example,Susman-Stillman and her colleagues found that infants rated by their mothers as high onsociability were more likely to be classified as A or B1 or B2 whereas infants high onirritability were more likely to be classified as B3, B4 or C.

The extent to which temperament has been found to directly influence attachment in partdepends on the method by which temperament constructs are assessed. Examination of theempirical literature yields some evidence of main effects of parent reports of temperamenton attachment classifications. For example, Calkins and Fox (1992) using Rothbart’s InfantBehavior Questionnaire (IBQ) reported that maternal ratings of infant activity level werepredictive of avoidant attachment. Furthermore, investigators using Bates Infant Character-istics Questionnaire (ICQ) have noted that resistant/ambivalent infants scored higher on theFussy/Difficult scale (e.g., Moran & Pederson, 1997; Seifer et al., 1996). Although a fewstudies have found main effects of maternal reports of temperament on attachment outcomes,it has been noted more frequently that maternal reports of infant temperament are predictiveof infant behaviors in the Strange Situation, but not to overall attachment classifications(Bates et al., 1985; Gunnar et al., 1989; Vaughn et al., 1989). Specifically, researchers havefound that temperamental measures of negative affectivity are related to negative affectivityexpressed during the Strange Situation. These findings are similar to those of Goldsmith andhis colleagues in their meta-analyses examining temperament and attachment (Goldsmith &Alanksy, 1987; Goldsmith, Bradshaw, & Reisser-Danner, 1986). They reported that negative

176 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

temperamental measures of negative affectivity were related to resistant behavior within theStrange Situation.

Fewer studies have examined the relations between observational measures of tempera-ment and attachment. Studies that have employed observational measures of temperamenthave found some associations between temperament and attachment. For example, in theMother-Child project at the University of Minnesota a direct association between scores onBrazelton’s Neonatal Assessment Scale (NBAS) and later resistant attachment was found(Egeland & Farber, 1984; Waters, Vaughn, & Egeland, 1980). A more recent investigationfrom the same longitudinal study was conducted by Susman-Stillman and her colleagues(Susman-Stillman et al., 1996) using a combination of both observational and maternalreports measures of temperament at 3 and 6 months to predict later attachment. Susman-Stillman et al. found among other things, a modest direct effect of infant irritability duringthe first year of life on later insecure resistant attachment.

Using observational measures of temperament from early infancy Calkins and Fox (1992)found that infants who cried more during pacifier withdrawal at 2 days of age were morelikely to be classified as insecurely attached at 14 months. Exploring both negative andpositive emotional expressivity from 2.5 to 7 months as antecedents of attachment, Malat-esta, Culver, Tesman, and Shephard (1989) found that infant who displayed increases insmiling across the first year were more likely to be securely attached, while infants whodisplayed decreases in smiling were more likely to be insecurely attached.

Seifer et al. (1996) employed both observers’ ratings of temperament aggregated acrossmultiple home observations, and maternal ratings temperament. Interestingly, they did notfind any associations between observations of temperament and Strange Situation classifi-cations. However, in a recent investigation using laboratory assessments of infant temper-ament Kochanska (1998) demonstrated that observed infant fearfulness did not differentiatebetween secure and insecurely attached infants, but was predictive of insecure resistantattachment versus insecure avoidant attachment. Thus, in her investigation infant tempera-ment predicted the type of insecurity the child manifested, but not whether s/he was securelyattached. Such research suggests that focusing exclusively on global measures of tempera-ment such as negative affectivity or proneness-to-distress as predictors of attachment may beless informative than examining specific aspects temperament, such as fearfulness, anger-proneness or positive affectivity. Furthermore, her results point to the importance of exam-ining the distinction between different types of insecurity, rather than just making a secure/insecure distinction.

The disparate findings in the literature on temperament and attachment may in part be duedifferences in sample sizes and temperamental variability across samples. When extremegroups are examined the associations between temperament and attachment appear muchmore robust. For example, higher rates of insecurity are found in research in which infantshave been selected specifically for being extreme on proneness-to-distress or irritability inearly infancy. For example, van den Boom (1989; 1994) selected infants who were in the top17% in irritability. In that sample, she found that 78% of highly irritable infants whosemother did not receive any form of parenting intervention were classified as insecurelyattached, in striking contrast to rates of insecurity of about 30–35% found in most normalsamples. Results from selected populations such as these suggest a number of possible

177S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

explanations for the mixed results observed across studies examining the impact of temper-ament on attachment. It is possible that infant characteristics are only influential in thedevelopment of attachment security if they are extreme. More likely, however, is that infantcharacteristics may still be influential, but their influence is more difficult to detect in normalsamples. Such effects may be present, but may not be easily identifiable because of theinadequate statistical power associated with small sample sizes (Goldsmith et al., 1986).Given the unequal distribution of avoidance, security and resistance in normal populations,it may be difficult to find significant differences when comparisons are made betweenattachment categories with small numbers of subjects in each group, as is typically true ofthe resistant and avoidant groups (Fox, 1992). Such findings suggest that it might be mostfruitful to examine the contributions of multiple dimensions of child behavior concurrentlyto examine their collective predictive power in predicting attachment security.

Sameroff and his colleagues (e.g., Sameroff & Chandler, 1975; Sameroff & Fiese, 1990)have proposed that in order to understand developmental outcomes we must examinetransactions between characteristics of the child, and characteristics of the caregivingenvironment. A number of investigators have emphasized the importance of transactionalmodels for understanding the development of attachment relationships, suggesting that wemust examine both maternal and infant contributions in the development of the attachmentrelationship (e.g., Belsky & Isabella, 1988; Crockenberg, 1981; Crockenberg, 1986).

One maternal characteristic which has received some attention in attachment research ispersonality. Considerable evidence suggesting that maternal personality is related to mother-infant attachment has been accrued by researchers using non-normative samples. For exam-ple, children of mother diagnosed with affective disorders are more likely to be insecurelyattached than children of normal parents (DeMulder & Radke-Yarrow, 1991; Gaensbauer etal., 1984; Radke-Yarrow et al., 1985; Teti et al., 1995). Few relations, however, have beenfound between normal range variations in maternal personality and security of attachment.For example, using a large battery of measures, Egeland and Farber (1984) found that onlytwo variables—maternal maturity and complexity of thinking—predicted security of attach-ment at 12 months. Similarly, Belsky and Isabella (1988) found that: 1) mothers of securebabies scored higher on interpersonal affection than mothers of insecure infants, and 2)mothers of avoidant infants displayed significantly poorer levels of ego strength than thoseof either secure or resistant infants. Weber, Levitt, and Clark (1986) used the Dimensions ofTemperament Scale (DOTS) to measure maternal temperament and reported that mothers ofavoidant infants scored significantly higher on intensity of reaction than mothers of eithersecure or resistant babies. Mothers of resistant infants, however, did not differ from those ofsecure infants on any of the DOTS dimensions.

Most studies which have examined both maternal characteristics and infant temperamentduring the first year have relied exclusively on maternal reports of infant temperament. Farfewer have included observational measures of temperament. Those which have usedobservational measures have tended to focus on global measures of infant negative affect orproneness-to-distress. For example, Mangelsdorf, Gunnar, Kestenbaum, Lang and Andreas(1990) found that infant proneness-to-distress as assessed at 9 months, in the Louisvilletemperament assessment (Matheny and Wilson) and by maternal report, was associated withproneness-to-distress as assessed in the Strange Situation at 13 months, but not to overall

178 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

security of attachment. Thus, there was no main effect of temperament on attachmentsecurity.

