The Relation Between Coordinated Interpersonal ... - Springer Link

Journal of Psycholinguistic Research, Vol. 32, No. 5, September 2003 (© 2003)

The Relation Between Coordinated Interpersonal Timing and Maternal Sensitivity in Four-Month-Old Infants Amie Ashley Hane,1,3 Stanley Feldstein,2 and Valerie H. Dernetz2 The relation between mother–infant coordinated interpersonal timing, an automated microanalytic measure of dyadic vocal coordination, and maternal sensitivity was explored. Thirty-five mothers and their developmentally normal 4-month-old infants were audio-recorded during a 20-min laboratory vocal interaction session, that was later analyzed for degree of vocal coordination. Maternal Sensitivity ratings (Ainsworth & Bell, 1969) were based on a video-taped 45-min unstructured laboratory interaction period. A significant curvilinear relation between the degree to which mother coordinated her noninterruptive co-occurring speech to that of her infant was found and revealed that mothers highest in sensitivity were characterized by moderate levels of coordination. Examining mother-infant interaction at the specific behavioral level, while incorporating tests of nonlinear trends, may provide important information about the nature of sensitive parenting. KEY WORDS: Coordinated interpersonal timing; mother–infant interaction; maternal sensitivity; vocal coaction.

INTRODUCTION The mother–infant attachment relationship evolves via reciprocal dyadic exchanges. These exchanges involve a strategy in which infants emit cues that indicate their need for safety, comfort, and social interaction, and mothers

Presented as a poster at the Society for Research in Child Development, 1999, Albuquerque, NM. We are indebted to the team of Research Assistants of the Interpersonal Communication Laboratory for their help in collecting data. Special thanks to Shyoon Loo for his careful attention to detail in coding the CIT data. We are also grateful for the time and commitment all of our participants dedicated to this project. Thank you to M. Demorest, D. Teti, and M. Jasnow for their contributions to this work while serving on A. Hane’s master’s thesis committee. 1 University of Maryland, College Park 2 University of Maryland, Baltimore County. 3 To whom all correspondence should be addressed: email: [email protected] 525 0090-6905/03/0900-0525/0 © 2003 Plenum Publishing Corporation

526

Hane, Feldstein, and Dernetz

accurately perceive and respond to these cues (Bowlby, 1969). Inherent to attachment theory is the notion that the early attachment relationship results in the formation of internal working models of behavior that serve to organize the infant’s future social interactions (Main, Kaplan, & Cassidy, 1985). Thus the early parent–child relationship is considered crucial to the long-term social outcomes of the child (Bowlby, 1969). Given the importance of the early mother–child attachment quality, Bowlby stated that the conditions in the early rearing environment that contribute to the development of a secure attachment relationship must be identified. Maternal Sensitivity Ainsworth, Blehar, Waters, and Wall (1978) were the first to study the antecedents of mother–infant attachment. After 70 h of home observation of their 26 mother–infant dyads, Ainsworth et al. discovered that ideal maternal behaviors reflected a high degree of maternal sensitivity, or “sensitive responsiveness to infant signals and communications” (Ainsworth et al., 1978, p. 152). Such maternal sensitivity cultivates synchronous, reciprocal, and jointly satisfying mother–infant interactions. These interactions, in turn, foster the development of a secure attachment relationship. In contrast, maternal insensitivity—maternal failure to accurately perceive and respond to her infant’s signals—results in poor timing, mutually unsatisfying interaction, and, ultimately, insecure attachment (Ainsworth et al., 1978). An abundance of research since Ainsworth’s seminal findings has somewhat clarified the role of maternal sensitivity in the evolving attachment relationship. The effect size reported by Ainsworth et al. (1978) has not been replicated, and what has emerged instead is a weak-to-moderate relationship between maternal sensitivity and security of attachment (De Wolff &van ljzendoorn, 1997; Goldsmith & Alansky, 1987). A more recent metaanalysis of maternal sensitivity and attachment reports the combined effect size of 66 studies to be moderate [r(1,097) ⫽ .24], based on Cohen’s (1988) conventional criteria for effect-size (De Wolff &van Ijzendoorn, 1997). Researchers claim that their modest but significant effect sizes for maternal sensitivity and attachment indicate partial replication of the original findings of Ainsworth et al. (1978). We propose here that one critical component of mother–infant interaction that is often missing from global sensitivity measures is the infant’s contribution to the behavioral exchange. Most measures of maternal sensitivity gauge the effectiveness and appropriateness of maternal interventions without explicitly measuring infant behaviors. The use of these unidirectional sensitivity measures persists, despite evidence that healthy development is contingent upon a “goodness of fit” between a child’s temperamental style and the

