CHAPTER 2

Development of the Attachment Styles Questionnaire1

More than six decades ago, psychoanalyst John Bowlby started his influential work on the relationship between mother and child, and in particular on the question why the mother is so important to children. This question set the stage for several years of study on the strong tie between mother and child, the so-called attachment bond, which resulted in amongst others- three books by Bowlby that would become known as the ground work of attachment theory: Attachment and Loss (1969/1982, 1973, 1980). During the years since Bowlby published his trilogy, research on attachment has flourished. Attachment has been studied in many fields of psychology, for example physiological, clinical, social, and developmental psychology and has been linked to a wide range of topics. Furthermore, studies on attachment have been conducted among people of every age period of life. In the last few decades, several instruments to measure attachment styles of children and adults have been proposed (see Cassidy & Shaver, 1999, for an overview). In the present study, we discuss a new instrument to measure attachment styles among adults, the Attachment Styles Questionnaire (ASQ). The ASQ is a multi-item instrument that measures attachment of adults to others in general, based on the two-dimensional framework of Bartholomew and Horowitz (1991). First, attachment theory is presented. Next, a historical overview of different approaches that serves as the theoretical background for the development of the ASQ is provided. Attachment Theory Bowlbys attachment theory states that during the first few years of life, children develop an attachment bond with the caregivers. This attachment bond serves an evolutionary function: it keeps the child close to the caregiver in times of stress which heightens its chance of survival. Bowlby distinguishes three features of the attachment
1 This chapter is partly based on Hofstra, J., & Van Oudenhoven, J.P. (2004). Hechtingsstijlen [Attachment styles]. In A.B. Dijkstra, J. Hofstra, J.P. van Oudenhoven, J.L. Peschar & M. van der Wal, Oud gedaan, jong geleerd? Een studie naar de relaties tussen hechtingsstijlen, competenties, EVLN-intenties en sociale cohesie. Amsterdam: Aksant; and Van Oudenhoven, J.P., Hofstra, J., & Bakker, W. (2003). Ontwikkeling en evaluatie van de Hechtingsstijlvragenlijst (HSL) [Development and evaluation of the Attachment Styles Questionnaire (ASQ)]. Nederlands Tijdschrift voor de Psychologie, 58, 95-102.

18

Chapter 2

bond with the caregivers (mostly the parents, and in particular the mother) which clearly describe the evolutionary benefits of attachment: proximity maintenance, secure base, and safe haven. With proximity maintenance is meant the childs need for being close to the caregivers. A secure base is important for the exploration of the environment: the child feels safe enough to discover the world around him. A safe haven refers to knowing that there is someone you can rely on. These characteristics are closely intertwined. When the child is near the caregiver it feels safe enough to explore its environment, because it relies on the caregiver to pay attention to potential threats. In case a threatening situation does arise for instance a stranger approaches the child- the explorative behaviour of the child stops and the child seeks proximity with the caregiver. In order to get the attention of and consequently the proximity with the caregiver in times of stress, the child shows attachment behaviour such as crying or vocalizing. When sufficient proximity is reached, the attachment behaviour stops. Over time, the interactions with the caregiver -in particular the reactions of the caregiver to the proximity seeking behaviour of the child- are internalized into mental schemas or internal working models of relations (Bowlby, 1973). These internal working models contain expectations and beliefs about whether or not the attachment figure is judged to be the sort of person who in general responds to calls for support and protection; and, whether or not the self is judged to be the sort of person towards whom anyone, and the attachment figure in particular, is likely to respond in a helpful way (Bowlby, 1973, p.204). The first refers to a model of others, and the second refers to a model of self. Bowlby stated that differential internal working models of relations lead to individual differences in attachment. Ainsworth, Blehar, Waters and Wall (1978) were the first to study and to describe individual differences in attachment patterns. They distinguished three attachment patterns or styles. A secure attachment style is developed when children perceive their caregiver as available and responsive. In contrast, children develop a resistant/ambivalent or an avoidant style when they perceive their caregiver as either inconsistently responsive or unavailable and not responsive. According to Bowlby (1973), the internal working models of the self and others and consequently the attachment styles, become increasingly resistant to change. The internal working models developed in childhood continue to guide for instance future relational choices and behaviour towards others, even in totally new contexts and with different people. Therefore, it is meaningful to study attachment styles of adults, and in the last few decades research on adult attachment has indeed flourished. In the present dissertation we used the model of adult attachment of Bartholomew and Horowitz

Development of the Attachment Styles Questionnaire

19

(1991). This model was the first to describe four adult attachment styles, based on the two dimensions model of self (positive versus negative) and model of others (positive versus negative) of the internal working models put forward by Bowlby (1973). The model will shortly be discussed in more detail. Measurement of Attachment Styles Through the years, several ways of measuring the concept of attachment have been proposed. The first studies on attachment were conducted by developmental and clinical psychologists and focused on the attachment patterns of infants. Observational methods, such as the Strange Situation of Ainsworth et al. (1978) and the Attachment Qsort of Waters (1987, 1995) which will be discussed later, were developed to measure these attachment patterns of children to their caregivers. In the mid 80s of the past century, attachment research was extended to adults. Two lines of research on adult attachment can be distinguished, which both developed their own attachment instruments. The first line of research stems from developmental and clinical psychology, and focuses on the attachment relationship of adults with their parents. Interview methods, such as the Adult Attachment Interview (George, Kaplan, & Main, 1984, 1985, 1996) were developed to measure this type of attachment of adults. The second independent line of research stems from social and personality psychology. Researchers from this research tradition focused on attachment relations of adults outside the childcaregiver dyad. To study attachment relations among large groups of respondents, alternatives to the lengthy and costly interview method were required, and therefore several self-report instruments were developed (e.g., vignettes, Hazan & Shaver, 1987; and the Relationship Questionnaire, Bartholomew & Horowitz, 1991). In the next sections, we will discuss a few of the first attachment instruments developed in the two lines of attachment research, which formed the basis of later developed self-report instruments. Observational Methods to measure Infant Attachment Strange Situation The best-known studies on the classification of the different attachment patterns among infants are observational studies in the laboratory (Ainsworth et al., 1978; Main & Solomon, 1986). These studies follow the Strange Situation paradigm for studying infantparent attachment. This paradigm has for a long time dominated the field of attachment