Belsky (1996) demonstrated that acombinationof infant and paternal characteristics(including infant temperament and paternal personality) could be used to predict attachmentusing discriminant function analysis, and found that the collective predictive power of thecombination of variables was significantly better than chance, even when individual predic-tors were not. A similar approach was also used by Notaro and Volling (1999) in a recentpaper using parental behavior to predict infant-parent attachment classifications. Such anapproach has not yet been used to examine the antecedents of infant-mother attachment.

The purpose of this investigation was to examine the relations between maternal person-ality, maternal behavior, specific aspects of infant temperament, and quality of attachment.Specifically we were interested in how combinations of maternal and infant characteristicscould be used to predict attachment classifications. In this investigation different types ofnegative affect in infancy, specifically fearfulness and anger proneness, were examined,rather than global measures of proneness-to-distress. In addition, individual differences inpositive affectivity in infancy were also explored. It was predicted that higher levels offearfulness and anger-proneness would be related to higher rates of insecurity, and thathigher levels of infant positive affect would be associated with security. It was also predictedthat lower levels of maternal positive affectivity, reflecting depression, would be associatedwith insecurity. In addition, we wished to examine whether infant and maternal character-istics could be used to predict the reactivity groupings of A1-B2 and B3-C2 as noted by otherresearchers (e.g., Belsky & Rovine, 1987; Frodi & Thompson, 1985).

2. Method

2.1. Participants

One hundred and two mother-infant dyads were recruited through birth announcementspublished in Ann Arbor, Michigan and surrounding areas. Infants were 8 months (M 5 8months, 4 days old,SD5 10 days) at the first visit and 12 months (M 5 12 months, 4 days,SD 5 8 days) at the second visit. Of the original 102 subjects who completed the firstassessment, 6 subjects did not complete the second assessment, for a completion rate of 94%(N 5 96). In addition, one subject was sick during the second lab visit and was thus droppedfrom the study, leaving a total of 95 subjects who completed both assessments. Fifty-one(54%) of the infants were males, and 35 (37%) were first-born. All infants were healthyfull-terms. Average maternal age was 31.5 (SD5 5.0 years). All but one of the mothers hadcompleted at least a high school education, and 61 (68%) of the mothers had completed atleast a four-year college degree. At the 12 month visit, 59 (63%) of the mothers wereemployed at least part-time; these mothers reported working an average of 23.9 hr/week(SD 5 13.8).

179S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

2.2. Procedure

Subjects came to the laboratory twice, when the infants were 8 and 12 months old.Mothers were sent two questionnaires, Tellegen’s (1982) Multidimensional PersonalityQuestionnaire (MPQ), and Rothbart’s (1981) Infant Behavior Questionnaire (IBQ), to com-plete and return at their first laboratory visit. At each lab visit, subjects were first brought intoa waiting room where mothers completed a short demographic questionnaire. Then mothersand infants were shown into a testing room (4.57 m3 4.57 m) with a one-way mirror. Duringthe first visit, mothers and infants participated in a short teaching task and a laboratoryassessment of temperament (described below). When infants were 12 months, mothers andinfants were assessed in a standard measure of attachment, the Strange Situation.

2.3. Measures

2.3.1. Multidimensional Personality Questionnaire.Maternal personality was assessed usingTellegen’s Multidimensional Personality Questionnaire (MPQ, Tellegen, 1982). The MPQcontains 300 forced-choice items and is designed to reveal individual differences in person-ality and behavioral tendencies within a normal population. The items yield 11 primaryfactors and three super factors: Positive Affectivity, Negative Affectivity, and Constraint.These three super factors have been interpreted in temperamental terms (Goldsmith et al.,1994; Tellegen, 1985), reflecting pleasure, anger, and fear, respectively. According toDiLalla, Gottesman, & Carey (1993), Positive Affectivity reflects extraversion, enjoyment ofsocial relationships, and well-being while low scores on Positive Affectivity reflect depres-sion (Tellegen, 1985). Negative Affectivity reflects anxiety, and anger, and Constraintreflects self-control, rigidity, and conventionality.

The MPQ correlates well with other measures of personality and exhibits good test-retestreliability (0.82 to 0.92, over 30 days) and internal consistency (Cronbach’s alphas rangedfrom 0.83 to 0.89). The MPQ also has validity scales which allow the investigator to identifyinvalid records. All subjects’ scores on the validity scales were within the acceptable range(Tellegen, 1982).

2.3.2. Infant Behavior Questionnaire.Maternal perceptions of temperament were assessedusing Rothbart’s Infant Behavior Questionnaire (IBQ, Rothbart, 1981). The IBQ is a widelyused parent report measure of individual differences in infant temperament and emotionality.The IBQ consists of 94 items that ask the parents to indicate how often their infants engagedin particular behaviors in specific situations during the last week. Each item is scored on aseven-point scale ranging from “Never” to “Always,” with a “Does Not Apply” option (itemsmarked this way were not included when calculating scale scores). An example of a typicalIBQ items is, “How often during the last week did the baby squirm or kick during feeding?”For the purposes of this investigation the following IBQ scales were used in later analyses:Smiling and Laughter (positive affect), Distress and Latency to Approach of Sudden orNovel Stimuli (fear), Distress to Limitations (anger), and Activity Level.

180 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

2.3.3. Demographic Questionnaire.Mothers completed a questionnaire which asked basicdemographic information about age, education, working status, and child care arrangements.This information was obtained at both the 8 and 12 month assessments.

2.3.4. Mother-infant interaction.During the first laboratory assessment at 8 months mothersand infants were videotaped through a one-way mirror doing a structured task. Specifically,mothers were given a shape sorter and were instructed to “try to get your infant to put theblocks in the bucket.” A structured task, rather than an unstructured free play task was used,because in laboratory assessments it has been found that a task that is challenging to the child(such as a shape sorter for an 8 month old) may reveal greater variability in parent-childinteraction than a free play context (e.g., Sroufe et al., 1985). Following Fish, Stifter, andBelsky (1991) maternal behavior was coded in 30 s intervals for five minutes. Using scalesadapted from Fish et al. (1991) maternal sensitivity and intrusiveness/overcontrolling be-havior were rated on 4 point scales, describing the extent to which the parent was tuned intothe baby’s agenda versus the parents’ own agenda. High scores for sensitivity indicated awell-timed interaction that followed the infant’s lead and signals. High scores on intrusive-ness/overcontrolling indicated behavior that was overstimulating or ill-timed, and out of syncwith the infant’s signals. Parental positive and negative affect and facilitation of infantbehavior were coded on three point scales for each 30 s epoch (Fish, 1993). Positive affectassessed how much the mother smiled or laughed during the episode, Negative affectassessed the extent to which the mother showed negative affect (facial or vocal) or behavior,and Facilitation assessed the extent to which mothers facilitated the baby’s response to theshape sorter task. High scores on each scale indicated that the subjects were high on thatdimension. Interactions were coded by two independent coders and inter-rater reliability, forall scales except negative affect, ranged from 0.70 to 0.91 (averager 5 0.76). Due to verylow frequency of occurrence, negative affect was dropped from all further analyses. Asbehavior across the episodes was highly correlated, totals were created for each scale bysumming across the episodes. Based upon patterns of intercorrelations two composites ofmaternal behavior were created. Sensitivity and intrusiveness were negatively correlated(r 5 20.78,p , .001), thus the first composite, termedSensitivity,subtracted intrusivenessfrom sensitivity. Positive Affect and facilitation were correlated (r 5 0.30,p , .01) thus thesecond composite, calledExpressivity,was the sum of positiveaffect and facilitation. Thesematernal behavior composites were used in all subsequent analyses.