The Relation Between Coordinated Interpersonal Timing ans Maternal Sensitivity

527

environment (Thomas & Chess, 1977). Indeed, the degree to which infants demand parental intervention varies based on temperament, with some easy infants requiring little more than quiet maternal stimulation, whereas other more challenging infants demand a high level of maternal involvement. Thus what is sensitive or appropriate for one child may seem intrusive or otherwise inappropriate for another. The goodness of fit model informs us that interaction variables that consider responsiveness to the child in terms of contingency (e.g., synchrony or reciprocity) in behavioral exchange must consider the role of the infant an integral part of the interaction measure. Research that analyzes interaction using measures of behavioral and vocal contingency demonstrates that the ideal “fit” between parent and child may be characterized by moderate, rather than high levels of contingent responsiveness, a finding that is in direct contrast to studies that have relied on global measures of sensitivity. Temporal Aspects of Mother–Infant Interaction The work of Belsky et al. (Belsky, Rovine, & Taylor, 1984; Belsky, Taylor, & Rovine, 1984; Isabella & Belsky, 1991; Isabella, Belsky, & von Eye, 1989) in behavioral synchrony provides empirical support for the existence of a curvilinear relation between interactional timing and attachment outcomes. In one of their studies of mother–infant synchrony, Isabella et al. (1989) used a time-sampling observational system in operationalizing mother–infant interactions to test the hypothesis that attachment outcomes in infants differ depending on the relative degree to which their interactional histories were characterized by synchronous or asynchronous exchange. Secure dyads at 12 months were characterized by synchronous temporal co-occurrences at 1 and 3 months. In contrast, attachment insecurity at 12 months was predicted by asynchronous mother–infant interactions, that is, was characterized by either under- or over-involvement. In a replication of this study, Isabella and Belsky (1991) again found that securely attached infants at 12 months engaged in well-timed, reciprocal interactions at 3 and 9 months. Insecure dyads were found to be characterized by unresponsive or intrusive maternal behaviors at 3 and 9 months. Research examining the role of vocal temporal coordination in attachment has revealed a similar pattern. Feldstein et al. (1993) explored the role of dyadic vocal contingency, which they call coordinated interpersonal timing (CIT), in infant social outcomes. Briefly, CIT occurs when the temporal pattern of each person’s vocal behavior is predictable from that of the other (Feldstein et al., 1993). In a study of mother–infant CIT and attachment, Jaffe, Beebe, Feldstein, Crown, and Jasnow (2001) found a curvilinear relationship between CIT at 4 months and attachment security at 12 months. Specifically,

528


vocal interactions characterized by high coordination (i.e., over-sensitivity or hypervigilance on the part of mother) as well as low coordination (i.e., undersensitivity or unresponsiveness on the part of mother) predicted insecure attachment. Secure attachments at 12 months were predicted by intermediate levels of coordinated timing at 4 months. These findings suggest that temporal measures play an intricate role in sensitive parenting. As stated by Schaffer (1977, p. 214), “What appears to matter most about ‘successful’ mother–infant interactions is above all the temporal integration of the two partners’ responses, and when we talk about mothers’ sensitivity it is often this temporal characteristic that we have in mind.” The purpose of the study presented here was to examine the role of CIT in maternal sensitivity. Indeed, both sensitivity and CIT have been demonstrated to be important antecedents to the mother–infant attachment relationship. However, the nature of their relations to mother–infant attachment are quite different, with the former yielding consistent linear trends in the data, and the latter relating to attachment in a curvilinear fashion. It is suggested here that coordinated interpersonal timing is one important component of maternal sensitivity, and that the nature of this relationship is curvilinear, such that moderate levels of vocal contingency (i.e., CIT) are associated with high levels of global sensitivity in 4-month-old infants. In contrast, high or low levels of CIT should be associated with relatively low levels of sensitivity.

