Presentation Format

O R I G I N A L A R T I C L E

The Brief Attachment Scale (BAS‐16): A short measure of infant attachment

Tim Cadman1 | Jay Belsky2 | Richard M. Pasco Fearon1

1 Department of Clinical, Educational and

Health Psychology, University College

London, London, UK

2 Department of Human Ecology, University of

California, Davis, California

Correspondence

Pasco Fearon, Department of Clinical,

Educational and Health Psychology, University

College London, WC1E 7HB WC1E 6BT, UK.

Email: [email protected]

Abstract

Background: Insecure attachment in infancy is associated with a range of later

socioemotional problems; therefore, it is important to identify at‐risk children so that

support can be provided. However, there are currently no well‐validated brief mea-

sures of infant attachment. The aim of this study is to create a brief version of the

Attachment Q‐Sort (AQS), one of the gold‐standard measures of attachment.

Method: Data was used from the National Institute of Child Health and Develop-

ment Study of Child Care and Youth Development (N = 1,364). The factor structure

of the AQS was explored, and Item Response Theory was used to select a reduced

number of items. Convergent validity of the shortened measure was assessed through

associations with the Strange Situation Procedure. Correlations with sensitivity,

externalising, and social competence were also examined.

Results: The Brief Attachment Scale (BAS‐16) was created consisting of two scales

of eight items, relating to (a) harmonious interaction with the caregiver and (b) prox-

imity‐seeking behaviours. The BAS‐16 showed comparable convergent, discriminant,

and concurrent validity to the full AQS.

Conclusion: This brief version of the AQS shows potential as a screening measure

for insecure attachment in infancy. Further development and validation is required

in separate samples.

KEYWORDS

AQS, attachment, Attachment Q‐Set, Attachment Q‐Sort, Brief Attachment Scale, Strange

Situation Procedure; SSP; TAS‐45

1 | INTRODUCTION

A growing body of research chronicles the negative long‐term

socioemotional correlates of insecure and disorganised attachment in

childhood (Fearon, Bakermans‐Kranenburg, Van IJzendoorn, Lapsley,

& Roisman, 2010; Groh et al., 2014; Groh, Roisman, van IJzendoorn,

Bakermans‐Kranenburg, & Fearon, 2012). Furthermore, there is an

evidence that interventions to improve parental sensitivity are effec-

tive at fostering security, especially in high‐risk populations

(Bakermans‐Kranenburg, Van Ijzendoorn, & Juffer, 2005). Accordingly,

it is important to identify infants at risk of insecure attachment at an

early age so that interventions can be offered (Allen, 2011).

At present, there are two “gold‐standard” measures of attachment

for infants—the Strange Situation Procedure (SSP; Ainsworth, Blehar,

& Waters, 1978; Main & Solomon, 1990) and the Attachment Q‐Sort

(AQS; Waters & Deane, 1985). The SSP is a laboratory procedure

involving two brief periods of separations and reunion between the

infant and their attachment figure. On the basis of time‐consuming

and video coding of their behaviour in these periods, the infant is clas-

sified as either Secure, Avoidant, Resistant, or Disorganised.

By contrast, the AQS is based on naturalistic observation of the

interaction between the child and the primary caregiver in a routine

situation, normally in the home. Following observation, raters sort

cards describing child behaviour into nine piles ranging from “most

Received: 15 September 2017 Revised: 25 June 2018 Accepted: 1 July 2018

DOI: 10.1111/cch.12599

766 © 2018 John Wiley & Sons Ltd Child Care Health Dev. 2018;44:766–775.wileyonlinelibrary.com/journal/cch

descriptive of this child” to “least descriptive of this child.” A security

score is calculated by correlating the individual sort with a criterion

sort created from an expert consensus on the behaviours of the

prototypically securely attached child. Support for the validity of

the observer, AQS has been provided by two meta‐analyses

(Cadman, Diamond, & Fearon, 2017; Van IJzendoorn, Vereijken,

Bakermans‐Kranenburg, & Marianne Riksen‐Walraven, 2004).

Despite the strengths of the SSP and AQS, in their current form,

these measures are arguable not suitable for use in routine clinical set-

tings, as both require considerable time and resources to administer.

The SSP requires a laboratory setting, video recording, and well‐

trained coders, whereas in its present form, the AQS requires a long

observation period followed by a lengthy period of sorting items.

A shorter measure, the Toddler Attachment Sort–45 (TAS‐45),

was developed to address this issue (Andreassen, 2007; Kirkland,

Bimler, Drawneek, McKim, & Schölmerich, 2004). Eight clusters of

items were derived from the original AQS along with six new items

to assess disorganised attachment. Two studies provide initial support

for the validity of the measure (Roisman & Fraley, 2008; Spieker,

Nelson, & Condon, 2011).

Although the TAS‐45 shows promise, it has potential limitations.

First, it is unknown whether it shows convergent validity with the

SSP. Second, there are limitations with the methodology used to

develop the measure. The eight clusters were identified on the basis

of the perceived semantic similarity between items; however, empirical

approaches to examine the factor structure of the AQS have reported

between three and five dimensions (Bailey, Moran, Pederson, & Bento,

2007; Bailey, Waters, Pederson, & Moran, 1999; Howes & Smith,

1995; Pederson & Moran, 1995; Posado, Waters, Crowell, & Lay,

1995). The TAS‐45 therefore may contain redundant dimensions,

making the instrument longer and of less clinical utility than it could be.

A promising alternative analytic approach is to use Q‐factor

analysis to explore the latent structure of the AQS (Bailey et al.,

1999; Bailey et al., 2007). Because the Q‐sort procedure requires cards

to be sorted in a forced normal distribution, conventional principal

component analysis (PCA) cannot be used because of violations of

the assumption of independence of measurement. Q‐factor analysis

identifies clusters of subjects with similar response patterns and can

be conceptualised as “inverted” factor analysis (with variables entered

as subjects and subjects as variables) (Brown, 1980; Kline, 2014). Using

this approach, Bailey and colleagues identified three clusters in the

AQS: (a) “Interacts Harmoniously with Mother,” (b) “Prefers Visitors,”

and (c) “Socially Withdrawn” (Bailey et al., 1999; Bailey et al., 2007).

