PEDERSON, DAVID R.; MORAN, GREG; SITKO, CAROLYN; CAMPBELL, KATHY; GHESQUIRE, KRISTEN; and ACTON, HEATHER.
University of Western Ontario
Maternal Sensitivity and the Security of Infant-Mother Attachment: A Q-Sort Study. CHILD DEVELOPMENT, 1990, 61, 1974-1983. 40 mothers and their 12-month-old infants were observed twice at home by 2 observers for 2 hours. After the second visit, the observers described the infant using the Waters Attachment Behavior Q-sort and the mother’s interactive behavior with the Maternal Behavior Q-sort developed by the present authors and Ainsworth’s rating scales. Maternal sensitivity was unrelated to maternal age, income, or SES, but correlated positively with maternal education. Mothers of more difficult children were less sensitive. A strong relation was
found between infant attachment and maternal sensitivity as measured by the Maternal Behavior Qsort and by the Ainsworth scales. Using the Q-sort procedure, mothers of more secure infants were more frequently characterized as noticing their babies’ signals and using these signals to guide their behavior; they also were more knowledgeable about their infant and appeared to enjoy them more than mothers of less secure infants.
The study of caregiver-infant relationships has escalated since the mid-1970s, stimulated by John Bowlby’s (1958, 1969) theoretical innovations and Mary Ainsworth’s (1982; Ainsworth, Blehar, Waters, & Wall, 1978; Ainsworth, Bell, & Stayton, 1974) longitudinal study. Bowlby amalgamated psychoanalytic concepts with ethological methods and evolutionary theory. The result has proven to be a very successful heuristic framework for the naturalistic study of mother-infant interactions. Ainsworth built on this foundation with the introduction of innovative assessment procedures and stimulating new views of early social relationships. The most important of these assessment procedures was the Ainsworth-Wittig (1969) Strange Situation and the description of qualitatively different attachment relationships (Ainsworth et al., 1978).
Assessments of attachment-The description of the infant-caregiver attachment relationship as “secure,” “anxious-avoidant,” or “anxious-resistant” on the basis of behavior in the Strange Situation has so predominated that this analytic method has become virtually synonymous with the assessment of attachment. The approach has not been without its critics (e.g., Lamb, Thompson, Gardner, & Charnov, 1985; Waters, 1985; Waters & Deane, 1985). Much of Lamb et al.’s misgivings concern the classification system as a way of describing infant behavior in the Strange Situation. These authors suggest that much information is lost in reducing Strange Situation outcomes ultimately to a binary secure-anxious code, and propose that cluster analyses may provide more satisfactory descriptions of the variations in infant attachment behaviors in the Strange Situation. Waters points out that the success of the Strange Situation procedures has resulted in the relative neglect of Ainsworth’s original focus on the description of the infant-mother relationship in the infant’s natural environment. While not denying the power of the Strange Situation classifications, he urges a return to studying attachment behavior in nonlaboratory settings, the actual developmental context of the relationship. Waters and Deane (1985) constructed the Attachment Behavior Q-set to facilitate the adoption of a more naturalistic assessment of the attachment relationship. The most recent version of this instrument (Waters, 1986) consists of 90 behavioral
descriptions that are sorted into nine piles according to similarity with the infant’s behavior. The correlation between this sort and a sort descriptive of a prototypically secure infant (i.e., the “criterion” sort) is taken as the subject infant’s security of attachment score. Attachment security so assessed has been related to interactions between siblings (Bosso, 1984) and peers (Denham, 1987), as well as to children’s adjustment to day-care (Vaughn, Deane, & Waters, 1985). There is also evidence of consistency between the homebased attachment Q-sort and Strange Situation classifications (Bosso, 1984; Vaughn, 1985; Waters, 1985). Initial applications, then, suggest that the Waters and Deane Attachment Q-set provides an important homebased complement to the Strange Situation. Moreover, because the Q-sort produces a continuous metric and may be applied repeatedly, it avoids some of the methodological shortcomings of the earlier assay.
Attachment and maternal sensitivity.- A central assumption made by attachment theorists (Ainsworth, 1982; Ainsworth et al., 1978; Bowlby, 1969; Sroufe & Fleeson, 1986) is that attachment quality is the cumulative product of the caregiver’s responses to the infant’s signals for proximity and contact. The importance assigned to caregiver behavior has stimulated considerable research on the differences in care received by securely and anxiously attached infants. Indeed, the focus of much of Ainsworth’s work has been the description of maternal behaviors that predict qualitative differences in attachment. Ainsworth reported that variations in maternal sensitivity were associated with differences in the infant-mother relationship reflected both in the Strange Situation and in infant behaviors observed at home (Ainsworth, Bell, & Stayton, 1971, 1974; Ainsworth et al., 1978).