EXPERIMENT Method Participants Data for this study were drawn from a larger longitudinal project that assesses mothers and infants across a variety of social and cognitive domains in the laboratory and home during the infants’ fourth, twelfth, and twenty-fourth month. The project employs the following eligibility criteria: English as their primary speaking language, married or living with the infant’s father, and an infant who is developmentally normal to the best of their knowledge. Thirty-five mothers and their 4-month-old developmentally normal infants were recruited from fliers and ads in local papers. Infants averaged 16.74 weeks, SD ⫽ 1.32, and 19 were female. All infants were screened for developmental normalcy using the Yale Neuropsychoeducational Scales (Shaywitz, 1982). Seventy-four percent (n ⫽ 26) were primiparous and the remainder were second born. Mothers averaged 30.05 years (SD ⫽ 5.66) and all were living


529

with the infant’s father. Roughly 83% percent of the sample were Caucasian (n ⫽ 29), 9% (n ⫽ 3) were African American, and the remaining 6% (n ⫽ 2) were of other ethnicity. Mothers were predominantly middle class; only one reported an annual household income of less than $10,000. Participation in the study was voluntary, and all the mothers signed informed consent forms for themselves and their infants. Measures Coordinated interpersonal timing. The audio-taped vocal behaviors from a 20-min mother–infant interaction session for each mother–infant dyad were analyzed by the Automatic Vocal Transaction Analyzer (AVTA) System (Jaffe & Feldstein, 1970). AVTA is a computer system comprised of an analog-todigital converter that determines only whether each partner in a dialogue is or is not speaking, and ignores the frequency or intonational contours of the speech. However, a person with unimpaired hearing sets the vocal intensity level that AVTA is to accept as a speech sound. The purpose of this is to generate a description of the time patterns of vocal behavior that would be perceived by a human listener. Once the audio signals have been detected, AVTA transforms and digitally stores them. The system programs then convert these codes into a set of vocal states for each person (Feldstein & Welkowitz, 1987; Jaffe & Feldstein, 1970) and provides descriptive statistics for these parameters for both a prespecified time unit and for the total time of the vocal interaction. For the purpose of calculating a time-series regression analysis, the prespecified time unit was five seconds, which had been previously determined to be long enough to include several occurrences of one or more of the states (Feldstein et al., 1993). The vocal states are vocalizations, defined as a vocal sound from either person that is bounded by the silences of the same person and is uninterrupted by silence longer than a quarter of a second; pauses, a mutual silence that is commenced and terminated by the vocalizations of the same person; switching pauses, an instance of simultaneous silence that has been initiated by one person, but terminated by the other; noninterruptive simultaneous speech, which occurs when both persons speak at the same time, but the person who had the turn before the simultaneous speech segment occurred continues to have the turn after the other person terminates (thus, the person speaking was never really interrupted); and lastly, interruptive simultaneous speech, which occurs when the person who has the turn loses the turn to the person who initiated the simultaneous speech. To obtain the degree to which the two partners coordinate their vocal states with each other, a time-series regression (TSR) analysis (Ostrum, 1978; Warner, 1988) was computed for each of these five states. The TSR partialed

530


the mothers’ and infants’ self-influence (i.e., autocorrelations) to determine the degree to which their contributions to the dialogue were mutually coordinated (i.e., cross-correlations). The partial cross-correlation coefficients (or their squares) represent each person’s degree of coordination. Five represent the mother’s degree of coordination with the infant (CIT-M), and another five represent the degree of coordination of the infant with the mother (CIT-I). The higher the coefficient, the greater the coordination. Because the computer completes all coding, this measure is objective and requires research assistants only to familiarize themselves with the computer procedures (Feldstein & Welkowitz, 1987). For this study, all AVTA processing was completed by a supervised undergraduate research assistant. The assistant achieved sound intrarater reliability across 10 dyads with correlation coefficients ranging from .90 to .95.