2.3.5. Laboratory Temperament Assessment Battery.Following the dyadic interaction, eachinfant participated in the prelocomotor version of Goldsmith and Rothbart’s (1988, 1991)Laboratory Temperament Assessment Battery (LABTAB). The LABTAB consists of a seriesof episodes in which the infant’s reaction to a specific stimulus is videotaped. Each stimulusattempts to elicit a specific emotional response from the infant. In this study we chose toexamine infants’ responses of anger, fear, and positive affect. There were three situationsdesigned to elicit each emotion, resulting in a total of nine episodes measuring temperament.Positive affect was elicited in a sound and light display (Display), a peek-a-boo game withmother (Peek-A-Boo), and a puppet game with a stranger (Puppet). Anger was elicited byrestraining the infants in a carseat (Restraint), placing a toy behind a barrier (Barrier), and

181S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

retracting a toy attached to a long pole (Retraction). Fear was elicited by showing threemasks (Masks), an approaching and unpredictable dinosaur (Dinosaur), and a sudden parasolopening (Parasol). Infants’ intensity of facial, vocal, and motoric responses to each stimuluswere rated in 5 s epochs according to LABTAB guidelines (see Goldsmith & Rothbart,1988). For each emotion, independent coders rated infant responses. Average inter-raterreliability for positive affect wasr 5 0.86 and for angerr 5 0.87. Four fear scales (parasolfacial and motoric responses, and dinosaur and masks motoric responses) did not achieveadequate inter-rater reliability were dropped from further analyses. For the remaining fearscales, average inter-rater reliability wasr 5 0.81.

As ratings across coding epochs were highly correlated, totals were created for each scaleby summing across epochs. Next, composites were created by standardizing scores and thensumming those correlated within each emotion. Three positive affect composites werecreated:Observed Display Positive Affect, Observed Peek-a-boo Positive Affect,and Ob-served Puppet Positive Affect.Two overall anger composites were created:Observed Re-straint Anger, Observed Toy Anger.Fear yielded one composite:Observed Fear.These sixLABTAB composites were used in all subsequent analyses.

2.3.6. Strange Situation.At 12 months infant-mother dyads were videotaped in Ainsworth’sStrange Situation (Ainsworth et al., 1978) which consists of eight 3-min episodes designedto be increasingly stressful for the infant. Based upon the infant’s proximity seeking, contactmaintenance, resistance, and avoidance of the mother following brief separations, infants areclassified as having one of three types of attachment relationships: Avoidant (A), Secure (B),or Resistant (C). All videotapes were coded by the first author (SM), who was trained byL. A. Sroufe. In addition, 22 tapes (25%) were coded by a second coder. The two codersagreed on 86% of the overall classifications, and Cohen’s kappa was 0.75. All disagreementswere resolved in favor of the senior coder (SM). Interrater reliability on the behavioral scales(averaged across both reunion episodes) was 87% on Proximity Seeking, 88% on ContactMaintaining, 90% on Avoidance 93% on Resistance. Three (3.2%) infants were judgedunclassifiable in the traditional Ainsworth classification system because they exhibited highlevels of both resistance and avoidance. These infants were dropped from analyses involvingthe traditional three group (A, B, C) classification system, but were included in analysescomparing secure and insecure groups. Of the 92 subjects who were classified using thetraditional three group coding system, 69.6% (N 5 64) were rated as secure, 13.0% (N 5 12)as avoidant, and 17.4% (N 5 16) as resistant. Others have reported similar distributions forcomparable samples (e.g., Belsky & Rovine, 1987; Notaro & Volling, 1999).

3. Results

The results section is divided into several sets of analyses. First, preliminary analysesexamined the associations between attachment security and the demographic variables thathave found to be of relevance to attachment, e.g., maternal employment and daycare.x2

analyses indicated that neither maternal employment nor child care status were related toattachment quality. Second, preliminary analyses examined the correlations among the

182 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

predictor variables and between the predictor variables and infants’ behavior during theStrange Situation are presented. Next, MANOVAs examining mean differences in predictorvariables by attachment classification are examined. Finally, in the central analyses we usedDiscriminant function analyses to from combinations of maternal and child characteristicspredicting attachment classification, security and reactivity follow.

3.1. Associations among predictor variables

The first set of analyses examined intercorrelations among predictor variables (see Table1). Maternal Negative Affectivity was modestly correlated with maternal ratings of temper-ament. Maternal temperament ratings of infant Distress to Novelty, Distress to Limits andActivity were moderately associated with one another. Maternal ratings of infant fearfulnesswere negatively associated with observations of positive affectivity in the laboratory. Angerobserved in the Restraint situation was negatively related to positive affect observed in thelaboratory.

3.2. Associations among predictor variables and Strange Situation behavior

Correlations were also conducted to examine the relations between predictor variables andbehavior during the Strange Situation. Maternal Facilitation was associated with greaterproximity seeking to mother during the second reunion,r (90) 5 29, p , 0.01. MaternalNegative Affectivity was not related to infants’ behavior during the Strange Situation.However, greater maternal Constraint was correlated with lower levels of Resistance duringthe reunion episodes (r (94) 5 0.30,p , 0.05; r (94)5 20.29,p , 0.01 for the first andsecond reunions respectively). Higher maternal Positive Affectivity was related to morecrying during the first separation episode,r (93) 5 0.23,p , 0.05. Maternal perceptions ofhigher Smiling and Laughter were associated with greater infant Proximity Seeking tomother during the first reunion,r (94) 5 0.32,p , 0.01, and to greater Contact Maintenanceduring the first,r (93) 5 0.21, p , 0.05, and second reunions,r (93) 5 0.21, p , 0.05.Maternal perceptions of infant Distress to Limitations were associated with greater Resistanceduring the second reunion,r (93) 5 0.22, p , 0.05. Observed infant anger proneness andfearfulness in the laboratory were not related to behavior during the Strange Situation. Observedinfant Display Positive Affect was related to greater Contact Maintenance,r (91) 5 0.24,p ,0.05, and to greater crying,r (90) 5 0.26,p , 0.05, during the second reunion episode.

3.3. Exploring main effects of maternal and infant characteristics

Next, four (maternal behavior, maternal personality, maternal ratings of infant tempera-ment and observed infant temperament) multivariate analyses of variance (MANOVAs) wereconducted to determine whether the predictor variables differed as a function of attachmentclassification. For each of these analyses, two planned contrasts were conducted: secureversus insecure, and insecure-resistant versus insecure-avoidant. None of the overall MANO-VAs were significant, although several of the planned contrasts were significant. The plannedcontrasts showed that: a) mothers of secure infants perceived their infants to be lower on

183S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

Tab

le1

Inte

rcor

rela

tions

ofpr

edic

tor

varia

bles

Var

iabl

e1

23

45

67

89

1011

1213

1415

Mat

erna

lbeh

avio

r:1.

Sen

sitiv

ity—

2.E

xpre

ssiv

ity.2

8**

—M

ater

nalp

erso

nalit

y:3.

Pos

itive

affe

ctiv

ity.2

1*.1

6—

4.N

egat

ive

affe

ctiv

ity2

.10

.12

.16

—5.

Con

stra

int

2.0

8.0

82

.24*

.03

—M

ater

nalI

BQ

ratin

gs:

6.S

mili

ng&

laug

hter

2.1

42

.03

.12

.05

2.0

3—

7.D

istr

ess

tono

velty

2.1

22

.04

.181

.25*

.00

2.1

3—

8.D

istr

ess

tolim

its.1

2.1

4.2

1*.3

1**

.06

2.0

5.3

7***

—9.