The research presented herein builds on the work of Bailey and

colleagues to develop a shortened version of the AQS with robust

psychometric properties. The first task is to replicate previous psycho-

metric work with the AQS to reliably identify its latent structure. The

second task concerns how to move from this identified structure to

create a shortened measure.

Increasingly, Item Response Theory (IRT) is being used in the

behavioural sciences to develop new assessment tools (For an acces-

sible explanation of IRT, see Fraley, Waller, & Brennan, 2000). IRT is

a statistical approach for modelling the relationship between a latent

trait (θ) and responses on test items (Reise & Waller, 2009). The

two‐parameter model expresses the probability of an individual

endorsing a particular response on a survey item as a function of the

individual's level of the underlying trait (e.g., attachment security),

the difficulty of the item (β), and item's discrimination properties (α).

The key advantage of IRT over classical test theory is that it allows

one to model, for each individual item, the relationship between these

variables. This allows one to select items that discriminate most

effectively at the required trait level. For our purposes, IRT offers

the potential to construct a measure containing items that are most

effective at identifying very insecurely attached infants.

To summarise, although the AQS is a well‐validated measure of

attachment, it is arguably not suitable for routine clinical use. The

current study therefore aimed to use modern psychometric tech-

niques to develop a short‐form version of the AQS that could serve

such purposes. It had three main aims: first, to use Q‐factor analysis

to identify the factor structure of the AQS and compare this with pre-

vious findings and second, to use IRT to create a shortened version

based on the identified factor structure. Finally, to evaluate the valid-

ity of this shortened measure by examining correlations with relevant

outcomes, including the SSP.

2 | METHOD

2.1 | Sample

Data from the National Institute of Child Health and Development

(NICHD) Study of Early Child Care and Youth Development was used

(NICHD Early Child Care Research Network, 1997). Analysis was based

on the 1,197 participants for whom there was complete AQS data.

2.2 | Measures

2.2.1 | Attachment Q‐Set

The AQS was completed by trained raters at 24 months. Raters

observed the child interacting with their mother in the home for a

period of 2 hr. The 90 statement cards were then sorted using a

forced normal distribution into nine piles, ranging from “most

describes this child” to “least describes this child.” An overall security

score, ranging from −1.0 to 1.0, is calculated by correlating the individ-

ual sort with an expert criterion sort (Waters, 1987). Scores for sub-

scales defined in previous studies were calculated by summing the

scores for individual items (i.e., the pile number in which the item is

Key messages

• We have developed a brief screening measure of

attachment—the BAS‐16—which in this sample

comparable validity to the full AQS. This has the

potential to be a useful screening measure for insecure

attachment in clinical settings.

• Further research is needed to test this measure in

clinical samples and to develop items to measure

disorganised attachment.

CADMAN ET AL. 767

placed), reverse‐scoring where necessary (Howes & Smith, 1995;

Pederson & Moran, 1995; Posado et al., 1995).

2.2.2 | Strange Situation Procedure

The SSP (Ainsworth et al., 1978) was administered at 15 and

36 months, with a modified procedure used at the later time point

(Cassidy et al., 1992). Videotapes of the SSP were coded indepen-

dently by two trained raters with children classified into one of four

categories: Avoidant (A), Secure (B), Resistant, (C) or Disorganised (D;

Ainsworth et al., 1978; Main & Solomon, 1990). The present analysis

excluded infants who were unclassifiable (U).

2.2.3 | Maternal sensitivity

Maternal behaviours were assessed at 24 months using videotaped

mother–child interactions during the completion of interactive tasks.

Behaviour was coded on 4‐point Likert scales measuring Supportive

Presence, Respect for Autonomy, and Hostility (reverse‐scored). A

composite maternal sensitivity score was calculated by summing the

three scales, with a possible range of 0–12.

2.2.4 | Child externalising behaviour

The externalising subscale of the Child Behaviour Checklist was

completed by parents at 24 months (Achenbach, 1991). The subscale

contains 50 questions rated on a 3‐point Likert scale.

2.2.5 | Social competence

Social competence was rated by teachers at 54 months using the

Social Skills Rating System (Gresham & Elliott, 1990). The Social Skills

Rating System contains 30 questions rated on a 3‐point Likert scale. A

standardised score is calculated using included age and gender norms.

2.2.6 | Temperament

The Early Infant Temperament questionnaire was completed by

mothers at six months (Medoff‐Cooper, Carey, & McDevitt, 1993).

The Early Infant Temperament questionnaire contains 76 items mea-

sure assessing child behaviour on a 6‐point Likert scale.

2.3 | Data analysis

2.3.1 | Q‐factor analysis

To identify the latent structure of the AQS, Q‐factor analysis was

conducted using PCA with varimax rotation. The number of factors

was chosen on the basis of scree plot analysis. The conceptual theme

for each factor was identified by examining the 10 items most and

least descriptive of individuals loading on the factor, identified by their

weighted aggregate sort (For further details, see Bailey et al., 1999

and Brown, 1980).

In contrast to traditional (“R”) factor analysis, it is not advanta-

geous to use a large number of participants for Q‐factor analysis. With

R‐factor analysis, the standard error of factor loadings decreases in

proportion to sample size, meaning that larger samples yield more sta-

ble solutions (MacCallum, Widaman, Zhang, & Hong, 1999). However,

as Q‐factor analysis inverts variables and participants, analysis con-

ducted with a much greater number of subjects than variables would

theoretically yield unstable solutions. To address this issue, Q‐factor

analysis was therefore performed on a randomly identified subset of

200 participants.

2.3.2 | Reliability of factor structure

The reliability of the identified factor structure was assessed through

comparison with previous findings. First, the items most and least

descriptive of the identified factors were compared qualitatively with

those identified by Bailey and colleagues (Bailey et al., 1999; Bailey

et al., 2007). Correlations were also calculated between participants'

factor loadings and scores on previously identified AQS subscales.

2.3.3 | Item Response Theory

For the remainder of the analysis, the entire dataset was randomly split

in two, with the first half (N = 598) used for the IRT analysis and the

second half (N = 599) used to assess the validity of the new measure.