Maternal Sensitivity To assess degree of maternal sensitivity, a modified version of the Ainsworth System for Rating Maternal-Care Behavior (Ainsworth & Bell, 1969) was used to code videotapes obtained from a 45-min feeding/unstructured interaction period. In its entirety, this measure consists of 22 9-point rating scales, which rate behavior across seven domains. Scale points 1, 3, 5, 7, and 9 are anchored in detailed behavioral descriptions. The higher the composite score across all 22 ratings, the more sensitive are the mother’s behaviors to her infant. Because these scales are based on detailed observations of mother– infant interaction as they occur in the home, not all of the scales are relevant to laboratory assessments. Thus, of the original 22 rating scales, only 16 were used to code laboratory sessions. These 16 scales are subdivided into six separate domains: General Attitude of Mother Toward Baby and Her Role as A Mother, which includes ratings for maternal perception of her baby, maternal delight in baby, and maternal acceptance of baby; Feeding, which is rated based on synchronization of mother’s interventions with baby’s rhythms, determination of amount of food at the end of feeding, and mother’s synchronization of rate of feeding to baby’s pace; Availability and Interaction, which rates total availability of caregiver, amount of interaction offered by mother, and appropriateness of mother’s initiations or interactions; Physical Contact, which rates the amount of physical contact and the quality of contact in holding baby; Response to Crying, which rates the effectiveness of mother’s responses to baby’s crying; and, lastly, Social Contact, which rates amount of visual and auditory contact, and frequency and appropriateness of play interactions. A global maternal sensitivity score


531

was obtained for each dyad by obtaining the mean rating across the 16 scales. All sensitivity coding was completed by two of the authors (A. A. H. and V. H. D.). The independent ratings of the two raters on five dyads chosen at random yielded an intraclass correlation (Haggard, 1958) of .91. Both sensitivity raters remained blind to the 20-min vocal data until the coding was complete. Intercorrelations of the modified maternal sensitivity scale for the entire sample of mothers and infants used in this study ranged from .08 to .87. The lack of agreement between certain scales (e.g., amount of interaction and appropriateness of maternal initiations) was hypothesized to be a function of the inclusion of scales that rate both frequency as well as quality of maternal behaviors. Thus those items that clearly rated the quality of mother–infant interaction were isolated from those that measured only frequency or amount of interaction. Items that were not clearly placed in either category were eliminated. Eliminated scales included the scales that rated feeding and the maternal availability scale, because these required both qualitative and quantitative judgements. The newly created Frequency of Maternal Interventions subscale consisted of the following five scales: amount of interaction, amount of physical contact, amount of visual contact, amount of auditory contact, and frequency of play interactions. Seven scales were placed in a Quality of Maternal Interventions subscale. These included: maternal perception of the infant, delight in the infant, maternal acceptance, appropriateness of interactions, effectiveness of response to crying, quality of physical contact, and appropriateness of play interactions. Subsequent intercorrelations between these two new subscales and the composite sensitivity score revealed that this newly created Frequency of Maternal Interventions subscale shares only 29% of its variability with the Quality of maternal interventions subscale [r(32) ⫽ .54, p ⬍ .01]. This suggests that mothers who engaged their infants in more frequent interactive behaviors were not necessarily characterized by engaging their infants in highly appropriate stimulation. Frequency of maternal interventions was strongly related to overall sensitivity scores [r(32) ⫽ .78, p ⬍ .001], as was quality of maternal interventions [r(32) ⫽ .92, p ⬍ .001]. Procedure Mothers and infants who met the inclusion criteria were scheduled for a laboratory visit between their infant’s fifteenth and eighteenth week. The age of 4 months was chosen specifically for this study, because developmentally, 4-month-olds are quite capable of (and interested in) partaking in vocal faceto-face interactions with their caregivers. Also, limited prehension and gross motor development prevents infants at this age from engaging in many of the