Reviews of research conducted since Ainsworth’s study (Goldsmith & Alansky, 1987; Lamb et al., 1985) have questioned the strength of the relation between maternal sensitivity and attachment security. Goldsmith and Alansky’s meta-analysis revealed substantial variability in reported differences between the sensitivity of mothers of securely and anxiously attached infants. The effect sizes ranged from +2.48 to -.18 SD units, with a median of about a third of a standard deviation. Goldsmith and Alansky conclude that the effects of maternal behavior on infant attachment may not be as robust as would be anticipated either by attachment theory or Ainsworth’s original study.
One interpretation of these findings is that the theoretical claim that maternal sensitivity is the primary, if not the exclusive, antecedent of infant security is wrong (Campos, Barrett, Lamb, Goldsmith, & Stenberg, 1983; Lamb et al., 1985). However, this interpretation cannot accommodate those studies that have reported a strong relation between maternal sensitivity and attachment security (e.g., Egeland & Farber, 1984; Grossmann, Grossmann, Spangler, Suess, & Unzner, 1985; Smith & Pederson, 1988). For example, Smith and Pederson (1988) correctly distinguished between securely and anxiously attached infants with 94% accuracy using maternal behaviors as the predictor variables in a discriminant function analysis.
Perhaps a more parsimonious interpretation, given the rather extreme variations in effect sizes, is that various measures of maternal sensitivity have not been uniformly effective; sensitivity may be a more elusive target than anticipated by many investigators. The original Ainsworth et al. (1971, 1974) study of the roots of the attachment relationship involved repeated home observations throughout the first year of the infant’s life. The descriptions of maternal sensitivity were based on over 60 hours of naturalistic observations for each dyad. This extensive contact provided these observers with much broader samples of maternal behavior in a greater variety of conditions than has been available to observers in most subsequent investigations.
Assessment of maternal sensitivity.-
Ainsworth et al. (1971, 1974) developed ninepoint rating scales that summarized the mother’s accessibility, acceptance, cooperation, and sensitivity. Each scale contained a lengthy description of the concept in addition to a long paragraph anchoring the odd points on the scale. Ainsworth’s descriptions (her operational definition of caregiver sensitivity) demand of the effective observer a considerable understanding of the mother’s psychological processes and the infant’s needs. Assessment of maternal sensitivity necessitates observations in contexts in which these more subtle attributes might be discerned.
The assessment problems are further complicated by the fact that, as Belsky, Rovine, and Taylor (1984) demonstrated, although maternal sensitivity implies that the mother is actively involved with her infant, maternal involvement per se does not distinguish between securely and anxiously attached infants. In their study, anxious avoidant infants had the most and anxious- 1976 Child Development resistant infants the least actively involved mothers, with securely attached infants having mothers who were moderately involved.
It is apparent from these considerations that the assessment of maternal sensitivity requires a context in which the subtlety of the mother’s awareness of her infant can be revealed. In his discussion of maternal sensitivity, Fromm (1956) provides a helpful analogy by suggesting that maternal sensitivity is like the receptiveness of an experienced driver to subtle, yet significant changes in the operation of the automobile. “Even a small, unaccustomed noise is noticed, … Yet, he is not thinking about all these factors; his mind is in a state of relaxed alertness, open to all relevant changes in the situation on which he is concentrated” (Fromm, 1956, p. 115). This analogy is useful because it reminds us that sensitivity involves an openness to signals in the context of the need to attend to other competing events. Settings, such as a brief period of free play in a laboratory, in which the mother’s attention can be entirely focused on the activities of her infant may mask all but the most conspicuous individual differences. Observations in situations that distract the mother from the infant’s more understated cues may be more effective in discerning less obvious differences. Smith and Pederson’s (1988) success at demonstrating a strong relation between sensitivity and attachment lends credence to this suggestion. Rather than observing the mother at free play with her infant, they required that she complete a questionnaire at the same time as her child was allowed to explore and find entertainment in an observation room devoid of toys. The situation quite clearly put the mothers in a situation of potentially conflicting demands: the attention and deliberation demanded by the questionnaire, and the need to be aware of her infant’s behavior. Finally, Smith and Pederson (1988) used a scoring system in which the appropriateness of the mother’s behavior was conditional upon the particular activities and signals of the infant.