Initial pools of items for IRT were chosen on the basis of the Q‐

factor analysis. To maximise reliability, items from subscales identified

by other researchers were also included in the relevant pools if they

correlated highly with participants' factor scores (r ≥ ± 0.5).

Item parameters were estimated using Samajima's Graded

Response Model (Samejima, 1969), an extension of the two parameter

logistic model (Birnbaum, 1968) to items with more than two response

categories. Items for the shortened measure were selected on the basis

of their discrimination and difficulty, with preference given to items

that discriminate most effectively at the low end of the trait range.

2.3.4 | Validity of shortened measure

The second half of the dataset was used to explore the validity of the

shortened scale in comparison with the full AQS. Convergent validity

was assessed by exploring the relationship with SSP classification at

15 and 36 months. Concurrent validity was assessed by examining

the associations with maternal sensitivity and measures of peer com-

petence and behavioural problems (Belsky & Fearon, 2008; Fearon

et al., 2010; Schneider, Atkinson, & Tardif, 2001; Wolff & Ijzendoorn,

1997). Discriminant validity was assessed by examining associations

with infant temperament, which should only be weakly related to

attachment (Groh, Fearon, IJzendoorn, Bakermans‐Kranenburg, &

Roisman, 2017; Vaughn, Bost, & van IJzendoorn, 2008).

Q‐factor analysis was conducted using PQMethod 2.35

(Schmolck, 2014), IRT was conducted using R‐Studio 3.03 (R Studio,

2015), and all other analyses were conducted using SPSS version 22

(IBM Corp, 2013). Missing data was handled using pairwise deletion.

3 | RESULTS

3.1 | Latent structure of AQS

Q‐Factor analysis revealed four factors, explaining in total 45% of the

variance (Appendix A). There was a large correlation between Factors

1 and 3 (r = 0.66), a moderate correlation between Factors 1 and 2

(r = 0.45), and small correlations between all other factors (r = 0.22–

0.39; all p values < 0.001).

768 CADMAN ET AL.

Factor 1, labelled “Harmonious Interaction,” accounted for 19% of

the variance with 95 subjects (48%) loading on this factor, including

one who loaded negatively. Items most descriptive of children loading

on this factor indicate that the child is generally happy and obedient,

whilst items least descriptive of children loading on this factor

describe children who are often angry and upset. This factor was

similar to the first factor described by Bailey et al. (1999), labelled

“interacts harmoniously with mother.” Participant loadings on this

factor also correlated highly with AQS subscales related to coopera-

tion and difficult interactions (Table 1).

Factor 2, labelled “Proximity‐seeking,” accounted for 10% of the

variance with 42 subjects (21%) loading on this factor, including three

who loaded negatively. Items most descriptive of children loading on

this factor indicate children who maintain close proximity to their

caregivers, enjoy physical contact with them, and prefer them to

strangers. Items least descriptive indicate children who are indepen-

dent, like exploring, but who also can become upset easily. This factor

also showed similarity to the second factor reported by Bailey et al.

(1999), labelled “prefers visitors.” Participant loadings on this factor

also correlated highly with AQS subscales describing a preference

for proximity to and physical contact with the mother.

Factor 3 accounted for 10% of the variance, with 39 subjects

(20%) loading on this factor. Children loading on this factor were

characterised as being even‐tempered and sociable. This factor

showed considerable similarity to the third factor described by Bailey

et al. (1999) labelled “Socially Withdrawn.”

Factor 4 accounted for 6% of the variance, with 27 subjects (14%)

loading on this factor. However, the items most and least descriptive

of children loading on this factor failed to display any particular

coherence.

3.2 | Item response analysis

3.2.1 | Item pool

Out of the four identified factors, the first two showed clear concep-

tual coherence and convergence with previous psychometric work on

the AQS. Whilst results also indicated a third factor, this contained sig-

nificant overlap with the first factor both in terms of the correlation

between factor scores and the items most and least descriptive of

individuals loading highly on that factor. In order to produce an abbre-

viated measure with distinct subscales, it was therefore decided to

conduct IRT on pools of items relating to the first two factors only.

The first pool consisted of the 10 items most and least descriptive

of the Harmonious Interaction factor identified above and in the two

studies by Bailey and colleagues (Bailey et al., 1999; Bailey et al.,

2007), and the items from the Smooth Interaction, Compliant, and

Fussy/difficult (reverse scored) subscales. The second pool consisted

of the 10 items most and least descriptive of the proximity‐seeking

factor identified above and in the Bailey studies and the items from

the Enjoys Physical Contact, Proximity to Mother, and Physical Con-

tact with Mother subscales (Pederson & Moran, 1995; Posado et al.,

1995). Duplicate items and items with unused response categories

(N = 4) were removed, leaving pools of 37 and 40 items, respectively.

3.2.2 | Model assumptions

PCA and scree‐plot analysis were used to assess unidimensionality for

both pools of items (Hambleton, 1991). Both pools showed a clear

first factor but some evidence of a second factor. Items with loadings

>0.4 on the first factor or < 0.1 were therefore removed. This left 29

items in the first pool and 27 in the second. PCA was repeated

resulting in clearer unidimensionality in both pools (Figures 1 and 2).

Whilst there was still evidence of a small second factor in the second

pool, IRT models may be robust to small violations of unidimensional-

ity with sufficient test length and sample size (De Ayala, 2009; Kirisci,

Hsu, & Yu, 2001).

3.2.3 | Model fit

Model fit was assessed through examination of item residual plots

(DeMars, 2010; Hambleton, 1991; Hambleton & Swaminathan,

1985) and using the S‐X2 statistic (Kang & Chen, 2011; Orlando &

Thissen, 2000). No items showed signs of significant misfit.