20

Chapter 2

research. The Strange Situation procedure is developed to examine the balance between attachment and exploratory behaviour (secure-base behaviour) under conditions of increasing, though moderate, levels of stress among infants (12 to 20 months old). The following seven episodes -which all last about three minutes-, make up the Strange Situation procedure. First, the parent and child are together in a room; the child explores the new surroundings. Then, a stranger enters the room. The first separation episode follows: the parent leaves the room, and the child is alone with the stranger. Next, the parent returns and the stranger leaves the room, this is the first reunion episode. Subsequently, in the second separation episode the parent leaves the child alone in the room. After a few minutes, the stranger enters the room again. Finally, in the second reunion episode, the parent returns and the stranger leaves. Trained observers classify the child in a certain attachment category based on the behaviour of the child to the parent in especially the two reunion episodes (see Ainsworth et al., 1978, for a detailed description of the classification procedure). Ainsworth et al. (1978), identified three types of attachment: secure, avoidant, and resistant/ambivalent. Secure children confidently explore the environment when entering the room; they miss the parent during separation and seek contact with and proximity to the parent when reunited. Avoidant children readily explore the environment; show minimal signs of distress when separated from the parent and they actively avoid and seek distance from the parent in the reunion episodes. Lastly, ambivalent children do not engage in exploration behaviour; are very distressed during separation from the parent; are not easily comforted when reunited with the parent and show both proximity seeking and proximity avoiding behaviour. Main and Solomon (1986, 1990) added a fourth attachment category disorganized/disoriented, as almost 15% of the children were not classifiable into the original three categories. The behaviour of disorganized/disoriented children appears to lack intentions or goals; it shows direct indications of fear, confusion and disorientation, such as for instance incomplete or interrupted movement and freezing. Attachment Q-sort Disadvantages of research with the Strange Situation procedure are that it takes place in a laboratory setting; it is an intrusive and stressful method for the child; and it is only designed for children between 12 and 20 months. Therefore, Waters (1987, 1995) developed the Attachment Q-sort, a less intrusive observational method to measure the quality of the secure-base behaviour of infants aged between 1 and 5 years in the home environment. The secure-base behaviour is defined as the smooth organization of and proper balance between proximity seeking and exploration. To operationalize this secure-

Development of the Attachment Styles Questionnaire

21

base behaviour of children, ninety items were developed which describe different types of attachment behaviours (a sample item is: When child is bored, he goes to mother looking for something to do). Trained observers or the parents sort the items into one of nine piles, ranging from most descriptive of the observed child, to least descriptive of the observed child. In the end, each pile eventually consists of ten items. Items in the first pile (most descriptive) are assigned a score of 9, items in the second pile receive a score of 8 and so on. Next, the Q-sort of the observed child is compared with a criterion sort (a description of a protypically secure child, developed trough the input of many experts on attachment), usually in the form of a Pearson correlation. So, this correlation reflects the degree of congruence between the individual and the criterion sort. Interview Method to measure Adult Attachment The Adult Attachment Interview (AAI; George, Kaplan, & Main, 1984, 1985, 1996) was originally developed to predict the childs attachment type from the parents state of mind regarding the own attachment history. It was assumed that this state of mind of the parent affected their parenting behaviour and subsequently the attachment patterns of the parents children. The AAI is a semi-structured, hour-long protocol and consists of 18 questions which focus on the relationship of adults with their parents during childhood; the current state of mind regarding the influence of the experiences with the parents on the adults personality; the current relationship with the parents; and traumatic losses or experiences. The interviews are audio taped and later transcribed verbatim. Trained coders rate the transcript of the interview on various scales, concerning childhood experience (e.g. loving mother, rejecting mother etc.), present state of mind regarding attachment-related experiences (e.g. idealization, lack of memory etc.) and coherence of discourse and collaboration during the interview. Adult attachment classifications are primarily based on the scales coherence of the discourse when speaking about emotion-laden attachment-relevant experiences and collaboration with the interviewer (Hesse, 1999). People are classified as secure/autonomous if they cooperate with the interviewer in coherently speaking about early attachment experiences and if they clearly value attachment relations, regardless of the nature (positive or negative) of that relationship. Dismissing individuals tend to minimize the discussion about attachment-related experiences; they devalue attachment relations and they tend to give a positive impression about the caregiver, but later discussion about experiences with the caregiver contradict this positive image. People are classified as preoccupied if they are incoherent in their stories; tend to maximize the attention to attachment-related issues; and seem to be

22

Chapter 2

enmeshed in earlier attachment relations. Finally, unresolved/disorganized individuals have substantial lapses in their discourse when discussing potentially traumatic events, and have not resolved the feelings associated with these traumatic events. Several studies have shown that the parents classification using the AAI was associated with their infants attachment classification using the Strange Situation procedure. Parents who were classified as secure, dismissing, preoccupied or unresolved/disorganized tended to have children who respectively showed secure, avoidant, resistant/ambivalent (Ainsworth et al., 1978) and disorganized attachment behaviour (Main & Solomon, 1986, 1990) in the Strange Situation (see e.g. Van IJzendoorn, 1995, for a review). Self-Report Instruments to measure Adult Attachment Vignettes The social psychologists Hazan and Shaver (1987) were the first to apply attachment theory to the study of adult relationships outside the infant-caregiver relationship. They argued that romantic love could be conceptualized as an attachment process (a process of becoming attached), experienced in different ways by different people because of the variations in attachment histories. Hazan and Shaver (1987) developed a single-item measure of attachment by translating Ainsworth et al.s (1978) description of attachment patterns of children. The measure consists of three typedescriptions, or vignettes, which correspond to the secure, avoidant and anxious/ambivalent types (or: styles, as social psychologists usually call them). After reading the vignettes, respondents have to indicate which of the vignettes describes their feelings in a romantic relationship best. A disadvantage of this forced-choice method of Hazan and Shaver (1987) is that it implies that people can have only one attachment style. The method passes over all sorts of intermediate forms of attachment. To our opinion, it is possible that people can score high on for instance avoidant and anxious/ambivalent attachment, due to the overlapping negative model of others. Besides, the extent to which a certain attachment category characterizes a person cannot be established using this instrument. As a consequence, information about the individual difference variability which exists within each category cannot be obtained. Furthermore, the determination of the attachment style of a person using only one item is assailable and does not allow to establish the internal reliability of the instrument. Therefore, Levy and Davis (1988) asked their respondents to indicate to what degree the three vignettes by Hazan and Shaver (1987) applied to them. However, also for this