Many contemporary models of parenting behaviors (e.g., Belsky, Hertzog, & Rovine, 1986) point out that maternal behaviors do not occur in isolation but are part of a complex system. Maternal sensitivity is expected to be influenced by the challenges of caring for a particular infant, family stress, and the mother’s social context. The mothers were asked to complete the Parenting Stress Index (Abidin, 1986) as an assessment of their perceptions of family stress and the challenges of their infants. Demographic information concerning the mothers’ social context was obtained in an interview.
-Families were recruited from a volunteer subject pool maintained by a university child development study group and from newspaper birth announcements. Forty mothers and their infants (aged 1 year + 2 weeks) were observed. Mothers’ ages ranged from 22 to 39 years (M = 29.9 years); fathers were slightly older (M = 32.1 years). Parental education ranged from 9 to 24 years (M = 14.9 years). Mean family income was approximately $35,000 (Canadian dollars). Socioeconomic status, according to the socioeconomic index for Canada (Blishen, Carrol, & Moore, 1987), ranged from 25 (laborer in a soft drink bottling plant) to 101 (physician), with a mean of 51 (e.g., commercial traveler; sales of animal pharmaceuticals). Families included one to four children with a mean of two.
Development of Maternal Behavior Qset.
-To construct the Maternal Behavior Qset, an initial pool of more than 150 items was created based on previous descriptions of maternal behavior related to attachment security. Because of the rich detail of exemplars of maternal behavior, Ainsworth’s writings provided a major source of items for this pool (Ainsworth et al., 1971, 1974). The pool also contained other items descriptive of the mother’s behavior that were not thought to be indicative of maternal sensitivity. Faculty and graduate students in developmental psychology who were familiar with research on attachment and with maternal behavior in natural contexts sorted this initial pool to describe the prototypically sensitive mother. Items that proved to be unclear or ambiguous or could not be reliably sorted were discarded. Ninety items were retained to form the final Maternal Behavior Q-set.
The Q-set encompasses descriptions of a mother’s tendency to detect and recognize signals or situations that might require her response, to respond promptly to these situations, and to respond appropriately. These items sample many aspects of maternal behavior, including child care (e.g., “Balances task and baby’s activities during feeding,” “Provides age-appropriate toys”), maternal affect (e.g., “Comments are generally positive when speaking about the baby,” “Seems to resent baby’s bids for attention and signals of distress”), attentiveness (e.g., “Arranges her location so that she can perceive baby’s signals,” “Preoccupied with interview, seems to ignore baby”), interaction style (e.g., “Points to and identifies interesting things in baby’s environment,” “Will often interfere with baby’s ongoing appropriate behavior”), and communication skills (e.g., “Interprets cues correctly as evidenced by baby’s response,” “Response so delayed that baby cannot connect mother’s response to the action that initiated it”).
The aggregate score of 10 judges (faculty and advanced graduate students in developmental psychology) using these items to describe the prototypical sensitive mother was used as the criterion score for each item in the final pool. In this, as in all other Q-sorts described here, the sorters were instructed to place the cards in nine piles of 10 cards each. Interrater reliability for the criterion sort of the final items was satisfactory (all pair-wise r’s > .82). Items rated as most like and rated as most unlike a sensitive mother are listed in Table 1. A mother’s sensitivity score is the correlation between the sort descriptive of her behavior and the criterion sort.
Attachment Behavior Q-sort.-The Waters (1986) Attachment Behavior Q-sort (version 2.0) was used to assess security of attachment between mothers and infants. Scoring procedures are similar to those used with the Maternal Behavior Q-sort. Ten developmental psychologists (faculty and advanced graduate students) familiar with the attachment literature were asked to sort the 90 items in the Q-set to describe the most securely attached 12-month-old infant. A child’s security of attachment score is the item-byitem correlation of the sort of his or her behavior with the aggregate of this criterion sort.