3.2.4 | Construction of the Brief Attachment Scale

Items showed minimal differences in difficulty and were therefore

selected to form the shortened measure (named the “Brief Attachment

Scale” [BAS‐16]) based on discrimination values alone. To determine

how many items to include, items were first ranked by their discrimi-

nation values and a cumulative total was calculated. This was plotted

TABLE 1 Correlations between factor loadings and previously defined subscales (N = 200)

Subscale Harmonious Interaction

Proximity seeking Sociable Factor 4

Howes and Smith (1995)

Secure base .10 .36** −.10 −.13

Avoid .01 −.50** −.37** .47**

Seek comfort −.10 .66** .24** −.50**

Positive negotiate .56** .24** −.32** −.28**

Difficult negotiate −.71** −.28** .30** .56**

Pederson and Moran (1995)

Secure base .41** .17* .08 −.25**

Affective sharing .00 .17* .03 −.03

Enjoyment of physical contact

−.14 .66** .27** −.49**

Compliance .60** .28** −.33** −.27**

Fussy/difficult −.84** −.13 .22** .51**

Posado et al. (1995)

Smooth interactions with mother

.82** .25** −.32** −.49**

Proximity to mother −.15* .73** .16* −.37**

Physical contact with mother

−.05 .64** .18** −.47**

Interactions with other adults

.61** −.56** .60** −.57**

Note. Correlations >0.5 marked in bold.

*Significant at p < 0.05,

**Significant at p < 0.01,

***Significant at p < 0.001.

CADMAN ET AL. 769

to graphically depict the relative increase in information as more items

were included (Figure 3). As there were no clear points of inflection, it

was decided (albeit arbitrarily) to construct a 16‐item measure con-

taining two eight‐item scales (BAS‐16; Table 2). Test information

curves showed that the Harmonious Information scale discriminated

effectively over a slightly wider trait range than the Proximity‐seeking

scale (Figure 4).

3.3 | Validity of the BAS‐16

3.3.1 | Correlations with full AQS

The final stage of analysis was to evaluate the validity of the new

scales in the second half of the dataset not used for conducting the

psychometric analyses just detailed (N = 599). Scores for the Harmoni-

ous Interaction and Proximity‐seeking scales were calculated by sum-

ming scores on the individual items, reverse scoring where necessary

(i.e., high scores on the scales therefore reflect harmonious interac-

tions and proximity‐seeking behaviours). Total scores for the BAS‐16

were calculated by summing the two subscale scores.

First, we examined how the new scales related to the full AQS by

regressing the 90‐item security score on the shortened scales. As

expected, the BAS‐16 correlated positively and significantly with the

overall security score. The two scales of the BAS‐16 collectively

FIGURE 2 Scree plot for second pool of items

FIGURE 3 Cumulative discrimination values for each factor

FIGURE 1 Scree plot for first pool of items

TABLE 2 Items comprising the BAS‐16

AQS item α Harmonious Interaction

74. When mother does not do what child wants right away, child behaves as if mom were not going to do it at all.

2.26

38. Child is demanding and impatient with mother Fusses and persists unless she does what he wants right away.

2.05

79. Child easily becomes angry at mother. 1.92

81. Child cries as a way of getting mother to do what he wants. 1.64

62. When child is in a happy mood, he is likely to stay that way all day.

1.36

18. Child follows mother's suggestions readily, even when they are clearly suggestions rather than orders.

1.22

32. When mother says “No” or punishes him, child stops misbehaving (at least at that time) and does not have to be told twice.

1.13

19. When mother tells child to bring or give her something, he obeys.

1.11

Proximity seeking

35. Child is independent with mother, prefers to play on his own, and leaves mother easily when he wants to play.

2.68

43. Child stays closer to mother or returns to her more often than the simple task of keeping track of her requires.

2.64

59. When child finishes with an activity or toy, he generally finds something else to do without returning to mother between activities.

1.33

11. Child often hugs or cuddles against mother, without her asking or inviting him to do so.

1.27

90. If mother moves very far, child follows along and continues his play in the area she has moved to.

1.25

44. Child asks for and enjoys having mother hold, hug, and cuddle him.

1.21

21. Child keeps track of mother's location when he plays around the house.

1.04

57. Child is fearless. 0.74

Note. italicised items are reverse‐scored. AQS: Attachment Q‐Sort; BAS‐ 16: Brief Attachment Scale.

770 CADMAN ET AL.

accounted for 67% of the variance in the overall security score

(Table 3). Inspection of the criterion sort showed that five of the items

from the Harmonious Interaction scale and two from the Proximity‐

seeking scale were placed in either the top or bottom of two piles.

3.3.2 | Convergent validity

Next, we examined whether the BAS‐16 distinguished between

secure and insecure SSP classifications (avoidant, resistant, and

disorganised) at 15 and 36 months. At 15 months, we found no signif-

icant differences in BAS‐16 scores or overall security score between

secure and insecure infants (Table 4). By contrast, at 36 months,

infants classified as insecure on the SSP showed significantly lower

scores on the BAS‐16 total scale and Harmonious Interaction subscale

compared with those classified as secure (Table 5). To further explore

this latter finding, a series of one‐way analysis of variance were con-

ducted to compare BAS‐16 scores between the four attachment sub-

groups from the SSP (Table 6). This showed only that the secure group

scored higher than the disorganised group on the Harmonious Interac-

tion subscale.

3.3.3 | Concurrent and discriminant validity

The Harmonious Interaction subscale and the BAS‐16 total score were

significantly correlated with sensitivity, externalising, and social com-

petence, to a similar magnitude as the full AQS (Table 7). By contrast,

the Proximity‐seeking scale showed only a weak correlation with

externalising. Neither the BAS‐16 nor the full AQS were significantly

correlated with temperament (r values between −0.07 and 0.06).

4 | DISCUSSION

The aim of this study was to develop a short form of the AQS without

sacrificing validity.

Replicating the work of Bailey and colleagues, we identified two

factors in the AQS relating to (a) happy and harmonious interactions

and (b) proximity and physical‐contact seeking behaviours. These fac-

tors showed only moderate correlations with each other, suggesting

they measure distinct aspects of child behaviour. IRT was then used

to create the BAS‐16, which showed good information across a wide

trait range. Scores from the BAS‐16 showed an impressively strong

association with the full AQS suggesting that removing a large propor-

tion of the items results in only a small loss of information. However,

some proportion of this correlation may be attributable to shared sys-

tematic error, as the shortened version was not administrated inde-

pendently from the full version (Peters, Sunderland, Andrews, Rapee,

& Mattick, 2012); future work should address this issue.