Development of the Attachment Styles Questionnaire

23

measure the reliability cannot be calculated, as only one item was used per attachment style. Relationship Questionnaire Bartholomew (Bartholomew, 1990; Bartholomew & Horowitz, 1991) compared the three attachment types defined by the Adult Attachment Interview and the vignettes of Hazan and Shaver (1987) and concluded that the two methods differed in their description of the avoidant type (or dismissing type as it is called in the AAI). The dismissing attachment category in the AAI is characterized by the denial of attachment needs and striving for self-sufficiency; the avoidant attachment category of Hazan and Shavers measure is characterized by fear of being rejected when others come too close. Bartholomew suggested to distinguish two distinct forms of avoidant attachment: dismissing-avoidant, and fearful-avoidant. So, according to Bartholomew (1990), four attachment styles exist2. Besides, she showed that the four attachment styles can be placed in a two-dimensional scheme, based on Bowlbys (1973) internal working models of relations (see Figure 2.1). Combinations of the two dimensions, model of self and model of others result in the following attachment styles. The secure style: people with this style have a positive model of the self and do not doubt others; interactions with others are faced with confidence. The dismissing style: dismissing people are secure about themselves, but they refrain from personal contacts with others. Dismissing people strive for independence of others. The preoccupied style: preoccupied individuals strive for personal contact with others, but they have a negative image of the self. They anxiously seek acceptance and validation from others. The fearful style: people with this style doubt themselves as well as others. They avoid personal contacts out of fear of being hurt or deceived.

Main and Solomon (1986, 1990) and George, Kaplan and Main (1984, 1985, 1996) also distinguished four attachment styles: the secure, avoidant or dismissing, resistant/ambivalent or preoccupied, and disorganized style. However, the disorganized style has no match in the self-report instruments.

24

Chapter 2

Figure 2.1 Two-dimensional Model of Adult Attachment (Bartholomew, 1990; Bartholomew & Horowitz, 1991). Model of Self Positive Model of Others Positive Negative Secure Dismissing Negative Preoccupied Fearful

To measure Bartholomews four attachment styles, the Relationship Questionnaire (Bartholomew & Horowitz, 1991) was developed. This self-report instrument is an adaptation of Hazan and Shavers (1987) measure. It consists of four short descriptions of the four attachment styles (see Figure 2.2). Respondents have to indicate on a 7-point scale to which extent each description applies to their feelings and behaviour in close relations, such as relationships with peers. However, this kind of measurement still has the shortcoming that the internal reliability of the scales cannot be determined because of the use of only one item.
Figure 2.2 Descriptions of Attachment Styles as used in the Relationship Questionnaire (Bartholomew & Horowitz, 1991). Secure It is relatively easy for me to become emotionally close to others. I am comfortable depending on others and having others depend on me. I don't worry about being alone or having others not accept me. Dismissing I am comfortable without close emotional relationships. It is very important to me to feel independent and self-sufficient and I prefer not to depend on others or have others depend on me. Preoccupied I want to be completely emotionally intimate with others, but I often find that others are reluctant to get as close as I would like. I am uncomfortable being without close relationships, but I sometimes worry that others don't value me as much as I value them. Fearful I am somewhat uncomfortable getting close to others. I want emotionally close relationships, but I find it difficult to trust others completely, or to depend on them. I sometimes worry that I will be hurt if I allow myself to become too close to others.

Development of the Attachment Styles Questionnaire

25

To overcome the above-mentioned difficulties of categorizing respondents into mutually exclusive attachment categories (as is the case with the instrument of Hazan and Shaver, 1987) and not being able to calculate the internal reliability of the attachment scales (as is the case with Bartholomew and Horowitz instrument), several researchers broke the type descriptions into several phrases that could be scored as items on a Likertscale. For instance, Simpson (1990) and Collins and Read (1990) developed multiple-item questionnaires based on Hazan and Shavers vignettes. A multiple-item questionnaire that is based on the descriptions of the Relationship Questionnaire of Bartholomew and Horowitz (1991), is the Relationship Scales Questionnaire of Griffin and Bartholomew (1994). A sample item is: It is very important to me to feel self-sufficient. Attachment Styles Questionnaire In the present dissertation, we wanted to measure adult attachment to others in general, using multiple items based on Bartholomew and Horowitz(1991) model. The Relationship Scales Questionnaire (RSQ) of Griffin and Bartholomew (1994) would be the obvious instrument to use for that purpose. However, the RSQ suffers from low internal reliability of the scale for secure attachment ( = .41). Therefore, we aimed to develop a new reliable instrument, the Attachment Styles Questionnaire (ASQ), to measure the four adult attachment styles, based on the theoretical model of Bartholomew (1990) and consequently on the RSQ of Griffin and Bartholomew (1994). In the next section, preliminary studies on earlier versions of the ASQ are discussed, followed by a description of the psychometric properties of the latest version of the ASQ. Earlier Versions of the Attachment Styles Questionnaire In a preliminary study among 366 psychology students (76% female; mean age 20.57 years) in which a 17-item version of the ASQ was used, Schrier and Van Oudenhoven (2001) found four factors which clearly corresponded with the four attachment styles. The subscales had reasonable to high internal consistencies, ranging from = .64 for the dismissing scale to = .80 for the fearful scale. Among a subgroup of the students who filled in the questionnaire for the second time after nine months (n = 133), stability coefficients of r = .55 for the secure style, r = .62 for the fearful style, r = .70 for the dismissing style, and r = .73 for the preoccupied style were found. As expected, the secure and to a lesser extent the dismissing style were positively related to

26

Chapter 2

an indicator of model of self, self-esteem, whereas the preoccupied and fearful style were negatively related to this indicator. In a second study among three groups of students and a group of Dutch emigrants, slightly adjusted versions of the ASQ, consisting of 24 or 26 items were tested (Van Oudenhoven, Hofstra, & Bakker, 2003). Of the three groups of students (N= 790) almost 75% was female. The mean age was 21 years. Among the group of Dutch emigrants (N = 1011) 44% was female. The mean age was 61 years. The majority of the emigrants migrated to Canada, the United States of America, and Australia. The results of this study showed comparable internal consistencies ( s ranging from .62 to .85), stability coefficients (which were measured among one of the groups of students; Pearsons correlations between .56 and .69) and pattern of relations between the attachment scales and self-esteem. Latest Version of the Attachment Styles Questionnaire In the present chapter we further developed the ASQ. We added nine new items to the previous 26-item version of the ASQ, so we started with an initial pool of 35 items. Eventually, we chose 24 items based on the results of a factor analysis (see Table 2.1). In order to be selected for the final version of the ASQ, the items had to meet three criteria. First, the items had to have a high loading (> .45) on the corresponding factor. Second, the items had to contribute to the internal reliability of the corresponding scale. Third, the items had to have face validity, that is the content of the items had to reflect the attachment style which they aimed to measure. Next, we discuss the psychometric qualities of this latest version of the ASQ, that is the internal structure (factor structure and reliability), the stability, and the convergent construct validity (the relation between the attachment styles and some important constructs). With respect to the construct validity, we first studied the relations between the attachment styles and indicators of model of self. Based on Bartholomew and Horowitz (1991) model of attachment styles (see Figure 2.1) one can expect a positive relationship between the secure and dismissing style on the hand, and indicators of model of self on the other; and a negative relation between fearful and preoccupied attachment and indicators of model of self. Second, the relations between the attachment styles and indicators of model of others were examined. According to the model of Bartholomew and Horowitz, a positive relation between the secure and preoccupied style and indicators of model of others can be expected. The fearful and dismissing style are expected to be negatively related to indicators of model of others.