-A pair of observers visited each mother-infant dyad in the home on two separate occasions for about 2 hours per observation. On the first visit, the mother was interviewed and given instructions on the Attachment Behavior Q-sort. In addition to reviewing the procedures detailed in written instructions, the mother was asked to practice by sorting the cards into three piles. During this practice sort, she was encouraged to askfor interpretations of items that seemed unclear to her. Following an interview on the second visit, the mother completed the Parenting Stress Index (Abidin, 1986). Thus, the attentional demands of her baby and the study procedures competed for the attention of the mother for about 20 to 30 min during each of the home visits; the rest of the time was less structured. The mothers completed the attachment Q-sort between the first and second visit. After the second visit, the two observers independently described the infant’s attachment behavior using the Attachment Behavior Q-sort. Both observers also described the mother’s interactive behavior using the Maternal Behavior Q-sort and then performed ratings of the mother’s acceptance, accessibility, cooperation, and sensitivity using procedures developed by Ainsworth et al.(1974).
The Parenting Stress Index (Abidin, 1986) was chosen because it is a wellestablished inventory of stress relevant to parenting. In addition to a total stress score, the Parenting Stress Index contains a Child Domain subscore, which reflects the parent’s perceptions of the challenges presented by the child’s behavior, and a Parent Domain
subscore, which reflects more general stresses of parenting, such as feelings of incompetence and isolation. Abidin (1986) reports alpha reliability coefficients of .95 for the Total Stress score, .89 for the Child Domain, and .93 for the Parent Domain.
-The correlations between the sensitivity and attachment Q-sort scores derived from the two observers across the 40 dyads are presented in Table 2. The sensitivity scores from the two observers were substantially intercorrelated (r = .75, p < .001) and were aggregated to obtain the maternal sensitivity Q-sort scores used in subsequent analyses. On the Ainsworth global rating procedures, the two observers agreed within 1 scale point for 85% of the mothers on the Acceptance-Rejection scale, 95% on the Accessibility-Ignoring Scale, 82% on the Cooperation-Interference scale, and 98% on the Sensitivity-Insensitivity scale. After these ratings were completed, the observers discussed their disagreements and arrived at a consensus rating to be used in the analyses. Consistent with the Ainsworth et al. (1971) results, the ratings on the four scales were highly intercorrelated (r’s from .63 to .88, p < .001 in all cases). Because the scales were highly intercorrelated, were based on similar procedures, and were assessments of the general construct of maternal sensitivity, each mother’s four ratings were averaged to form an aggregate, called the Ainsworth ratings in subsequent analyses.
The correlation between the two observers’ Q-sort attachment scores was .72 (p < .001), and the correlations between the mothers’ attachment security scores and the observers’ security scores were .57 (p < .001) for Observer 1 and .40 (p < .01) for Observer 2 (see Table 2). These intercorrelated scores provided three independent assessments of a single factor using the same assessment instrument but deriving from two quite distinct sources, the mother and trained observers. An aggregated attachment security score was constructed by averaging the observers’ and mothers’ scores.
Measures of maternal sensitivity.
–The mean sensitivity score derived from the Maternal Behavior Q-sort was .73 (SD = .18). The mean aggregated Ainsworth maternal behavior rating was 7.11 (SD = .93). These two assessments of maternal sensitivity were highly intercorrelated (r = .90, p < .001) and independent of maternal age, family income, and parental occupational status. Years of maternal education was associated with both the Ainsworth ratings (r = .47, p < .001) and with the Q-sort maternal sensitivity measure (r = .29, p < .05).
Maternal sensitivity and attachment.
-The mean attachment security score was .40 (SD = .17). Attachment security was not significantly correlated with family demographic variables of income, occupational status, maternal age, and education. Security was correlated with the Q-sort maternal sensitivity scores (r = .52, p < .001) and with the Ainsworth ratings (r = .53, p < .001).
The Q-sort attachment measure was used to identify the 10 most securely attached and 10 least securely attached infants. A series of 90 t tests were then conducted to identify those items in the Maternal Behavior Q-sort that distinguished the two groups (p < .05). Those items are listed in Table 3.
Nonaggregated measures of security and attachment.-Aggregations of the Q-sort measures of attachment and of maternal sensitivity provide the best assessments of the respective constructs because the aggregated scores include data from all coders. For this reason, aggregated scores were used in the primary analyses. Because the home visitors were contributing data to the maternal sensitivity and to the attachment scores, we were concerned about halo effects that could arise if an observer’s global impressions of the dyad affected both Q-sorts. In recognition of this possibility, correlations between each observer’s maternal sensitivity scores and the other observer’s attachment scores were computed to determine if the relations reported in the primary analyses would be substantiated if this potential confound was eliminated. Observer 1’s sensitivity score correlated .42 (p < .01) with Observer 2’s attachment scores, and Observer 2’s sensitivity score correlated .44 (p < .01) with Observer l’s attachment score (see Table 2). The mothers also supplied independent attachment scores; their scores correlated .29 (p < .05) with the Q-sort measure of maternal sensitivity and .29 (p < .05)
with the Ainsworth ratings.