Promisingly, we also found that the BAS‐16 (especially the Har-

monious Interaction subscale) showed comparable convergent, con-

current, and discriminant validity to the full AQS. At 36 months,

infants categorised as secure or insecure on the SSP showed signifi-

cant differences in BAS scores, with disorganised infants showing

the lowest scores. The BAS‐16 was not significantly associated with

infant temperament, suggesting that shortening the measure and

focusing on these specific domains does not create correlations with

temperament that were not there in the original set. However, given

the previously reported associations between the AQS and infant

temperament (Van IJzendoorn et al., 2004), this finding may reflect

FIGURE 4 Test information function for Brief Attachment Scale

TABLE 3 Relationship between subscales and overall security score

Independent variable Beta R2

BAS‐16

Harmonious interaction 0.72*** 0.67

Proximity seeking 0.36***

Note. BAS‐16: Brief Attachment Scale.

***Significant at p < 0.001.

TABLE 4 Differences in BAS‐16 and security scores between inse- cure and secure infants at 15 months

Mean (SD) Secure versus Insecure SS 15 months Secure (N = 340)

Insecure (N = 234) T P

Security score 0.30 (0.20) 0.30 (0.19) 0.19 0.85

BAS‐16 Total 86.69 (15.30) 86.50 (14.64) −0.15 0.57

Harmonious interaction

42.66 (10.70) 43.18 (10.48) 0.57 0.42

Proximity seeking 44.03 (10.06) 43.32 (10.50) −0.81 0.88

Note. BAS‐16: Brief Attachment Scale: SD: standard deviation.

TABLE 5 Differences in BAS‐16 and security scores between inse- cure and secure infants at 36 months

Mean (SD) Secure versus Insecure SS 36 months Secure (n = 342)

Insecure (n = 221) T P

Security score 0.31 (0.19) 0.26 (0.20) −2.91 0.004

BAS‐16 Total 87.92 (14.73) 84.40 (15.50) −2.70 0.007

Harmonious interaction

43.65 (10.45) 41.49 (10.45) −2.37 0.02

Proximity seeking 44.27 (10.17) 42.92 (10.42) −1.52 0.13

Note. BAS‐16: Brief Attachment Scale.

CADMAN ET AL. 771

the large interval between measurement rather than the discriminant

validity of the AQS.

Contrary to expectations, we failed to find any significant rela-

tionships between the Proximity‐seeking scale—which describes more

prototypical secure base behaviour—and SSP classification at 15 or

36‐month olds. This suggests that in the home, secure attachment pri-

marily manifests in harmonious parent–child interaction rather than

safety‐seeking/exploration behaviour. This fits with previous findings

that questions on the AQS related to noncompliance and fussiness

were most strongly associated with disorganised attachment at

15 months (Van Bakel & Riksen‐Walraven, 2004).

One explanation for this is that typical secure base behaviours are

most prominently displayed when the attachment system is activated—

in stressful situations (Cassidy & Shaver, 1999). In nonstressful, familiar

situations, these behaviours may not be observed, and related AQS items

will be placed towards the middle of a sort and will thus fail to differentiate

securely and insecurely attached children. Indeed, we found that a number

of items describing responses to separation had few to no responses in

the highest or lowest piles suggesting the events had not occurred. A

promising area of research is to introduce a structured period of separa-

tion within the observation to activate the attachment system (e.g., Van

Bakel & Riksen‐Walraven, 2004); this may enable items related to separa-

tion to be more effective in identifying insecurely attached children.

It was also surprising that neither the full AQS nor BAS‐16 signifi-

cantly related to SSP classification of 15‐month olds. This is despite the

relatively robust associations discerned in meta‐analytic reviews

(Cadman et al., 2017; Van IJzendoorn et al., 2004). Given that this asso-

ciation was equally weak for the full AQS, it may reflect the particular

methodology included in this study (e.g., the 11‐month interval of mea-

surement between the AQS and SSP at a time of rapid infant develop-

ment). Further assessment of the validity of the BAS‐16 against infancy

SSP classifications is required using different study samples.

Taken together, our findings show that in this sample, the BAS‐16

performs comparably to the full AQS. Indeed, our results suggest that a

very brief measure comprising the eight‐item Harmonious Interaction

scale could be as effective as the full AQS at identifying insecure attach-

ment in the home. Whilst this scale does not cover the same range of

behaviours as the full measure, it appears to capture the important

manifestations of insecure attachment within a nonstressful setting.

This has important clinical implications. Given its brevity, the scale

could plausibly be used by healthcare professionals who have routine

contact with at‐risk families. It could also be used in longitudinal stud-

ies where brevity is paramount given the large batteries of measures

involved.

This study had a number of strengths. It is the first study to use

IRT to reduce the number of items in the AQS based on an empirically

determined factor structure and was conducted on a large sample. The

main limitation of the study was that despite the forced distribution of

the sorting process, it was assumed that AQS items were independent.

The justification for treating them as independent is the vast number

of possible ordered combinations of AQS items (90! = 1.49 E + 145).

However, it is unknown how similar or different the pattern of

responses would have been if rated on independent scales.

Future studies are therefore needed to assess the validity of these

new scales when used on their own (i.e., not extracted from the total

set of 90 items that was administered), for example, in a Likert‐type

response format and when they are applied in different populations

such as clinical or at‐risk groups. It should also be examined how much

time (if any) a shortened measure could save, given that a lengthy

period of observation may still be necessary to capture the relevant

behaviours. Potential drawbacks of using healthcare professionals to

assess attachment with a brief measure should also be investigated,

for example, reliance on parent report or response bias. Finally, whilst

the BAS‐16 did successfully distinguish disorganised from secure

TABLE 6 One‐way ANOVAs with strange situation classification at 36 months (N = 563)

Mean (SD) SS (36 months) F Effect size (η2p)

A (N = 29) B (N = 342) C (N = 87) D (N = 105)

Security score 0.31 (0.20) 0.31 (0.19)a 0.27 (0.22) 0.24 (0.18)b 3.82* 0.020

BAS‐16 Total 86.83 (15.79) 87.92 (14.73) 85.10 (16.23) 83.16 (14.83) 2.99 0.016

Harmonious interaction 46.10 (11.30) 43.65 (10.45)a 41.20 (10.92) 40.45 (10.19)b 4.11** 0.022

Proximity seeking 40.72 (9.79) 44.27 (10.17) 43.90 (9.79) 42.70 (11.06) 1.50 0.008

Note. ANOVA: analysis of variance; BAS‐16: Brief Attachment Scale: SD: standard deviation. Subscripts a and b represent statistically significant differ- ences between attachment categories at p<0.05.