Development of the Attachment Styles Questionnaire

27

Furthermore, we studied the relations between the attachment scales of the ASQ with the scales of an other attachment measure, the Relationship Questionnaire (RQ; Bartholomew & Horowitz, 1991). The RQ consists of four short paragraphs, each describing one of the four attachment styles (see Figure 2). After reading the descriptions, the respondents are asked to indicate to what extent the description fits their feelings in close relations. The construct validity of the ASQ is satisfactory when the attachment scales assessed by the ASQ correlate highest with the corresponding descriptions of the RQ. In addition to relations with indicators of model of self and model of others and the RQ, we expected relations between the attachment styles and indicators of social competencies, ways of dealing with frustrating situations (or Exit, Voice, Loyalty and Neglect intentions [EVLN]; Hirschman, 1970; Farrell & Rusbult, 1992). Like attachment styles, these EVLN-intentions can roughly be described along the dimensions model of self or orientation towards the self and model of others or orientation towards others. People using exit when having a problem with someone, only take their own interests into account; people using voice take the interests of both parties into account; people using loyalty primarily focus on the interests of the other party; and people using neglect neither consider the own interests nor the interests of the other party. As attachment styles form the basis for effective social competencies (Waters & Sroufe, 1983), one might expect a connection between attachment styles and the reactions to frustrating situations. For instance, a positive relation between secure attachment and the intention to use voice in a frustrating situation is expected. In the literature, voice is defined as an active and constructive reaction to conflicting interests and values in interpersonal relationships (e.g. Farrell & Rusbult, 1992; Rusbult, Zembrodt, & Gunn, 1982). It is an active reaction because an individual using voice is doing something about the relationship; it is a constructive reaction because it is intended to maintain or revive the relationship. Individuals who react with voice in frustrating situations try to solve the problem by taking the interests of both themselves and the other party into account. Secure attachment is also considered as a constructive way of relating to others, as it is characterized by a positive image of the self and of others. Interactions with others are faced with confidence. Therefore, it is plausible that the secure attachment style is positively related to voice. In addition, we expect negative relations between the secure attachment style and the less constructive and more passive ways of dealing with frustrating situations (exit, loyalty, and neglect). In contrast, we expect the insecure styles to be negatively related to the intention to use voice, and positively to the intention to use exit, loyalty and neglect in frustrating

28

Chapter 2

situations. For instance, as the fearful attachment style is characterized by a negative image of the self and a negative image of others, one might expect that people scoring high on fearful attachment have the intention to choose neglect as this intention is characterized by not standing up for the own interests nor for the interests of others in conflicting situations. As a final way to determine the construct validity of the ASQ, we explored the relations between attachment styles and the Big Five personality traits. The Big Five model is a well known empirically-based framework describing major individual differences in personality (see e.g. Digman, 1990; John, 1990; Goldberg, 1993). This model distinguishes five robust personality traits which are believed to form the basic structure of personality. Extraversion refers to the tendency to be outgoing, assertive and active; agreeableness refers to the tendency to be kind, trusting and trustworthy; conscientiousness refers to the tendency of being organized, dependable, perseverant and achievement-oriented; emotional stability refers to the tendency to remain calm in stressful situations. Intellect refers to the tendency to be creative, perceptive and to be independent of others. This last trait is also referred to as autonomy (Hendriks, 1996). Shaver and Brennan (1992) were the first to study the relations between attachment styles and the Big Five traits and since then, many other researchers followed. Noftle and Shaver (2006) reviewed 11 studies using different methods of measurement of the attachment styles and the Big Five traits. On average, it appeared that secure attachment was consistently positively related to extraversion, agreeableness, emotional stability and conscientiousness. A high score on negative model of self, which is characteristic of fearful and preoccupied attachment, was on average strongly negatively related to emotional stability. Finally, dismissing attachment was consistently negatively related to extraversion and agreeableness. In the present chapter, we studied how the attachment styles as measured with the ASQ relate to the Big Five traits. We expected to replicate the conclusions of the review study by Noftle and Shaver (2006), also because the correlations found between the Big Five traits and the attachment styles in the studies seem to fit the image of securely, fearfully, preoccupied and dismissingly attached individuals. For instance, people scoring high on secure attachment feel comfortable in social contacts as they are self-confident and perceive others as trustworthy and reliable, and therefore will be more socially outgoing or extraverted. People scoring high on preoccupied attachment have a negative image of the self and a positive image others; they seek acceptance and validation of others. Therefore, it is plausible that they will score higher on neuroticism (or lower on emotional stability).