Act

ivity

.09

.14

.09

.24*

2.0

6.0

4.3

2**

.42*

**—

Obs

erve

dte

mpe

ram

ent:

10.

Res

trai

ntan

ger

.15

2.0

2.1

02

.28*

*.0

92

.15

.07

.181

.12

—11

.T

oyan

ger

2.0

3.0

7.0

22

.08

.03

2.0

42

.04

2.0

3.1

7.3

2**

—12

.F

ear

.201

2.0

22

.02

.01

.05

2.0

4.1

2.1

5.0

9.1

32

.03

—13

.D

ispl

aypo

sitiv

eaf

fect

2.2

6*2

.05

.04

.04

2.0

8.1

2.0

02

.17

2.0

42

.25*

2.0

52

.23*

—14

.P

uppe

tpo

sitiv

eaf

fect

.05

.042

.12

.04

.09

.16

2.1

72

.05

.06

2.3

0**

.06

2.0

5.0

3—

15.

Pea

k-a-

boot

pos.

aff.

.12

2.0

12

.10

2.1

22

.181

2.0

12

.23*

2.0

4.0

02

.27*

.06

2.1

3.2

7*.1

91—

1p.

,0

.1;

*p.

,.0

5;

**p.

,.0

1;

***

p.,

.00

1.

184 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

Distress to Novelty than mothers of insecure infants; b) mothers of avoidant infants werelower on Positive Affectivity than were mothers of resistant infants; and c) avoidant infantswere higher on observed fear than were resistant infants. In order to further explore thePositive Affectivity finding, a MANOVA was conducted on the four subscales (well-being,social potency, social closeness, and achievement) that comprised the Positive Affectivityscale to determine which subscales differed as a function of attachment classification. Theoverall MANOVA was significant,F(8,174)5 2.12,p , 0.05. Planned contrasts indicatedthat mothers of secure infants were higher on Well-Being than were mothers of insecureinfants, t(2,89) 5 2.10,p , 0.05. Planned contrasts also showed that mothers of resistantinfants were higher on Achievement than were mothers of avoidant infants,t(2,89)5 2.09,p , 0.05. Given that the Achievement subscale was the only subscale which showed asignificant contrast between avoidant and resistant infants, it can be inferred that theAchievement subscale was responsible for the difference between mothers of avoidant andresistant infants on Positive Affectivity reported earlier. Means and standard deviations ofeach attachment group on the predictor variables and the results of the planned comparisonsare presented in Table 2.

3.4. Exploring maternal and child characteristics collectively

In order to examine whether infant attachment classification could be collectively pre-dicted by maternal and infant characteristics at eight months, discriminant function analyses

Table 2Means and standard deviations of predictor variables by attachment classification

Insecureavoidant

Secure Insecureresistant

Secure vs.insecuret-value

Avoidant vs.resistantt-value

Maternal behavior:Sensitivity 9.83 (10.58) 10.34 (14.55) 10.60 (7.85) ns nsExpressivity 18.83 (7.57) 19.29 (7.07) 19.07 (7.72) ns ns

Maternal personality:Positive affectivity 146.17 (8.83) 153.91 (10.86) 157.00 (11.63) ns 2.63**Negative affectivity 124.58 (11.99) 123.16 (11.52) 123.38 (8.83) ns nsConstraint 166.50 (12.33) 166.41 (11.28) 159.75 (14.51) ns ns

Maternal IBQ ratings:Smiling & laughter 4.51 (0.76) 5.03 (0.79) 4.99 (0.85) ns nsDistress to novelty 3.22 (1.16) 2.70 (0.81) 2.97 (0.57) 2.14* nsDistress to limitations 3.19 (0.61) 3.45 (0.79) 3.71 (0.76) ns 1.741Activity 4.37 (0.98) 4.18 (0.86) 4.56 (0.60) ns ns

Observed temperament:Restraint anger 20.21 (0.97) 20.05 (0.86) 0.31 (0.80) ns nsToy anger 20.20 (0.45) 20.04 (0.58) 0.05 (0.91) ns nsFear 0.39 (1.17) 20.01 (0.58) 20.16 (0.47) ns 2.09*Display positive affect 20.14 (0.71) 0.09 (0.93) 0.03 (0.83) ns nsPuppet positive affect 0.23 (0.74) 0.02 (0.82) 20.15 (0.82) ns nsPeak-a-boo positive affect 0.09 (0.81) 0.08 (0.81) 20.22 (0.70) ns ns

1 p. , .10; * p. , .05; **p. , .01.

185S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

were conducted. Discriminant function analysis can be used to classify subjects into groupsbased upon various discriminating variables (Huberty, 1984). In this study, we attempted topredict which subjects would be classified as secure, insecure/avoidant, or insecure/resistantbased upon maternal personality and behavior variables, maternal perceptions of infanttemperament, and laboratory measures of infant temperament. We then compared the “hitrate” of those predictions with the subjects’ actual attachment classification as measured inthe Strange Situation at twelve months. The z statistic tests whether the number of subjectscorrectly classified is greater than would be correctly classified by chance, and the I statisticindicates to what extent the obtained classification rate is better than that expected by chance(Huberty, 1984).

Three separate discriminant analyses were conducted to predict attachment classifications(ABC), security/insecurity, and high/low reactivity groups from the predictor variables. Eachof these analyses will be discussed in order. For each of these three analyses first thepredicted vs. actual group membership results are presented (see Table 3). Thereafter, wepresent the discriminant functions for each of these analyses (see Table 4).

3.4.1. Predicting avoidance, resistance and security.As shown in Table 3, the greatestpercentage of subjects (78%) were correctly classified in the ABC analysis. In this analysis,security was correctly predicted in 94% of the cases, whereas subjects were correctlyclassified as A’s 58% of the time, and as C’s 29% of the time. Resistant infants were neverincorrectly classified as avoidant infants, and avoidant infants were never incorrectly clas-sified as resistant infants. This analysis was significant at thep , 0.001 level and correctlyclassified subjects at a rate of 53% greater than chance. The large percentage of subjectscorrectly classified indicates that attachment security can be successfully predicted by acombination of maternal and infant characteristics.

Table 3Summary of discriminant function analyses classification tables

Actual group Predicted group membership

Avoidant Secure Resistant

Avoidant 7 (58%) 5 (42%) 0 (0%)Secure 2 (3%) 58 (94%) 2 (3%)Resistant 0 (0%) 10 (71%) 4 (29%)Percent of cases correctly classified: 78%Percent improvement over chance5 53% Z 5 4.59,p. , .0001

Insecure SecureInsecure 8 (28%) 21 (72%)Secure 4 (6.5%) 58 (93.5%)Percent of cases correctly classified: 73%Percent improvement over chance5 37% Z 5 3.07,p. , .001

Low reactive High reactiveLow reactive (A1, A2, B1, B2) 14 (39%) 22 (61%)High reactive (B3, B4, C1, C2) 8 (15%) 44 (85%)Percent of cases correctly classified: 66%Percent improvement over chance5 29% Z 5 2.68,p. , .01