Significant at *P<0.05, **P<0.01

TABLE 7 Correlations with sensitivity, externalising, and social competence

Sensitivity (24 months) Externalising (24 months) Externalising (Grade 6) Social competence (54 months)

Security score 0.23*** −0.25*** −0.08 0.12**

BAS‐16 Total 0.15** −0.23*** −0.12** 0.10*

Harmonious interaction 0.18*** −0.24*** −0.07 0.13**

Proximity seeking 0.05 −0.10* −0.11* 0.01

Note. BAS‐16: Brief Attachment Scale.

*p < 0.05.

**p < 0.01.

***p < 0.001.

772 CADMAN ET AL.

infants, the inclusion of specific questions to identify disorganised

attachment could further improve the sensitivity of the measure.

In summary, we have developed a shortened measure of attach-

ment based on the AQS—the BAS‐16. This new measure shows good

psychometric properties and could be a brief yet valid alternative to

much longer existing measures of attachment.

ORCID

Tim Cadman http://orcid.org/0000-0002-7682-5645

Jay Belsky http://orcid.org/0000-0003-2191-2503

Richard M. Pasco Fearon http://orcid.org/0000-0003-1847-8443

REFERENCES

Achenbach, T. M. (1991). Integrative guide for the 1991 CBCL/4‐18, YSR, and TRF profiles. Department of Psychiatry, University of Vermont Burlington.

Ainsworth, M., Blehar, M., & Waters, E. (1978). Patterns of attachment: A psychological study of the strange situation. Oxford, England: Lawrence Erlbaum.

Allen, G. (2011). Early intervention: The next steps, an independent report to her majesty's government by Graham Allen MP. The Stationery Office.

Andreassen, A. F. P. (2007). Early childhood longitudinal study, birth cohort (ECLS‐B) psychometric report for the 2‐year data collection (NCES 2007–084). Washington, D.C.: National center for education statistics, Institute of education sciences, U.S. department of education.

Bailey, H., Moran, G., Pederson, D. R., & Bento, S. (2007). Understanding the transmission of attachment using variable‐and relationship‐centered approaches. Development and Psychopathology, 19(02), 313–343.

Bailey, H., Waters, C. A., Pederson, D. R., & Moran, G. (1999). Ainsworth revisited: An empirical analysis of interactive behavior in the home. Attachment & Human Development, 1(2), 191–216.

Bakermans‐Kranenburg, M. J., Van Ijzendoorn, M. H., & Juffer, F. (2005). Disorganized infant attachment and preventive interventions: A review and meta‐analysis. Infant Mental Health Journal, 26(3), 191–216.

Belsky, J., & Fearon, R. (2008). Precursors of attachment security. In J. Cassidy, & P. R. Shaver (Eds.), Handbook of attachment: Theory, research, and clinical applications (pp. 295–316). New York: Guilford Press.

Birnbaum, A. (1968). Some latent trait models and their use in inferring an examinee's ability. In F. M. Lord, M. R. Novick, & A. Birnbaum (Eds.), Statistical theories of mental test scores (pp. 397–479). Reading, Mass: Addison‐Wesley.

Brown, S. R. (1980). Political subjectivity: Applications of Q methodology in political science. New Haven, CT: Yale University Press.

Cadman, T., Diamond, P. R., & Fearon, P. (2017). Reassessing the validity of the attachment Q‐sort: An updated meta‐analysis. Infant and Child Development, 27. https://doi.org/10.1002/icd.2034

Cassidy, J., Marvin, R., & Group, w. t. M. W. (1992). Attachment organiza- tion in preschool children: Procedures and coding manual. Unpublished manuscript, University of Virginia, 125–131.

Cassidy, J., & Shaver, P. R. (1999). Handbook of attachment: Theory, research, and clinical applications. New York: Guildford Press.

De Ayala, R. J. (2009). Theory and practice of item response theory. New York: Guilford Press.

DeMars, C. (2010). Item response theory. Oxford: Oxford University Press.

Fearon, R., Bakermans‐Kranenburg, M. J., Van IJzendoorn, M. H., Lapsley, A. m., & Roisman, G. I. (2010). The significance of insecure attachment and disorganization in the development of children's externalizing behavior: A meta‐analytic study. Child Development, 81(2), 435–456.

Fraley, R. C., Waller, N. G., & Brennan, K. A. (2000). An item response theory analysis of self‐report measures of adult attachment. Journal of Personality and Social Psychology, 78(2), 350–365.

Gresham, F. M., & Elliott, S. N. (1990). Social skills rating system (SSRS). MN: American Guidance Service.

Groh, A. M., Fearon, R., IJzendoorn, M. H., Bakermans‐Kranenburg, M. J., & Roisman, G. I. (2017). Attachment in the early life course: Meta‐analytic evidence for its role in socioemotional development. Child Development Perspectives, 11(1), 70–76.

Groh, A. M., Fearon, R. P., Bakermans‐Kranenburg, M. J., Van IJzendoorn, M. H., Steele, R. D., & Roisman, G. I. (2014). The significance of attach- ment security for children's social competence with peers: A meta‐ analytic study. Attachment & Human Development, 16(2), 103–136.

Groh, A. M., Roisman, G. I., van IJzendoorn, M. H., Bakermans‐Kranenburg, M. J., & Fearon, R. (2012). The significance of insecure and disorganized attachment for children's internalizing symptoms: A meta‐analytic study. Child Development, 83(2), 591–610.