Development of the Attachment Styles Questionnaire

29

Method Respondents The psychometric qualities of the ASQ were examined among three groups of Dutch respondents (in total N = 3533): 1. Students of psychology (n = 1960); 74% female, 26% male; mean age was 20.55 (SD = 4.10). The group of students is a combination of seven subsamples of students which are described in more detail in Appendix 1. 2. Adults (n = 1010); 55% female; mean age was 48.78 (SD = 13.07). The group of adults is a combination of six subsamples of adults (see Appendix 1). The adults were approached by the researchers and a few research assistants. 3. A group of Dutch emigrants mainly to Canada, Australia and the United States of America (n = 563); 46% female; mean age was 59.88 (SD = 14.67). See Bakker, Van Oudenhoven, & Van der Zee (2004) for a more detailed description of the sample. Instruments Attachment styles were measured with 24 items (see Table 2.1). These items refer to attachment to others in general. The items were formulated based on the four vignettes as described by Bartholomew and Horowitz (1991) and the Relationship Scales Questionnaire by Griffin and Bartholomew (1993). The secure style was measured by seven items, such as I find it easy to get engaged in close relationships with other people. The scale for fearful attachment consisted of five items, a sample item is: I feel uncomfortable when relationships with other people become close. The preoccupied style was measured by seven items, such as: I have the impression that usually I like others better than they like me. Finally, the dismissing scale contained five items, such as: I prefer that others are independent of me, and that I am independent of others. All attachment items were measured on a 5-point scale, ranging from 1 (strongly disagree) to 5 (strongly agree). Two items were mirrored. Scores on the scales were computed by summing up the scores on the items and dividing the sum score by the number of items of the scale. The internal consistencies of the scale will be discussed shortly. In order to determine the construct validity of the scale, a number of related constructs were included. First, we measured model of self with Rosenbergs (1965) 10-item Self-Esteem Inventory (Cronbachs was .82 for subgroups 1, 6 and 7 of the students and subgroups 10 and 13 of the adults, see Appendix 1). A sample item is: I think

30

Chapter 2

positively about myself. Participants were asked to indicate the level of agreement with the items on a 5-point scale, ranging from 1 (strongly disagree) to 5 (strongly agree). As a second way to measure model of self, a 6-item scale was used from the Adult SelfPerception Profile developed by Harter (1988). A sample item is: I am happy about the way I live my life. The internal reliability of this scale was .80 (for subgroup 9 of the adults, see Appendix 1). The answers were scored on a 5-point scale, ranging from 1 (not true, thats not the way I am) to 5 (true, thats the way I am). Second, we measured model of others using six self-constructed items reflecting image of others ( = .76 for subgroup 1 of the students) a sample item is I like doing things with other people. Answers could be given on a 5-point scale, ranging from 1 (NO! Not applicable to me) to 5 (YES! Totally applicable to me). Moreover, a scale consisting of seven items ( = .73 for subgroup 9 of the adults) reflecting trust in others which had proven to be a reliable instrument (see Dijkstra, Hofstra, Van Oudenhoven, Peschar, & Van der Wal, 2004) was used. A sample item is: If somebody is friendly towards me, I become suspicious (-). A 5-point scale was used, ranging from 1 (strongly disagree) to 5 (strongly agree). Third, the Relationship Questionnaire (RQ; Bartholomew & Horowitz, 1991) was included (among subgroup 10 of the adults). This measure is an adaptation of the attachment instrument developed by Hazan and Shaver (1987). The RQ consists of four short paragraphs, describing the four attachment styles (see Figure 2.2). After reading the descriptions, respondents had to indicate on a 7-point scale the degree to which they resemble each of the four styles. Fourth, we measured ways of dealing with frustrating events or EVLN-intentions: exit, voice, loyalty and neglect (Farrell & Rusbult, 1992; Hagedoorn, 1998; Hirschman, 1970). These intentions were measured using a self-constructed questionnaire which consisted of 19 items. Respondents were asked to indicate on a 5-point scale, ranging from 1 (never) to 5 (always) for each item to what extent they would react in such a way when faced with a problem with someone. A sample item of the five items of the scale measuring voice ( = .66 and .75 for subgroup 1 of the students and subgroups 8 and 9 of the adults respectively) is: Together with the other party I think of a solution which is acceptable for both of us. Exit ( = .85 and .79 for subgroup 1 of the students and subgroups 8 and 9 of the adults respectively) was measured with four items such as: I do not associate with that person anymore. The scale measuring loyalty ( = .75 and .81 for subgroup 1 of the students and subgroups 8 and 9 of the adults respectively) contained five items like: Then, I believe that eventually everything will work out just fine. Finally, a sample item of the neglect scale consisting of five items ( = .74 and .68 for subgroup 1 of the students

Development of the Attachment Styles Questionnaire

31

and subgroups 8 and 9 of the adults respectively) is: Then, I do not care that much anymore. Finally, we measured the Big Five personality traits using the Five Factor Personality Inventory (Hendriks, Hofstee, & De Raad, 1999) which contains 20 items per scale, and a shortened version of this questionnaire which contains six items per scale. Extraversion ( = .87 and .86 for subgroup 4 of the students and subgroup 11 of the adults; .61 for subgroup 14, the emigrants, respectively) was measured by items such as: Avoids company (-) and Likes to chat. Sample items of the agreeableness scale ( = .82 and .76 for subgroups 1 and 11; and .56 for subgroup 14 respectively) are: Takes others interests into account and Empathizes with others. The scale measuring conscientiousness ( = .88 and .75 for subgroups 1 and 11; and .65 for subgroup 14 respectively) contained items like: Does things according to a plan and Is well prepared. Emotional stability ( = .87 and .86 for subgroups 1 and 11; and .62 for subgroup 14 respectively) was measured by items such as: Readily overcomes setbacks and Panics (-). Finally, the autonomy scale ( = .81 and .79 for subgroups 1 and 11; and .63 for subgroup 14 respectively) consisted of items like: Can easily link facts together and Takes the lead. Participants were asked to indicate on a 5-point scale ranging from 1 (not applicable at all) to 5 (totally applicable) whether the items were applicable to them. Results Internal Structure Factor Structure of the Scales Exploratory factor-analysis (oblimin principal components analysis) with a forced 4-factor solution showed four factors with Eigenvalues above 1.2, which clearly corresponded with the four attachment styles (see Table 2.1 for the items and the factor loadings). The four factors explained 49.27% of the variance. As can be seen from the table, all items had reasonably high factor loadings. Reliability In Table 2.2 the results with regard to the internal consistency (Cronbachs alpha) of the four scales for the different groups of respondents are presented. The items of the four attachment scales completely correspond with the items as mentioned in Table 2.1. The attachment scales reached the minimum reliability level of .60 set by Nunnally (1978), except for the dismissing scale in the emigrant group. In general, we can conclude that the

32

Chapter 2

internal consistencies of the attachment scales are satisfying. Table 2.3 shows the scale inter-correlations of the attachment scales. The scales are not orthogonal. The intercorrelations support Bartholomew and Horowitz model: opposing styles in the model the secure and fearful style, and the preoccupied and dismissing style- were negatively correlated. The inter-scale correlations between adjacent styles in the model are in line with other research on attachment styles (e.g. Tsagarakis, Kafetsios & Stalikas, 2007; Holmes & Lyons-Ruth, 2006). Stability A group of adults (n = 175) completed the ASQ twice with a one year interval, so we were able to calculate the stability of the attachment styles3. We calculated the stability coefficients of the attachment scales using Pearson correlations (see Table 2.2, last column). The stability coefficients ranged from .59 to .76. These coefficients are slightly higher than the ones that were found among the groups of students from the studies mentioned in the introduction section of this chapter. The lower stability of the attachment styles among students is understandable considering the phase of life of students: going to college and the transition from childhood to adulthood brings about many changes and insecurities. Moreover, our findings that the stability of the attachment styles is higher for adults than for students is in line with previous research by for instance Bowlby (1973) and Caspi (1998) who stated that the stability of personality characteristics increases with the age of the respondents.