Parenting Stress Index.
-Matemal sensitivity assessed by the Q-sort was unrelated to the parent domain of Parenting Stress Index (r = -.10), but a significant correlation with the Parenting Stress Index child domain (r = -.36, p < .05) indicated that more sensitive mothers tended to describe their child as less difficult than less sensitive mothers. These patterns were confirmed by the analyses of the aggregated rating of maternal sensitivity derived from the Ainsworth global scales, which was not related to the parent domain of the Parenting Stress Index (r = -.16, p > .10) but was correlated significantly with the child domain (r = -.42, p < .01).
Scores on the parent domain of the Parenting Stress Index showed a modest relation to security of attachment (r = -.35, p < .05); however, the child domain scores showed a substantial correlation with attachment security (r = -.65, p < .001). The same patterns of correlations were present, but at lower magnitudes, when the mothers’ data were excluded from the attachment security scores (r’s of -.26, p > .05, and -.54, p < .001, for the parent and child domains, respectively). Children reported by their mothers to be less difficult tended to be judged as more securely attached.
Attachment and maternal sensitivity.–
Attachment theorists (Ainsworth et al., 1971; Bowlby, 1969; Sroufe, 1985) make explicit assertions that maternal sensitivity is the major experiential antecedent to the development of a secure relationship. In contrast to many previous investigations (see Goldsmith & Alansky, 1987), the present study provides a robust empirical endorsement of this assertion. Although there is broad agreement on the importance of sensitivity, theorists differ in the relative emphasis they place on particular aspects of maternal behavior in explaining the mechanism of its impact on attachment. Bowlby (1969), reasoning from an evolutionary-adaptational perspective, hypothesizes that it is the caregivers’ responses to infant signals for proximity and contact that is the principal feature of maternal sensitivity. Ainsworth et al. (1971), reflecting on their experiences with extensive home observations of mothers, conclude that it is not possible to point to any one aspect of maternal behavior as the essential feature of maternal sensitivity. Instead, they cite high interrelations among responses to crying, enjoyment of physical
contact, and contingent pacing of interactions (reviewed in Ainsworth et al., 1978). Ainsworth et al. (1971) surmised that the underlying characteristic related to infant security was the mother’s ability to establish an atmosphere of harmony between herself and her baby.
The pattern of items from the Q-set of maternal characteristics that distinguished mothers of securely attached infants from those of less securely attached infants (see Table 3) are consistent with Ainsworth et al.’s characterization of the relation between sensitivity and attachment. Mothers of highly secure infants noticed their babies’ signals, effectively used these signals to guide their behaviors, knew a lot about their infants, enjoyed cuddling, and spoke positively about their babies. In sharp contrast, mothers of less secure infants were less responsive and more resentful of their babies. The observed negative correlation between the child domain of
the Parenting Stress Index and attachment security indicates that mothers of secure infants perceive their infants as generating less stress than do mothers of less secure infants. This finding lends additional credence to Ainsworth’s conclusion that the establishment of a harmonious relationship is the essential feature of maternal sensitivity.
-The observers in our study provided assessments of the mothers’ sensitivity and contributed to the infants’ security score. Rather than showing a functional link between these two constructs, the parallel ratings could reflect the fact that the observers’ descriptions of both the mother and the infant are confounded by a common
factor-the observer’s social perception or evaluation of the dyad. There are no perfect solutions to this problem since even independent observers probably would share common biases. However, we have reason to be confident that such biases do not account for a substantial portion of the relation between scores on the maternal sensitivity and attaching Q-sorts.
First, the attachment security scores produced by one observer were significantly correlated with the maternal sensitivity scores produced by the second observer, and vice versa. These correlations are between measures derived from independent sources. The finding that the correlations of the individual measures are lower than the correlation between the aggregates is consistent with the principles of aggregation (Rushton, Pressley, & Brainerd, 1983). Second, the Q-sort measure of maternal sensitivity was independent of social class and education. A correlation between these variables would have been expected if the observers were biased because they were more comfortable in some homes than in others. Third, the Ainsworth rating scales also were correlated with attachment (r = .53), at the same magnitude as the sensitivity scores derived from the Maternal Behavior Q-sort (r = .52). Because the opportunity for observer bias would be greater on the Ainsworth 9-point scales than for the forced distribution of the Q-sort, similar relations between these two data sets and attachment argue against a response-bias explanation. A final reason to discount an observer-bias explanation is that, as an inspection of Tables 1 and 3 will reveal, there is no simple one-to-one correspondence between high- and low sensitivity items on the criterion sort and those maternal sensitivity items that distinguish most securely and least securely attached infants. This finding suggests that the concomitants of secure attachment fit a pattern more complex than a simple abstract ranking of the behavioral characteristics of sensitivity. That is, a simple bias would be expected to produce a relatively straightforward correspondence between the characteristics of mothers of the most secure infants and those of the prototypically sensitive mother.