Hambleton, R. K. (1991). Fundamentals of item response theory (Vol. 2). Newbury Park, CA: Sage publications.

Hambleton, R. K., & Swaminathan, H. (1985). Item response theory: Princi- ples and applications. Boston MA: Kluwer‐Nijhoff.

Howes, C., & Smith, E. W. (1995). Relations among child care quality, teacher behavior, children's play activities, emotional security, and cog- nitive activity in child care. Early Childhood Research Quarterly, 10(4), 381–404. https://doi.org/10.1016/0885‐2006(95)90013‐6

IBM Corp. (2013). IBM SPSS Statistics for Windows (Version 22.0). Armonk, NY.

Kang, T., & Chen, T. T. (2011). Performance of the generalized S‐X2 item fit index for the graded response model. Asia Pacific Education Review, 12(1), 89–96.

Kirisci, L., Hsu, T.‐c., & Yu, L. (2001). Robustness of item parameter estima- tion programs to assumptions of unidimensionality and normality. Applied Psychological Measurement, 25(2), 146–162.

Kirkland, J., Bimler, D., Drawneek, A., McKim, M., & Schölmerich, A. (2004). An alternative approach for the analyses and interpretation of attachment sort items. Early Child Development and Care, 174(7–8), 701–719. https://doi.org/10.1080/0300443042000187185

Kline, P. (2014). An easy guide to factor analysis. London: Routledge.

MacCallum, R. C., Widaman, K. F., Zhang, S., & Hong, S. (1999). Sample size in factor analysis. Psychological Methods, 4(1), 84–99.

Main, M., & Solomon, J. (1990). Procedures for identifying infants as disor- ganized/disoriented during the Ainsworth Strange Situation. In M. T. Greenberg, D. Cicchetti, & E. M. Cummings (Eds.), The John D. and Cath- erine T. MacArthur Foundation series on mental health and development. Attachment in the preschool years: Theory, research, and intervention (pp. 121–160). Chicago, IL, US: University of Chicago Press.

Medoff‐Cooper, B., Carey, W. B., & McDevitt, S. C. (1993). The Early Infancy Temperament Questionnaire. Journal of Developmental and Behavioral Pediatrics, 14(4), 230–235.

NICHD Early Child Care Research Network (1997). The effects of infant child care on infant‐mother attachment security: Results of the NICHD Study of Early Child Care. Child Development, 68, 860–879.

Orlando, M., & Thissen, D. (2000). Likelihood‐based item‐fit indices for dichotomous item response theory models. Applied Psychological Mea- surement, 24(1), 50–64.

Pederson, D. R., & Moran, G. (1995). A categorical description of infant‐ mother relationships in the home and its relation to Q‐sort measures of infant‐mother interaction. Monographs of the Society for Research in Child Development, 60(2–3), 111–132.

Peters, L., Sunderland, M., Andrews, G., Rapee, R. M., & Mattick, R. P. (2012). Development of a short form Social Interaction Anxiety (SIAS) and Social Phobia Scale (SPS) using nonparametric item response theory: The SIAS‐ 6 and the SPS‐6. Psychological Assessment, 24(1), 66–76.

Posado, G., Waters, E., Crowell, J. A., & Lay, K. L. (1995). Is it easier to use a secure mother as a secure base? Attachment Q‐sort correlates of the adult attachment interview. Monographs of the Society for Research in Child Development, 60(2–3), 133–145.

R Studio. (2015). R Studio (Version 3.03). Boston, MA. Retrieved from http://www.rstudio.org/

CADMAN ET AL. 773

Reise, S. P., & Waller, N. G. (2009). Item response theory and clinical mea- surement. Annual Review of Clinical Psychology, 5, 27–48.

Roisman, G. I., & Fraley, R. C. (2008). A behavior‐genetic study of parenting quality, infant attachment security, and their covariation in a nationally representative sample. Developmental Psychology, 44(3), 831–839. https://doi.org/10.1037/0012‐1649.44.3.831

Samejima, F. (1969). Estimation of latent ability using a response pattern of graded scores. Psychometrika Monograph Supplement, 34(4, Pt. 2), 100.

Schmolck, P. (2014). PQMethod Version 2.35. from http://schmolck. userweb.mwn.de/qmethod/

Schneider, B. H., Atkinson, L., & Tardif, C. (2001). Child–parent attachment and children's peer relations: A quantitative review. Developmental Psy- chology, 37(1), 86–100.

Spieker, S., Nelson, E. M., & Condon, M. C. (2011). Validity of the TAS‐45 as a measure of toddler‐parent attachment: Preliminary evidence from Early Head Start families. Attachment & Human Development, 13(1), 69–90. https://doi.org/10.1080/14616734.2010.488124

Van Bakel, H. J., & Riksen‐Walraven, J. M. (2004). AQS security scores: What do they represent? A study in construct validation. Infant Mental Health Journal, 25(3), 175–193.

Van IJzendoorn, M. H., Vereijken, C. M., Bakermans‐Kranenburg, M. J., & Marianne Riksen‐Walraven, J. (2004). Assessing attachment security with the attachment Q sort: Meta‐analytic evidence for the validity of the observer AQS. Child Development, 75(4), 1188–1213.

Vaughn, B. E., Bost, K. K., & van IJzendoorn, M. H. (2008). Attachment and temperament: Additive and interactive influences on behavior, affect, and cognition during infancy and childhood.

Waters, E. (1987). Attachment Behaviour Q‐Set (Revision 3.0). Unpublished instrument, State University of New York at Stony Brook, Department of Psychology.

Waters, E., & Deane, K. E. (1985). Defining and assessing individual differences in attachment relationships: Q‐methodology and the orga- nization of behavior in infancy and early childhood. Monographs of the Society for Research in Child Development, 50, 41–65.

Wolff, M. S., & Ijzendoorn, M. H. (1997). Sensitivity and attachment: A meta‐analysis on parental antecedents of infant attachment. Child Development, 68(4), 571–591.