3 The instrument used to measure the stability of the preoccupied scale was based on six of the seven items which we identified as items for the preoccupied style in the present chapter.

Development of the Attachment Styles Questionnaire

33

Table 2.1 Items of the Attachment Styles Questionnaire and their Factor Loadings on the corresponding Factors.

Secure attachment style

I feel at ease in emotional relationships. I avoid close ties. I trust other people and I like it when other people can rely on me. I find it easy to get engaged in close relationships with other people. I feel at ease in intimate relationships. I think it is important that people can rely on each other. I trust that others will be there for me when I need them. .60 -.62 .63 .58 .73 .58 .56 .74 .78 .77 .72 .53 .82 .59 .80 .60 -.68 .68 .48 .49 .69 .65 .64 .57

Fearful attachment style

I would like to be open to others, but I feel I cant trust other people. I would like to have close relationships with other people, but I find it difficult to fully trust them. Im afraid that my hopes will be deceived when I get too closely related to others. I am wary to get engaged in close relationships because Im afraid to get hurt. I feel uncomfortable when relationships with other people become close.

Preoccupied attachment style

I often wonder whether people like me. I have the impression that usually I like others better than they like me. I am often afraid that other people dont like me. I fear to be left alone. I dont worry whether people like me or not. I find it important to know whether other people like me. I usually find other people more interesting than myself.

Dismissing attachment style

I feel comfortable without having close relationships with other people4. It is important to me to be independent. I prefer that others are independent of me, and that I am independent of others. I like to be self-sufficient. I dont worry about being alone: I dont need other people that strongly.

4 This item had comparable factor loadings on the dismissing (.49) and the secure attachment style (.53). We decided to place this item in the scale for dismissing attachment based on the face validity (this item reflects the dismissing style more than the secure style) and the contribution to the internal consistency ( of the dismissing scale rises from .58 to .62 when this item is included; of the secure style rises with .75 to .77 when this item is included).

34

Chapter 2

Table 2.2 Internal Consistencies and Stability of the Attachment Styles Questionnaire. Attachment Scales Secure Fearful Dismissing Preoccupied Students (n=1960) .77 .81 .60 .81 Adults (n=1010) .73 .75 .61 .80 Emigrants (n=563) .68 .78 .59 .73 Total group (N=3533) .75 .79 .62 .80 Stability (n=175) .69 .76 .59 .68

Construct Validity Attachment Styles and Indicators of Model of Self and Model of Others To determine the construct validity, we first examined the relations between the attachment styles on the one hand and indicators of model of self and model of others on the other hand among students and adults (see Table 2.4). Among both groups, the relations were largely in line with Bartholomew and Horowitz (1991) model. The secure attachment style correlated positively with self-esteem/self-perception, and image of others/trust in others, and the fearful style was negatively related to self-esteem/selfperception and image of others/trust in others.

Development of the Attachment Styles Questionnaire

35

Table 2.3 Correlations between the Four Attachment Styles. Attachment scale Students (n = 1960) 1. Secure 2. Fearful 3. Dismissing 4. Preoccupied Adults (n = 1010) 1. Secure 2. Fearful 3. Dismissing 4. Preoccupied Emigrants (n = 563) 1. Secure 2. Fearful 3. Dismissing 4. Preoccupied Note. **p < .01; ***p < .001. --.37*** --.12** .31*** --.19*** .59*** .02 ---.50*** --.26*** .22*** --.19*** .42*** -.12*** ---.57*** --.26*** .19*** --.22*** .42*** -.16*** -1. 2. 3. 4.

As expected, the dismissing style correlated negatively with image of others. However, no meaningful relationship was found between dismissing attachment and selfesteem/self-perception. Finally, preoccupied attachment was, as expected, negatively related to self-esteem/self-perception. Different from what Bartholomew and Horowitz (1991) model predicts, preoccupied attachment correlated negatively with trust in others. Probably, this distrust in others of preoccupied individuals reflects distrust based on a negative image of the self, rather than distrust based on presumed bad intentions of others. For instance, preoccupied attached individuals might not trust others during social contacts, as they can hardly believe that the others sincerely want contact with them.

36

Chapter 2

Table 2.4 Correlations between Attachment Styles, Self-Esteem/Self-Perception, and Image of Others/Trust in Others. Model of Self Self-esteem (Rosenberg) Students (n = 658); Adults (n = 516) Secure Students Adults Fearful Students Adults Dismissing Students Adults Preoccupied Students Adults -.43*** -.49*** -.57*** -.01 -.30*** .08* .04 -.01 -.27*** -.14 -.29*** -.42*** -.35*** -.27*** -.55*** .29*** .26*** .22** .50*** .51*** Self-perception (Harter) Adults (n = 175) Students (n = 379) Adults (n = 175) Model of Others Image of others Trust in others

Note. * p < .05; ** p < .01; ***p < .001.

Development of the Attachment Styles Questionnaire

37

Attachment Styles and Relationship Questionnaire To further establish the validity of the four attachment scales of the ASQ, correlations between these scales and another instrument to measure attachment styles, the Relationship Questionnaire (RQ; Bartholomew & Horowitz, 1991), were calculated among a group of 461 adults. As can be seen from Table 2.5, the attachment scales of the ASQ correlated highest with the corresponding vignette of the RQ. However, the correlations are moderate and the differences between correlations are not impressive. This can be the result of the low reliability of the RQ.
Table 2.5 Correlations between the Scales of the Attachment Styles Questionnaire and the Scores on the Vignettes of the Relationship Questionnaire (Bartholomew & Horowitz, 1991) among 461 Adults. Vignette Secure ASQ-Secure ASQ-Fearful ASQ-Dismissing ASQ-Preoccupied .32*** -.31*** .16** -.33*** Vignette Fearful -.28*** .34*** -.07 .27*** Vignette Dismissing -.28*** .10 .49*** -.14** Vignette Preoccupied .07 .23*** -.21*** .34***

Note. ** p < .01; ***p < .001.