Related to this issue is the observation that security of attachment scores based on maternal Q-sorts were not as highly correlated with maternal sensitivity scores as were attachment scores derived from observers. The difference between these correlations is consistent with the argument that the relation between observer-based attachment and sensitivity scores is a product of common observer bias if one assumes that motherbased attachment scores are a more valid indicator of the relationship. However, further inspection of the pattern of data suggests instead that the low correlation between observer-based sensitivity scores and motherbased attachment scores is reasonably attributed to larger error variance in the Q-sorts done by the mothers. Recall that the aggregate attachment score is constructed by summing the scores produced by the sorts of both the mothers and the observers. The correlation of maternal sensitivity scores with this aggregate attachment score (r = .52) is much more similar to the correlation of sensitivity with each of the observer-based attachment scores (r’s = .58 and .46) that with the correlation of sensitivity with the mother-based attachment scores (r = .29) (see Table 2). This pattern emerged from the analyses despite the fact that the aggregate is an equal combination of observer- and mother-based attachment scores, suggesting higher error variance in the mother-based scores. More reliable ratings by the mothers might have been obtained with more extensive familiarization procedures and the aggregation of repeated sorts.
Future studies might use an assessment of sensitivity based on behavioral coding similar to Smith and Pederson (1988) to provide independent validation of the Q-sort measure. The Maternal Behavior Q-sort also provides a means to compare the attachment relationship based on behavior in the Ainsworth-Wittig (1969) Strange Situation with measures of sensitivity from the home.
The relatively strong relation found in this study between maternal sensitivity and attachment supports our suggestion that sensitivity is best examined under circumstances where the mother is unable to devote her attention solely to the demands of her infant. Smith and Pederson (1988) observed interactions under analogous conditions and also found a relation between sensitivity and attachment that surpassed that found in much earlier research. This pattern of results suggests that observations of mothers and their infants in relatively low-demand situations are unlikely to uncover variability that is relevant to studies of individual differences in
Advantages of the Q-sort methodology.
-We believe that the Q-sort procedures offer methodological and conceptual advantages over ratings scales or behavior coding. Admittedly, it is much faster for observers to complete rating scales; however, the specific nature of the Q-set items forces observers to attend to the details of the mothers’ and infants’ behaviors. In our experience, these effects on the observers are analogous to doing ratings of proximity seeking, contact maintenance, resistance, and avoidance in coding the Strange Situation. These interactive ratings are not essential to the Strange Situation classifications, but completing the ratings provides a framework for viewing the Strange Situation. In addition, the Q-sort procedures, being grounded in behavioral observations, share the descriptive advantages of behavior codes. Unlike behavior codes, the Q-sort procedures allow the observers to provide differential weightings for the psychological significance of the behaviors.
The Q-sort methodology goes beyond the production of a single abstract metric of sensitivity or attachment and forces an empirically based elaboration of the underlying constructs. In the present study, we described maternal sensitivity in terms of contingent pacing in response to communicative signals, prompt and effective reactions to distress, and maternal availability. Finally, as Waters (1985) has pointed out, Q-sorts describe behavior independent of any particular concept that might be of interest. This advantage is essential for the evolution of theoretical constructs. For example, in the present study, there appear to be subtle differences in our concept of maternal sensitivity and maternal characteristics that distinguished mothers of infants differing in security. Mothers of highly secure infants were characterized not only by their responsiveness to infant signals but also by their knowledge and enjoyment of the infant.
In summary, the results of the present study support the long-standing claim of a central relation between maternal sensitivity and infant attachment. Moreover, the findings reinforce the feasibility and importance of studies of the origins of attachment that focus on interaction in the home. At the same time, they bolster our speculation that assessments of maternal sensitivity are most likely to reflect meaningful individual differences under circumstances where multiple demands are put on the attentional resources of the mother.