How to cite this article: Cadman T, Belsky J, Pasco Fearon

RM. The Brief Attachment Scale (BAS‐16): A short measure

of infant attachment. Child Care Health Dev. 2018;44:766–775.

https://doi.org/10.1111/cch.12599

APPENDIX A

ITEMS MOST AND LEAST DESCRIPTIVE OF ATTACHMENT Q‐SORT FACTORS

Factor 1 (“Harmonious Interaction”)

Factor 2 (“Proximity‐seeking”)

Most descriptive

9. Child is light‐hearted and playful most of the time.

48. Child readily lets new adults hold or share things he has, if they ask to.

18. Child follows mother's suggestions readily, even when they are clearly suggestions rather than orders.

77. When mother asks child to do something, he readily understands what she wants (may or may not obey).

1. Child readily shares with mother or lets her hold things if she asks to.

(Continued)

15. Child is willing to talk to new people, show them toys, or show them what he can do, if mother asks him to.

89. Child's facial expressions are strong and clear when he is playing with something.

62. When child is in a happy mood, he is likely to stay that way all day.

41. When mother says to follow her, child does so.

5. Child is more interested in people than in things.

Least descriptive

75. At home, child gets upset or cries when mother walks out of the room (may or may not follow her).

79. Child easily becomes angry at mother.

54. Child acts like he expects mother to interfere with his activities when Morn is simply trying to help him with something.

58. Child largely ignores adults who visit the home and finds his own activities more interesting.

65. Child is easily upset when mother makes him change from one activity to another (even if the new activity is something child often enjoys).

30. Child easily becomes angry with toys.

38. Child is demanding and impatient with mother. Fusses and persists unless Mom does what he wants right away.

63. Even before trying things himself, child tries to get someone to help him.

6. When child is near mother and sees something he wants to play with, he fusses or tries to drag mother over to it

31. Child wants to be the center of mother's attention. If mom is busy or talking to someone, he interrupts.

Most descriptive

11. Child often hugs or cuddles against mother, without being asked or invited to do so.

44. Child asks for and enjoys having mother hold, hug, and cuddle him.

28. Child enjoys relaxing in mother's lap.

43. Child stays closer to mother or returns to her more often than the simple task of keeping track of her requires.

1. Child readily shares with mother or lets her hold things if she asks to.

18. Child follows mother's suggestions readily, even when they are clearly suggestions rather than orders.

21. Child keeps track of mother's location when he plays around the house.

50. Child's initial reaction when people visit the home is to ignore or avoid them, even if he eventually warms up to them.

64. Child enjoys climbing all over mother when they play.

53. Child puts his arms around mother or puts his hand on her shoulder when she picks him up.

Least descriptive

35. Child is independent with mother, prefers to play on his own, and leaves mother easily when he wants to play.

54. Child acts like he expects mother to interfere with his activities when she is simply trying to help him with something.

67. When the family has visitors, child wants them to pay a lot of attention to him.

30. Child becomes easily angry with toys.

7. Child laughs and smiles easily with a lot of different people.

65. Child is easily upset when mother makes him change from one activity to another.

774 CADMAN ET AL.

Factor 3 (“Sociable”)

Factor 4 (Not named)(Continued)

66. Child easily grows fond of adults who visit his home and are friendly to him.

52. Child has trouble handling small objects or putting small things together.

79. Child easily becomes angry at mother.

12. Child quickly gets used to people or things that initially made him shy or frightened him.

Most Descriptive

37. Child is very active, always moving around, and prefers active games to quiet ones.

5. Child is more interested in people than in things.

67. When the family has visitors, child wants them to pay a lot of attention to him.

44. Child asks for and enjoys having mother hold, hug, and cuddle him.

48. Child readily lets new adults hold or share things he has, if they ask to.

11. Child often hugs or cuddles against mother, without her asking or inviting him to do so.

28. Child enjoys relaxing in mother's lap.

77. When mother asks child to do something, he readily understands what she wants (may or may not obey.)

85. Child is strongly attracted to new activities and new toys.

15. Child is willing to talk to new people and show them toys.

Least descriptive

58. Child largely ignores adults who visit the home and finds his own activities more interesting.

50. Child's initial reaction when people visit the home is to ignore or avoid them even if he eventually warms up to them.

76. When given a choice, child would rather play with toys than with adults.

75. At home, child gets upset or cries when mother walks out of the room (may or may not follow her).

63. Even before trying things himself, child tries to get someone to help him.

82. Child spends most of his playtime with just a few favourite toys or activities.

04. Child is careful and gentle with toys and pets.

39. Child is often serious and businesslike when playing away from mother or alone with his toys.

32. When mother says “No” or punishes him, child stops misbehaving (at least at that time) and does not have to be told twice.

65. Child is easily upset when mother makes him change from one activity to another.

Most descriptive

81. Child cries as a way of getting mother to do what he wants.

68. On the average, child is a more active type person than mother.

77. When mother asks child to do something, he readily understands what she wants (May or may not obey)

38. Child is demanding and impatient with mother. Fusses and persists unless Mom does what he wants right away.

74. When mother does not do what child wants right away, child behaves as if Mom were not going to do it at all.

85. Child is strongly attracted to new activities and new toys.

50. Child's initial reaction when people visit the home is to ignore or avoid them, even if he eventually warms up to them.

37. Child is very active, always moving around, and prefers active games to quiet ones.

89. Child's facial expressions are strong and clear when he is playing with something.

27. Child laughs when mother teases him

Least descriptive

75. At home, child gets upset or cries when mother walks out of the room. (May or may not follow her.)

66. Child easily grows fond of adults who visit his home and are friendly to him.

11. Child often hugs or cuddles against mother, without her asking or inviting him to do so.

67. When the family has visitors, child wants them to pay a lot of attention to him.

63. Even before trying things himself, child tries to get someone to help him.

16. Child prefers toys that are modelled after living things (e.g., dolls and stuffed animals).

32. When mother says “No” or punishes him, child stops misbehaving (at least at that time) and does not have to be told twice.

52. Child has trouble handling small objects or putting small things together.

56. Child becomes shy or loses interest when an activity looks like it might be difficult.

24. When mother speaks firmly or raises her voice at him, child becomes upset, sorry, or ashamed about displeasing her.

CADMAN ET AL. 775

This document is a scanned copy of a printed document. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material.