Attachment Styles and Reactions to Frustrating Situations Next, the relations between the attachment styles and the different ways of reacting to frustrating situations were calculated among students and adults (Table 2.6). The pattern of relations between the variables was the same for both groups. As we expected, the secure style correlated positively with voice, the constructive reaction to conflicts. Besides, secure attachment was negatively related to the less constructive reactions exit, loyalty, and neglect. For the insecure attachment styles, the pattern is reversed. In line with our expectations, the fearful, dismissing and preoccupied styles correlated negatively with voice and positively with the less constructive reactions. However, the relation between these insecure styles and loyalty is weak or not significant. The relations between the dismissing style and the reactions to conflicts are inconsistent and therefore difficult to interpret.

38

Chapter 2

Table 2.6 Correlations between Attachment Styles and Reactions to Frustrating Situations among 379 Students and 218 Adults. Exit Secure Students Adults Fearful Students Adults Dismissing Students Adults Preoccupied Students Adults .23*** .14* -.19*** -.16* .04 .10 .25*** .18* .11* .08 .01 -.24*** .10 -.01 .15** .10 .18** .20** -.19*** -.29*** .10* .08 .29*** .32*** -.19*** -.20** .26*** .31*** -.23*** -.17* -.37*** -.25*** Voice Loyalty Neglect

Note. * p < .05; ** p < .01; ***p < .001.

Attachment Styles and the Big Five Personality Traits As a final way of determining the construct validity of the ASQ we related the attachment styles to the Big Five personality traits among students, adults and emigrants (see Table 2.7). In all three samples, the secure style correlated positively with extraversion, as we expected, and to a lesser extent positively with agreeableness and autonomy. The fearful style is negatively related to extraversion, and to lesser extent negatively to emotional stability and autonomy. The dismissing style is negatively associated with extraversion. Finally, the preoccupied attachment style is primarily negatively correlated with emotional stability, as predicted, and to a lesser extent with autonomy and extraversion. Our results are largely in line with the review study of Noftle and Shaver (2006).

Development of the Attachment Styles Questionnaire

39

Table 2.7 Correlations between the Four Attachment Styles and the Big Five Traits among Students (n = 450), Adults (n = 133) and Emigrants (n = 563). Extraversion Secure Students Adults Emigrants Fearful Students Adults Emigrants Dismissing Students Adults Emigrants Preoccupied Students Adults Emigrants -.24*** -.19* -.13* .04 .01 -.10* .15** -.07 -.04 -.46*** -.46*** -.37*** -.37*** -.30** -.32*** -.14* -.28** -.28*** -.08 -.02 -.03 -.04 -.17 .07 .09 .14 .22*** .17** .10 -.04 -.33*** -.36*** -.22*** -.13* -.19* -.09 .03 -.23* .04 -.33*** -.34*** -.17** -.18** -.20* -.27*** .42*** .41*** .41*** .18** .39*** .18** .02 .40*** .03 .20*** .18 .04 .17** .22* .12* Agreeableness Conscientiousness Emotional Stability Autonomy

Note. * p < .05; ** p < .01; *** p < .001.

Group Differences in the Attachment Styles Next, we were interested in whether the mean scores on attachment styles differed by group (students, adults and emigrants; see Table 2.8). As the three groups differed with respect to mean age (M = 20.55; M = 48.78 and M = 59.88, for students, adults and emigrants, respectively) we performed a MANCOVA, in which we controlled for the effect of age by entering it as a covariate. Group (students, adults, and emigrants) was entered in the analysis as an independent variable, and the four attachment styles as dependent variables. At the multivariate level, a significant main effect was found for age,

40

Chapter 2

F (4, 2859) = 33.00, p <.001, 2 =.04. Univariately, this effect could be attributed to the secure attachment style, F (1, 2862) = 21.33, p <.001, 2 =.01; the fearful style; F (1, 2862) = 58.74, p <.001, 2 =.02; and the dismissing style, F (1, 2862) = 80.08, p <.001, 2 = .03. Moreover, we also found a significant main effect for group, F (8, 5720) = 18.90, p <.001, 2 =.03. Univariately, this effect could be attributed to all four attachment styles, F (2, 2862) = 15.43, p <.001, 2 =.01 for the secure style; F (2, 2862) = 10.73, p <.001, 2 =.01 for the fearful style; F (2, 2862) = 15.48, p <.001, 2 = .01 for the dismissing style; and F (2, 2862) = 15.36, p <.001, 2 = .01 for the preoccupied style. So, after controlling for the effect of age, the effect of group on the means on the attachment styles was significant. Further analysis showed that students and emigrants scored significantly higher on secure attachment than adults (ps < .01). Next, students scored higher on fearful and preoccupied attachment than adults and emigrants (ps < .01). In addition, emigrants scored higher on dismissing attachment than students and adults (p < .001).
Table 2.8 Mean Scores on the Four Attachment Styles for the Three Groups controlled for the Effect of Age. Scales Secure Fearful Dismissing Preoccupied Students (n = 1960) 4.00a 2.50a 3.23a 2.93a Adults (n =1010) 3.84b 2.27b 3.15a 2.65b Emigrants (n =563) 3.97a 2.28b 3.35b 2.67b

Note. Cell means with different subscripts (per row) differ significantly (p <.01).

In addition to studying the differences in means on attachment styles for the three groups, we were also interested in whether the means on attachment styles differed for males and females (see Table 2.9). A MANOVA in which gender was entered as the independent variable and the attachment styles as the dependent variables, showed a significant main effect for gender, F (4, 3235) = 19.93, p < .001, 2 = .02. This effect could be attributed to the secure, dismissing and preoccupied attachment style, F (1, 3238) = 39.45, p < .001, 2 = .01; F (1, 3238) = 17.19, p < .001, 2 = .01; F (1, 3238) = 19.36, p < .001, 2 = .01, respectively. Further tests revealed that women scored higher than men on secure and preoccupied attachment, and lower than men on dismissing attachment (all p s < .001).

Development of the Attachment Styles Questionnaire

41

Table 2.9 Mean Scores on the Four Attachment Styles for Females and Males with Standard Deviations between Brackets. Females (n = 2263) Secure Fearful Dismissing Preoccupied 4.00a (.56) 2.39a (.81) 3.13a (.63) 2.89a (.78) Males (n = 1262) 3.83b (.57) 2.39a (.78) 3.35b (.67) 2.66b (.68)

Note. Cell means with different subscripts (per row) differ significantly (p <.001).