186 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

In order to interpret the functions that discriminated the groups in each analysis, corre-lations between the discriminating variables and the canonical discriminant functions wereexamined (see Table 4). For the ABC analysis, maternal and infant positive affectivitycorrelated strongly and positively with function one, whereas maternal ratings of andobserved fearfulness correlated negatively with function one. Thus, this function seems toreflect high maternal positive affectivity and maternal perceptions of infant positive affectand low infant fearfulness. Examination of the correlations between the discriminatingvariables and the second function indicates that the second function reflects high maternalConstraint, low infant Activity, and low reported Distress to Novelty. In order to examinehow the functions discriminated between the attachment groups, mean discriminant scoreson the functions for each attachment group were calculated. Multivariate analyses ofvariance were used to examine whether the means varied as a function of attachmentclassification, with two planned contrasts, secure versus insecure infants and avoidant versusresistant infants. The MANOVA was significant (F(4, 170)5 11.07,p , 0.001), as were theplanned contrasts between secure and insecure infants and between avoidant and resistantinfants. The secure/insecure contrast showed that secure infants were significantly higher onfunction one than insecure infants,t(2,85) 5 4.15, p , 0.001. Thus, infants classified assecure in the Strange Situation were higher on positive affect and lower on fearfulness andhad mothers higher on positive affectivity than insecurely attached infants. In addition,resistant infants were significantly higher than avoidant infants on function one, indicatingthat they were higher on reported positive affect and lower on reported and observedfearfulness than avoidant infants,t(2,85) 5 5.85, p , 0.0001. The planned contrasts also

Table 4Correlations between discriminating variables and canonical discriminant functions for discriminant analyses

Discriminating variables ABC classification Secure/insecureclassification

Hi/lo reactiveclassification

Function 1 Function 2 Function 1 Function 1

Maternal behavior:Sensitivity .03 .02 2.03 .07Expressivity .02 .06 .11 2.09

Maternal personality:Positive affectivity .45 2.08 .20 .48Negative affectivity 2.05 2.12 2.05 2.19Constraint 2.17 .61 .34 2.32

Maternal IBQ ratings:Smiling & laughter .36 .09 .42 .74Distress to novelty 2.25 2.39 2.37 2.09Distress to limitations .24 2.16 .05 .05Activity .00 2.46 2.34 .15

Observed temperament:Restraint anger .19 2.26 2.11 .08Toy anger .16 .09 .07 2.03Fear 2.33 .01 2.22 2.11Display positive affect .11 2.03 .00 .28Puppet positive affect 2.20 .03 2.15 .06Peak-a-boo positive affect .11 .29 .23 .12

187S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

revealed that secure infants were higher than insecure infants on function two, indicating thatsecurely attached infants were lower on activity and distress-to-novelty and had mothers whowere higher on Constraint than insecurely attached infants,t(2,85)5 2.47,p , 0.05.

3.4.2. Predicting security vs. insecurity.For the discriminant function analysis predictingsecurity vs. insecurity, 73% of the subjects were correctly classified, a rate of 37% greaterthan chance, which was lower than the correct classification rate for the ABC analysis. Thecorrelations between the predictor variables and the function suggest that the discriminantfunction reflects high maternal perceptions of infant Smiling and Laughter and low percep-tions of Distress to Novelty and Activity and high maternal Constraint. (see Table 4, column3). An analysis of variance examining mean differences on the function scores by attachmentsecurity revealed that securely attached infants had mothers higher on Constraint andmothers who perceived their infants to be higher on Smiling and Laughter and lower onActivity and fearfulness,F(1,89)5 14.90,p , 0.001.

3.4.3. Predicting reactivity.The final discriminant analysis involved predicting Belsky andRovine’s (1988) reactivity groups by categorizing A1, A2, B1 and B2 infants as low reactiveand B3, B4, C1 and C2 as high reactive. Infants were correctly classified as high reactive85% of the time, and correctly classified as low reactive 39% of the time, for an overall “hitrate” of 66% (see Table 3). Inspection of the correlations of the discriminating variables withthe discriminant function indicate that the discriminant function reflects high maternal andinfant Positive Affectivity and low maternal Constraint (see Table 4, column 4). An analysisof variance on mean discriminant function scores showed that the high reactive group washigher on this function than the low reactive group,F(1,86)5 13.98,p , 0.001. That is, highreactive infants were reported to be higher on positive affect and had mothers who werehigher on Positive Affectivity and lower on Constraint than infants classified as low reactive.

4. Discussion

The results of this study highlight the importance of examining both child characteristicsand parent characteristics in the development of attachment relationships. As was predicted,both parent and child characteristics examined collectively served to discriminate among thethree attachment groups, although there were few main effect of maternal or child charac-teristics on attachment quality. The results of this investigation suggest that any individualcharacteristic of either child or mother may be less important than the relationship contextwithin which that characteristic occurs. That is, attachment security seems to be betterpredicted by examining a combination of maternal and child characteristics, rather thanfocusing on individual characteristics in isolation. Such findings are consistent with atransactional approach to the study of early parent-child relationships, (e.g., Crockenberg,1986; Sameroff & Fiese, 1990), in which a developmental outcome, in this case attachmentsecurity, in seen as neither a function of the individual alone, nor the experiential context(parenting environment) alone. Specifically, the findings of this investigation revealed thatinfants classified as secure in the Strange Situation were higher on positive affect and lower

188 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

on fearfulness and had mothers higher on positive affect than insecurely attached infants.Likewise, insecurely attached infants were found to be higher on activity and distress tonovelty and had mothers who were lower on Constraint than securely attached infants. Thefinding that insecurely attached infants may be higher on activity replicates Calkins and Fox(1992) finding, although in this investigation the activity level difference only emerges as asignificant difference between secure and insecure infants in combination with other infantand maternal characteristics. It is possible that temperamental characteristics only place achild at risk for developing an insecure relationship when they occur in conjunction withother temperamental characteristics, and characteristics of the caregiver. For example, theresults of the current investigation suggest that an active and fearful infant who has motherwho is impulsive and takes risks (low on constraint) may be more at risk for developing aninsecure attachment than a child with only one of these characteristics.

4.1. Predicting avoidance and resistance

The results of this study also indicate that child temperamental and maternal antecedentsof the two patterns of insecure attachment are quite distinct. In fact, the predictive power ofthe analysis predicting security vs. insecurity was lower than the analysis predicting all threegroups. This supports a basic theoretical assumption of attachment theory that patterns ofinsecurity have distinct correlates in the history of interaction between parents and child(e.g., Ainsworth et al., 1978). Many studies have failed to support this assumption, in partdue to inherent problems with statistical power because of the small ratio of avoidant andresistant dyads found in normative samples. To deal with these problems researchers haveoften combined avoidant and resistant dyads into a single insecure groups. The results of thisstudy indicate that that practice may in fact hamper attempts to understand the antecedentsof attachment. The results of this study are consistent with studies of mother child interaction(Isabella, Belsky & von Eye, 1989; Isabella & Belsky, 1991; Isabella, 1993) which havenoted different antecedent patterns of maternal characteristics for the three major attachmentpatterns, and are also consistent with the suggestions made by temperament researchers (e.g.,Kagan, 1982; Buss & Plomin, 1986) that different temperamental antecedents may underlinethe three attachment patterns.

One particularly noteworthy finding of the current investigation is that variations inspecific emotions were predictive of attachment. Although many investigations have exam-ined how infant negative emotionality is related to secure/insecure attachment, far fewerhave examined how specific types of negative affect,. e.g., anger-proneness vs. fearfulnessmay relate to attachment outcomes. In this investigation fearfulness and anger-pronenesswere differentially predictive of attachment outcomes, highlighting the importance of ex-amining these emotions separately. Likewise, although there has been much focus onnegative affect as an antecedent of attachment, far less attention has been paid to individualdifferences in positive affect. The results of this investigation indicate that infant andmaternal positive affect are also important antecedents of attachment as suggested byMalatesta and her colleagues (Malatesta et al., 1989).

189S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

4.2. Predicting reactivity

Surprisingly the analysis predicting high and low reactive groups was the one with thelowest predictive power; the reactivity groupings were accurately predicted only 66% of thetime. However, the prediction of high reactivity was far more accurate (85% of the time) thanof low reactivity (39%). High reactivity was predicted by a combination of high maternal andchild positive affect, and low maternal constraint. It may be that the low reactive group is aheterogeneous group with a variety of different temperamental and maternal antecedentswhich we were not able to detect by combining them.