532


behaviors that come later in infancy, such as grasping for objects or crawling around the room. It is for these reasons that this age has been referred to as the age of “interactional primacy,” or the stage in which infants are more interested in human interaction rather than didactic exchanges (Bakeman, Adamson, Brown, & Eldridge, 1989). For the 20-min vocal interaction session, mother and infant were fitted with wireless clip-on microphones and escorted to a simulated living room environment in which the infant was placed in an infant seat on a low table. Mothers were instructed to engage their baby vocally as they would at home, but asked to leave the child in the seat to ensure adequate distance between the microphones (to reduce the possibility of auditory feedback into each other’s microphones). Ordinarily, this session lasted 20 min and was audio-taped and also recorded by video- and wall-mounted cameras in the room and through a one-way mirror. For eight dyads, however, the session was terminated early because of infant protest. But ample timing data were generated by these prematurely terminated episodes (ranging in duration from 10.5 to 19.5 min, M ⫽ 14.44, SD ⫽ 3.32), thus permitting their inclusion in the analyses. For the unstructured/feeding interaction session, mother and infant were escorted to a different room similarly furnished. Mothers were instructed to engage the child as they would at home, taking advantage of the surroundings whichever way they chose, and feeding or changing the child as they deemed necessary. This session lasted exactly 45 min and was also audio- and videotaped. Videotapes obtained from this session were subsequently coded for maternal sensitivity. The order of the 20- and 45-min interaction sessions was counterbalanced. However, modifications to the counterbalancing system were employed when the order of the episodes had to be tailored to the infant’s own schedule. Following both interaction sessions, the mothers were asked to complete a battery of questionnaires. With the exception of a demographic sheet and the Yale Neuropsychoeducational Scales (Shaywitz, 1982) used for developmental screening, data generated from these questionnaires were not utilized for this study.

RESULTS Preliminary Analyses A Fischer’s r to z⬘ transformation was used to normalize all coefficients of coordination before data analysis. Table I presents the descriptive statistics of the coefficients of timing; Table II presents a summary of the maternal sensitivity data.


533

Table I. Coefficients of Coordinated Timing Vocal parameter

n

M

SD

34 34 34 33 34

.49 .53 .50 .57 .59

.38 .58 .69 .25 .13

34 33 34 33 34

.50 .51 .53 .56 .57

.27 .38 .37 .19 .14

Mother Vocalizations Pauses Switching Pauses Noninterruptive SS Interruptive SS Infant Vocalizations Pauses Switching Pauses Noninterruptive SS Interruptive SS

Note: The entries in the column labeled M are the average degrees of temporal coordination (r), which have been transformed from z⬘ scores back into original form for ease of interpretation. SS, Simultaneous speech.

Bivariate correlations were run to test for order effects and revealed no significant relation between the order of the episodes the dyads were assigned and maternal sensitivity (overall, quality, and frequency) or degree of CIT. Pearson product-moment correlations were calculated among overall maternal sensitivity ratings, maternal sensitivity frequency scores, and maternal sensitivity quality scores and the following demographic information: mother age, infant age, infant gender, and parity. A significant positive correlation between frequency of interaction and infant gender was found [r(32) ⫽ .350, p ⬍ .05], revealing that the mothers of female infants were more likely to engage their infants in interaction than were mothers of males. Parity and infant

Table II. Maternal Sensitivity Scores Score Maternal Sensitivity Overall Quality Subscale Frequency Subscale