Discussion The present research aimed to study the psychometric qualities of a new instrument to measure adult attachment styles. The development of the instrument seems to be -on the whole- successful. Factor analyses on our data showed four factors which clearly corresponded to the four attachment styles as put forward by Bartholomew and Horowitz (1991). The reliability of the secure, fearful and preoccupied scales is reasonable to good. However, the internal consistency of the dismissing scale is moderate (s between .59 and .62). This may be due to the variability of the dismissiveness concept. Several motives can lead to being dismissing of others. One can be dismissing because one does not need others that strongly; one has a negative image of others; or one finds it handy to be self-sufficient. The items of the dismissing scale referred to the different motives, which could account for its lower internal consistency. It is important to realize that low reliability may also explain the in general weak relations between dismissing attachment and the other variables. As could be expected, the stability of the attachment styles is lower than the stability of basic personality traits, like the Big Five, which is in general about .80 (Hendriks, Hofstee, & De Raad, 1999). However, the stability coefficients are reasonably high considering the length of the period between the two measurements, roughly one year. Even the dismissing scale had a stability coefficient of .59, despite the moderate internal consistency of the scale. The stability was the highest for the fearful attachment style. The construct validity of the four scales is reasonable. As expected, the secure style correlated positively with indicators of model of self, that is self-esteem/selfperception, whereas the fearful and preoccupied style correlated negatively with these variables. We found one weakly-significant and two non-significant correlations between

42

Chapter 2

the dismissing style and self-esteem/self-perception. As indicated before, this might be attributed to the different motives which can lead to a dismissing style. Nevertheless, our results are in line with a study by Onishi, Gjerde and Block (2001) who also found that dismissing attachment did not significantly correlate with self-esteem. The dismissing style did have a significant negative relation with an indicator of model of others: Image of others. Negative image of others thus appears to be more defining for dismissing people than self-esteem. For the secure and fearful style we found, as predicted by Bartholomew and Horowitz model, positive and negative relations with image of others/trust in others, respectively. For the preoccupied style we found a negative relation with trust in others, which is different from the model of Bartholomew and Horowitz. As preoccupied people in general are afraid of being abandoned and of other people not liking them, it might be that preoccupied people come to distrust others. This distrust of others of preoccupied people might stem from their negative image of the self instead of supposed bad intentions of other people. Furthermore, the scales of the ASQ correlated highest with the corresponding scales of the Relationship Questionnaire (Bartholomew & Horowitz, 1991). In addition, the attachment styles appeared to be associated with reactions to frustrating situations in a predictable way. People scoring high on the secure style tend to solve conflicts in a constructive manner, that is taking the interests of the self and the other party into account. At the same time, they do not tend to react in a way in which the interests of the other party are neglected (exit and neglect), or in which the own interests are neglected (loyalty). People scoring high on fearful and preoccupied attachment tend to show a reversed pattern: they do not tend to react with voice, but with exit and neglect. The correlations between the dismissing style and the EVLN-intentions were less clear, although people scoring high on dismissing attachment somewhat tend to use the less constructive strategies. Finally, the correlations between the attachment styles and the Big Five traits were in line with the expectations. People scoring high on secure attachment, showed high levels of extraversion, agreeableness and autonomy. These results fit the image of people with a positive model of the self and self-confidence during social interactions. Highly fearful individuals appeared to score low on extraversion, emotional stability and autonomy, which fits the image of people with a negative model of the self and a tendency to avoid social contacts out of fear of being hurt. People scoring high on dismissing attachment, showed low levels of extraversion. This result reflects the lower need for others which is characteristic of dismissing attachment. The preoccupied attachment style correlated negatively with emotional stability, autonomy and

Development of the Attachment Styles Questionnaire

43

extraversion. These results fit the image of people with a negative image of the self and a fear of being rejected by others in social interactions. Our results are largely in line with the results of Noftle and Shavers (2006) review study. In the past years, Brennan, Clark, and Shaver (1998), recommended that adult attachment should be assessed by measuring two underlying factors or dimensions, anxiety and avoidance. Factor analyses on our data, however, showed four factors, in stead of two, which corresponded with Bartholomew and Horowitz (1991) attachment styles. In addition, the four attachment scales are clearly distinguishable, based on the correlations with self-esteem/self-perception, image of others/trust in others, the descriptions of the Relationship Questionnaire, the EVLN-intentions and the Big Five traits. So, we prefer, based on theoretical and empirical arguments, the measurement of four attachment styles instead of two dimensions. A remarkable outcome of the present study is the differences between the three groups, students, adults and emigrants, in their mean scores on the attachment styles after controlling for the effect of age. First of all, the students scored higher on the fearful and preoccupied styles than both the adult and emigrant group. Although speculative, the differences between the groups might be attributed to the divergent phases in life. It is understandable that students, who are in a transition phase of entering adulthood and going to college which brings about many changes and insecurities, are more insecure about themselves and more focused on the question whether other people like them and find them interesting enough, than the more mature adults. Second, emigrants appeared to score higher on dismissing attachment than students and adults. The dismissing style is characterized by a certain degree of distrust of others. It could be that, because of the experiences and circumstances in an other country, emigrants get used to having less warm personal contacts, more aloofness towards others and keeping more distance from other people. Generally, women and men differ on relational variables. Therefore, it is not surprising that we found gender differences with respect to attachment styles. Results showed that women are more secure in relations. Besides, women are more concerned about whether other people like them than men, as indicated by the higher score on preoccupied attachment. Moreover, males appeared to be more dismissing. This indicates that women are somewhat more socially oriented than men, which is in line with studies that for instance showed that women are more relationship oriented (Cross & Madson, 1997), and provide more social support than men (Wellman, 1992). In future studies, a few things need to be addressed. First of all, the internal consistency of the dismissing scale is too low. More items for this subscale should be

44

Chapter 2

developed in order to improve the reliability. Second, a few items of the scale for fearful attachment are double-barrelled. Items that only consist of one statement will probably be clearer for the respondent and consequently the score on the fearful scale will better reflect the true score of the respondent. Future research could look more deeply into this issue. Also, more research is needed on the relations between the four attachment styles, particularly the dismissing style, and other theoretically relevant variables, for instance coping styles or number of friends. Nevertheless, to our opinion, the Attachment Styles Questionnaire we presented in this chapter is a valid and reliable new instrument for measuring adult attachment to others in general.

Acknowledgements We are grateful to Winny Bakker and Dick Barelds who generously gave access to the data of the emigrants, and one group of students and two groups of adults, respectively.