4.3. Examination of individual characteristics

There were a few significant differences among the three attachment groups when parentand child characteristics were examined individually. For example, mothers of resistantinfants were higher on Positive Affect than mothers of avoidant infants. Thus, as predicted,there was an association between maternal positive affect and insecurity, but this associationwas for a specific type of insecurity, avoidance, and not resistance. Mothers of secure infantsscored more similarly to the mothers of resistant infants than to the mothers of avoidantinfants, but were not significantly different from either group. Follow-up analyses exploringwhich scales of the Positive Affect Factor were responsible for these differences between thegroups indicated that mothers of secure infants scored higher on Well-being than mothers ofinsecure infants. According to Tellegen (1982), individuals scoring high on Well-being havehappy cheerful dispositions and feel good about themselves, in contrast to low scorers, whoare “seldom really happy.” Well-being is primarily (and negatively) related to depression(Hall, 1977; Tellegen, 1982). Thus the findings of this investigation indicate that mothers ofsecurely attached infants are less likely to be depressed than the mothers of insecurelyattached infants. This finding is consistent with other investigations which have explored theassociation between maternal depression and security of attachment (e.g., Teti et al., 1995).In this investigation mothers of resistant infants scored higher on achievement than mothersof avoidant infants. Individuals who score high on achievement “like to work hard and putwork and accomplishments before many other things, and are perfectionists” (Tellegen,1982). The achievement scale includes items such as, “I set extremely high standards formyself in my work.” It is easy to imagine how such an orientation, putting work before otherthings, perhaps before parenting, might not be conducive to the development of a secureattachment relationship with one’s child. It is not clear why this dimension of maternalpersonality is specifically related to resistant rather to avoidant attachment. However,according to Ainsworth (e.g., Ainsworth et al., 1978) mothers of insecure resistant infants areinconsistent in their sensitivity and availability. It is possible that a mother with highstandards for her performance in her work might be more likely to inconsistently sensitivein her sensitivity and availability as a function of her demands at work. Although consid-erable attention has been paid to the role of maternal employment in the development ofinsecure attachment (e.g., Belsky & Rovine, 1988; NICHD Early Child Care Network, 1997)less attention has been paid to how maternal employment may interact with maternal

190 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

personality in the prediction of insecure attachment. Future research should explore suchlinkages.

Examination of maternal reports of infant temperament indicated that infants whosemothers’ rated them as more distressed to novelty or fearful at 8 months of age were morelikely to be insecurely attached at 12 months of age. Thus as predicted, fearfulness wasrelated in insecurity. Similarly there was a trend for infants whose mothers rated them ashigher on distress to limitations, or anger at 8 months to be more likely to be resistant thanavoidant. Both fearfulness and anger-proneness may both be components of what haspreviously been described as proneness-to-distress. Thus, these findings are consistent withprevious investigations which have noted that higher irritability is associated with resistantattachment (Miyake, Chen & Campos, 1985; Seifer et al., 1996; Susman et al., 1996) andwith insecure attachment in general, both avoidance and resistance (Calkins & Fox, 1992;van den Boom, 1989; 1994).

Temperament, both as reported by mothers and as observed in the laboratory, was relatedto behavior within the Strange Situation. For example, infants rated as higher on anger weremore likely to display resistance in the Strange Situation. This finding is consistent withprevious research in which associations between proneness-to-distress and resistance havebeen noted (Goldsmith & Alansky, 1988; Susman-Stillman et al., 1996).

4.4. Associations between maternal report and laboratory assessments of temperament

Surprisingly, there were few associations between temperament observed in the laboratoryand maternal reports of temperament. However, interestingly maternal reports of infantdistress to novelty were associated with behavior in the first episode in the laboratoryassessment. Distress to novelty may be most easily detected at the outset of a laboratoryassessment, when the situation is truly novel (Nachmias et al., 1996). Other investigationsusing observers’ ratings of temperament have also sometimes failed to reveal associationsbetween parental reports and observations of temperament (e.g., Seifer et al., 1994). Thislack of association between parents’ reports and laboratory assessments may be because theframe of reference or the contexts in which the behaviors are observed are so different. Forexample, parents are basing their ratings across months of observing their own child in avariety of different contexts, whereas laboratory assessments are observations of behavior oflarge numbers of children in specified contexts. Results of other investigations reveal thatobserver-parent convergence in ratings is higher if the context is clearly specified (e.g.,Bridges et al., 1993; Hagekull, Bohlin, & Lindhagen, 1984; Mangelsdorf, 1992). Interest-ingly, although there was little convergence between maternal reports and laboratory obser-vations of temperament in this investigation, both were predictive of behavior at three yearsof age (Mangelsdorf et al., 1994). Such findings suggest that maternal reports and laboratoryassessments may both provide important, although often nonoverlapping, information re-garding infant behavior.

It should also be noted that maternal negative affectivity was related to maternal ratingsof infant temperament. Specifically, maternal negative affect was modestly associated withDistress to Novelty, Distress to Limits (Anger), and to Activity. Similar associations havebeen noted in other research (e.g., Diener, Goldstein & Mangelsdorf, 1995; Mangelsdorf et

191S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

al., 1990; Vaughn et al., 1987). Such associations may be caused by maternal bias, mothershigh on negative affect perceiving everything, including their infants, in a more negativelight, and/or by objective differences in infants, e.g., anxious mothers may indeed have morefearful babies.

4.5. Maternal behavior and attachment

Maternal behavior assessed in the laboratory was associated with behavior within theStrange Situation, with greater maternal expressivity being associated with greater proximityseeking to mothers in the Strange Situation. Also, maternal self-report ratings of positiveaffect were associated with higher observed sensitivity in the laboratory. However, maternalbehavior was not associated with overall Strange Situation classifications. Seifer et al. (1996)point out that the sensitivity/Strange Situation association has been noted in some investi-gations and not in others, suggesting that it is a significant, but modest association. Thus bychance one would expect it to be detected in some investigations, while not in others. But,when researchers look across studies significant effects are noted (De Wolff & van Ijzend-dorn, 1997). It also seems that the sensitivity attachment security association might be harderto detect when briefer laboratory assessments are used rather than lengthy home observa-tions. However, in previous investigations researchers have found home observations andlaboratory observations of maternal behavior to be significantly correlated (e.g., Mangelsdorfet al., 1990), thus we had chosen in this investigation to use only the laboratory assessment.Furthermore, in a recent meta-analysis of parental antecedents of infant attachment, DeWolff and van Ijzendoorn (1997) found that the duration of home observations was notrelated to the magnitude of the association between sensitivity and attachment, nor was thereany effect of home versus laboratory observations. It also may be that if we had assessedmaternal sensitivity earlier in infancy the sensitivity-security association may have beendetected. Isabella (1993) has noted that observations of sensitivity may be most robust duringearly infancy when patterns of interaction are being established, rather than later in infancy.Thus, observations of mother-infant conducted earlier in infancy may have been moreinformative.

Other investigators have highlighted the importance of social support for maternal sen-sitivity (Goldstein, Diener & Mangelsdorf, 1996) and security of attachment (Crockenberg,1981). Likewise marital quality has also been identified as an important predictor of thequality of attachment relationships (e.g., Belsky, 1996; Frosch, Mangelsdorf & McHale,2000). It is likely that the predictive power of the analyses reported in this paper might havebeen even more robust if such familial contextual variables had been included.

In sum, the results of this investigation clearly highlight the importance of examiningmultiple determinants of attachment security by examiningboth child and maternal contri-butions. Thus perhaps we should lay to rest whether parent or child characteristics are moreimportant for relationship outcomes, but rather acknowledge that both are important. Thisinvestigation also suggests that conceiving of child characteristics as variations in specificemotions (Izard et al., 1991) including positive emotions, rather than merely focusing onmore global measures of proneness-to-distress or irritability, may help us to further elucidateattachment processes.

192 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

Acknowledgments

This research was supported by grants from the Office of the Vice-President for Researchat the University of Michigan, and from the Research Board at the University of Illinois. Theauthors would like to thank all those who helped in the research including Kristin Boelcke,Douglas Clingan, Lisa Pillola, Mitchell Shapiro and Kendra Ward, and thanks to CarolynAnderson and Margaret Fish who provided statistical advice. Thanks also to Cynthia Froschand Karl Rosengren who read and commented on earlier drafts of this paper.

References

Ainsworth, M. D. S., Blehar, M., Waters, E., & Wall, S. (1978).Patterns of attachment: a psychological studyof the Strange Situation.Hillsdale, NJ: Erlbaum.

Bates, J., Maslin, C. A., & Frankel, K. A. (1985). Attachment security, mother-child interaction, and temperamentas predictors of behavior problem ratings at age three years. In I. Bretherton & E. Waters (Eds.),Growingpoints of attachment theory and research(pp. 167–193).Monographs of the Society for Research in ChildDevelopment, 50,(1–2, Serial No. 209).

Belsky, J. (1997). Theory testing, effect-size evaluation, and differential susceptibility to rearing influences: thecase of mothering and attachment.Child Development, 64,598–600.

Belsky, J., & Isabella, R. (1988). Maternal, infant, and social-contextual determinants of attachment security. InJ. Belsky & R. Isabella (Eds.),Clinical implications of attachment(p. 41–94). Hillsdale, NJ: Erlbaum.

Belsky, J., & Rovine, M. (1987). Temperament and attachment security in the Strange Situation: an empiricalrapprochement.Child Development, 58,787–795.

Belsky, J., & Rovine, M. (1988). Nonmaternal care in the first year of life and the security of infant-parentattachment.Child Development, 59,157–167.

Belsky, J., Rovine, M., & Taylor, D. G. (1984). The Pennsylvania infant and family development project, III. Theorigins of individual differences in infant-mother attachment: maternal and infant contributions.ChildDevelopment, 55,718–728.

Bretherton, I. (1985). Attachment theory: retrospect and prospect. In I. Bretherton & E. Waters (Eds.),Growingpoints of attachment theory and research(pp. 3–35).Monographs of the Society for Research in ChildDevelopment,50, (1–2, Serial No. 209).

Bridges, L. J., Palmer, S. A., Morales, M., Hurtado, M., & Tsai, D. (1993).Infant Behavior and Development,16, 501–506.

Buss, A. H., & Plomin, R. (1986). The EAS approach to temperament. In R. Plomin, & J. Dunn (Eds.),The studyof temperament: changes, continuities, and challenges(pp. 67–79). Hillsdale, New Jersey: Lawrence ErlbaumAssociates.

Calkins, S. D., & Fox, N. A. (1992). The relations among infant temperament, security of attachment, andbehavioral inhibition at 24 months.Child Development, 63,1456–1472.

Crockenberg, S. B. (1981). Infant irritability, mother responsiveness, and social support influences on the securityof infant-mother attachment.Child Development, 52,857–867.

Crockenberg, S. B. (1986). Are temperamental differences in babies associated with predictable differences incare giving? In J. V. Lerner, & R. M. Lerner (Eds.),Temperament and social interaction during finance andchildhood(pp. 53–73).New Directions for Child Development,No. 31. San Francisco: Jossey-Bass.

DeMulder, E. K., & Radke-Yarrow, M. (1991). Attachment with affectively ill and well mothers: concurrentbehavioral correlates.Development and Psychopathology, 3,227–242.

De Wolff, M. S., & van Ijzendoorn, M. H. (1997). Sensitivity and attachment: a meta-analysis on parentalantecedents of attachment.Child Development, 68,571–591.

Diener, M., Goldstein, L., & Mangelsdorf, S. (1995). The role of prenatal expectations in parents’ reports of infanttemperament.Merrill Palmer Quarterly, 41(2), 172–190.

193S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

DiLalla, D. L., Gottesman, I. I., & Carey, G. (1993). Assessment of normal personality traits in a psychiatricsample: dimensions and categories. In L. J. Chapman, J. P. Chapman, & D. C. Fowles (Eds.),Progress inexperimental personality and psychopathology research, vol. 16(pp. 137–162). New York: Springer.

Egeland, B., & Farber, E. (1984). Infant-mother attachment: factors related to its development and changes overtime. Child Development, 55,753–771.

Fish, M. (1993).Maternal behavior scales.Unpublished scales.Fish, M., Stifter, C. A., & Belsky, J. (1991). Conditions of continuity and discontinuity in infant negative

emotionality: newborn to five months.Child Development, 62,1525–1537.Fox, N. A. (1992). The role of individual difference in infant personality in the formation of attachment

relationships. In E. J. Susman, L. V. Feagaons, & W. J. Ray (Eds.),Emotions, cognition, health anddevelopment in children and adolescents(pp. 31–52). Hillsdale, NJ: Erlbaum.

Fox, N. A., Kimmerly, N. L., & Schafer, W. D. (1991). Attachment to mother/attachment to father: a meta-analysis.Child Development, 62,210–225.

Frodi, A., & Thompson, R. (1985). Infants’ affective responses in the Strange Situation: effects of prematurityand of quality of attachment.Child Development, 56,1280–1290.

Frosch, C. A., Mangelsdorf, S. C., & McHale, J. L. (2000). Marital behavior and the security of preschooler-parent attachment relationships.Journal of Family Psychology, 14,144–161.

Goldsmith, H. H., & Alansky, J. A. (1987). Maternal and infant temperamental predictors of attachment: ameta-analytic review.Journal of Consulting and Clinical Psychology, 55,805–816.

Goldsmith, H. H., Bradshaw, D. L., & Reiser-Danner, L. A. (1986). Temperament as a potential developmentalinfluence on attachment. In J. V. Lerner & R. M. Lerner (Eds.),Temperament and social interaction duringinfancy and childhood. New Directions for Child Development, 31,5–34. San Francisco: Jossey-Bass.

Goldsmith, H. H., Losoya, S. H., Bradshaw, D. L., & Campos, J. J. (1994). The genetics of personality: a twinstudy of the five factor model and parental offspring analyses. In C. Halverson, R. Martin, and G. Kohnstamn(Eds.),The developing structure of temperament and personality from infancy to adulthood.Hillsdale, NJ:Erlbaum.

Goldsmith, H. H., & Rothbart, M. K. (1988). The Laboratory Temperament Assessment Battery (LAB-TAB):Locomotor Version, Edition 1.2. Oregon Center for the Study of Emotion technical report #88–01.

Goldsmith, H. H., & Rothbart, M. K. (1991). Contemporary instruments for assessing early temperament byquestionnaire and in the laboratory. In J. Strelau & A. Angleitner (Eds.),Explorations in temperament:international perspectives on theory and measurement(pp. 249–272). New York: Plenum.

Goldstein, L. H., Diener, M. L., & Mangelsdorf, S. C. (1996). Maternal characteristics and social support acrossthe transition to motherhood: associations with maternal behavior.Journal of Family Psychology, 10,60–71.

Gunnar, M. R., Mangelsdorf, S. C., Larson, M., & Hertsgaard, L. (1989). Attachment, temperament, andadrenocortical activity in infancy.Developmental Psychology, 25,355–363.

Huberty, C. J. (1984). Issues in the use and interpretation of discriminant analysis.Psychological Bulletin, 95,156–171.

Isabella, R. A. (1993). Origins of attachment: maternal interactive behavior across the first year.Child Devel-opment, 64,605–621.

Isabella, R. A., & Belsky, J. (1991). Interactional synchrony and the origins of infant-mother attachment: areplication study.Child Development, 62,373–384.

Izard, C. E., Haynes, O. M., Chisholm, G., & Bask, K. (1991). Emotional determinants of infant-motherattachment.Child Development, 62,906–917.

Kagan, J. (1982).Psychological research on the human infant: an evaluative summary.New York: W.T. GrantFoundation.

Kochanska, G. (1998). Mother-child relationship, child fearfulness, and emerging attachment: a short-termlongitudinal study.Developmental Psychology, 34,480–490.

Malatesta, C. Z., Culver, C., Tesman, J. R., Shepard, B. (1989). The development of emotion expression duringthe first two years of life.Monographs of the Society for Research in Child Development, 54,(1–2, Serial No.219).

Mangelsdorf, S., Diener, M., & Boelcke, K. (June, 1994).Individual differences in emotional expressiveness:

194 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

continuities from 8 months to 3 years.Paper presented at a symposium at the International Conference ofInfant Studies, Paris, France.

Mangelsdorf, S., Gunnar, M., Kestenbaum, R., Lang, S., & Andreas, D. (1990). Infant proneness-to-distresstemperament, maternal personality, and mother-infant attachment: associations and goodness-of-fit.ChildDevelopment, 61,820–831.

Mangelsdorf, S. (1992). Developmental changes in infant-stranger interaction.Infant Behavior and Development,15, 191–208.

Miyake, K., Chen, S., & Campos, J. J. (1985). Infant temperament, mother’s mode of interaction, and attachmentin Japan: an interim report. In I. Bretherton & E. Waters (Eds.),Growing points of attachment theory andresearch(pp. 276–297).Monographs of the Society for Research in Child Development, 50,(1–2, Serial No.209).

Moran, G., & Pederson, D. R. (April, 1997).Proneness to distress and ambivalent relationships observed in thehome.Paper presented at the biennial meeting of the Society for Research in Child Development, Washington,DC.

Nachmias, M., Gunnar, M., Mangelsdorf, S. C., Parritz, R., & Buss, K. (1996). Behavioral inhibition and stressreactivity: the moderating role of attachment security.Child Development, 67,508–522.

NICHD Early Child Care Research Network (1997). The effects of infant child care on infant-mother attachmentsecurity: results of the NICHD study of early child care.Child Development, 68,860–879.

Notaro, P. C., & Volling, B. L. (1999). Parental responsiveness and infant-parent attachment: a replication studywith fathers and mothers.Infant behavior and development, 22(3), 345–352.

Radke-Yarrow, M., Cummings, E. M., Kuczynski, L., & Chapman, M. (1985). Patterns of attachment in two- andthree-year-olds in normal families and families with parental depression.Child Development, 56,884–892.

Rothbart, M. K. (1981). Measurement of temperament in infancy.Child Development, 52,569–578.Rothbart, M., & Bates, J. E. (1997). Temperament. In W. Damon (Series Ed.) & N. Eisenberg (Vol. Ed.),

Handbook of child psychology: vol. 3. social, emotional, and personality development(5th ed.). New York:Wiley.

Sameroff, A., & Chandler, M. J. (1975). Reproductive risk and the continuum of caretaking casualty. In F. D.Horowitz, M. Hetherington, S. Scarr-Salapatek, & G. Siegel (Eds.),Review of child development research(Vol. 4, pp. 187–244). Chicago: University of Chicago Press.

Sameroff, A., & Fiese, B. H. (1990). Transactional regulation and early intervention. In S. J. Meisels & J. P.Shonkoff (Eds.),Handbook of early childhood intervention(pp. 119–149). New York: Cambridge UniversityPress.

Seifer, R., Sameroff, A. J., Barrett, L. C., & Krafchuk, E. (1994). Infant temperament measured by multipleobservations and mother report.Child Development, 65,1478–1490.

Seifer, R., & Schiller, M. (1995). The role of parenting sensitivity, infant temperament, and dyadic interaction inattachment theory 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 growingpoints of attachment theory and research(pp. 146–174).Monographs of the Society for Research in ChildDevelopment, 60,(2–3, Serial No. 244).

Seifer, R., Schiller, M., Sameroff, A. J., Resnick, S., & Riordan, K. (1996). Attachment, maternal sensitivity, andinfant temperament during the first year of life.Developmental Psychology, 32,12–25.

Sroufe, L. A. (1985). Attachment classification from the perspective of infant-caregiver relationships and infanttemperament.Child Development, 56,1–14.

Sroufe, L. A. (1996).Emotional development: the organization of emotional life in the early years.New York:Cambridge University Press.

Sroufe, L. A., Jacobvitz, D., Mangelsdorf, S., DeAngelo, E., & Ward, M. J. (1985). Generational boundarydissolution between mothers and their preschool children: a relationship system approach.Child Development,56, 317–325.

Susman-Stillman, A., Kalkoske, M., Egeland, B., & Waldman, I. (1996). Infant temperament and maternalsensitivity as predictors of attachment security.Infant Behavior and Development, 19,33–47.

195S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

Tellegen, A. (1982).Brief manual for the Multidimensional Personality Questionnaire. Minneapolis: Universityof Minnesota.

Tellegen, A. (1985). Structures of mood and personality and their relevance to assessing anxiety, with anemphasis on self-report. In A. H. Tuma & J. D. Maser (Eds.),Anxiety and the anxiety disorders(pp. 681–706).Hillsdale, NJ: Erlbaum.

Teti, D. M., Gelfand, D. M., Messinger, D. S., Isabella, R. (1995). Maternal depression and the quality of earlyattachment: an examination of infants, preschoolers, and their mothers.Developmental Psychology, 31,364–376.

van den Boom, D. C. (1989). Neonatal irritability and the development of attachment. In G. Kohnstamm, J. E.Bates, & M. K. Rothbart (Eds.),Temperament in childhood(pp. 299–318). New York: Wiley.

van den Boom, D. C. (1994). The influence of temperament and mothering on attachment and exploration: anexperimental manipulation of sensitive responsiveness among lower-class mothers with irritable infants.ChildDevelopment, 65,1457–1477.

Vaughn, B. E., Bradley, C. F., Joffe, L. S., Seifer, R., & Barglow, P. (1987). Maternal characteristics measuredprenatally are predictive of ratings on temperamental “difficulty” on the Carey Infant Temperament Ques-tionnaire.Developmental Psychology, 23,152–161.

Vaughn, B., Lefever, G. B., Seifer, R., & Barglow, P. (1989). Attachment behavior, attachment security, andtemperament during infancy.Child Development, 60,728–737.

Waters, E., Vaughn, B., & Egeland, B. (1980). Individual differences in infant-mother attachment relationshipsat age one: antecedents in neonatal behavior in an urban, economically disadvantaged sample.Child Devel-opment, 51,208–216.

Weber, R. A., Levitt, M. J., & Clark, M. C. (1986). Individual variation in attachment security and StrangeSituation behavior: the role of maternal and infant temperament.Child Development, 57,56–65.

196 S.C. Mangelsdorf et al. / Infant Behavior & Development 23 (2000) 175–196

Recommended