n

M

SD

35 35 35

6.42 6.11 6.80

1.36 1.73 1.32

534


gender were not significantly correlated with overall maternal sensitivity or maternal sensitivity quality scores. Thus neither gender nor parity was treated as a covariate. Relation Between CIT and Maternal Sensitivity One hierarchical multiple regression analysis was employed to test the hypothesis that moderate degrees of maternal CIT are associated with high levels of maternal sensitivity. Implied by this hypothesis is the expectation that each of the parameters that make up CIT is curvilinearly related to sensitivity. A hierarchical model was used for two reasons. First, previous research examining the CIT of mother–infant interaction has found that each parameter, when considered separately, reveals important distinctive information about the nature of mother–infant vocal exchange. Second, the procedure for the testing of a quadratic trend in multiple regression requires that scores on each predictor be squared and entered into the regression analysis immediately following the entry of the variable’s linear component. Entering the variables in this fashion yields the unique contribution of the quadratic trend, as the variance associated with a given predictor’s linear relation to the criterion has been removed (for a complete description of this procedure, see Cohen & Cohen, 1983). Thus a hierarchical model was used to ensure that each parameter’s curvilinear component (i.e., its respective squared value) entered the model directly following its linear counterpart. Results of this analysis yielded a significant curvilinear relation between CIT and coordination of noninterruptive simultaneous speech [ri (29) ⫽ .384, F(1,22) ⫽ 4.63, p ⬍ .05, r2 ⫽ .15] and reveals that mothers who coordinated their noninterruptive simultaneous speech to that of their infants in moderate levels were characterized by high levels of overall sensitivity. In contrast, mothers who were characterized by high or low levels of the coordination of noninterruptive simultaneous speech were more likely to be low in sensitivity. No significant linear trends were found (Fig. 1). Exploratory Analyses To explore the role of infant sex as a moderator in the relation between CIT and maternal sensitivity, overall sensitivity scores were regressed onto the following predictors (in order): infant sex, vocal parameters of CIT, and the product of infant sex and CIT. No significant linear or interactions effects were found. Thus the relation between CIT and sensitivity is not dependent on infant sex.

4

The incremental correlation coefficient (ri) is the relation between the independent variable and the dependent variable controlling for all prior variables entered into the regression equation.


535

Fig. 1. The quadratic trend of the relation between global maternal sensitivity mean scores and maternal noninterruptive simultaneous speech coordination.

Two post-hoc regression analyses were computed to explore the relation between CIT and the Frequency and Quality of Maternal Interventions scores. In the first analysis, Frequency of Maternal Interventions scores was regressed on each of the five coefficients of timing and their respective squares. No significant linear or curvilinear relation was found. The second regression analysis using the Quality of Maternal Interventions ratings as the criterion were employed to explore the linear and curvilinear relations between quality of interaction and CIT. A significant curvilinear relation between noninterruptive simultaneous speech and Quality of Maternal

536


Interventions emerged [ri1(29) ⫽ .38, F(1,22) ⫽ 4.63, p ⬍ .05, r2 ⫽ .15]. Given the high degree of correlation between the overall sensitivity and Quality of Maternal Interventions scores, the shape of this trend is nearly identical to that of the quadratic trend of maternal noninterruptive simultaneous speech and overall sensitivity illustrated in Fig. 1.

DISCUSSION The findings of this study provide only partial support for the hypothesis that the degree to which mother coordinates her vocal behavior to that of her infant is related—in a curvilinear fashion—to maternal sensitivity. Indeed, the significant quadratic trend of maternal noninterruptive simultaneous speech coordination and maternal sensitivity conforms to our prediction, in that those mothers who were highest in sensitivity were characterized by moderate levels of noninterruptive simultaneous speech coordination. The exploratory expectation that the infant’s degree of vocal coordination to the mother would also relate to maternal sensitivity was not supported. This lack of support may be a function of limited statistical power and/or the nature of the criterion variable employed, which focuses almost exclusively on maternal behavior. Results of the remaining post-hoc analyses that explored the contributions of CIT to separate domains of maternal sensitivity generate some interesting questions regarding current conceptualizations of sensitive parenting. The findings of this study suggest that a moderate degree of synchronization of vocal coaction in the mother–infant vocal exchanges is associated with high levels of globally rated maternal sensitivity, whereas both low and high degrees of vocal synchrony of vocal coaction are associated with less sensitivity. Thus a mother who fails to interrupt her infant with brief interjections (e.g., reinforcing praise or comforting vocal soothing) to the same degree that her infant does so to her (be it more or less) was more likely to be rated as less sensitive, both in terms of overall sensitivity and the quality of her interventions. In contrast, a moderate level of noninterruptive simultaneous speech timing characterizes a vocal exchange in which a mother times her interjections to those of her infant, thereby engaging her infant in a smoothly paced, moderately predictable vocal interaction that is associated with high levels of sensitivity. The form of this curvilinear relation is consistent with research in CIT that has found that moderate levels of vocal (Ashley et al., 1997; Feldstein et al., 1993; Jaffe et al., 2001) and behavioral synchrony (Belsky et al., 1984a; Belsky et al., 1984b; Isabella et al., 1989; Isabella &Belsky, 1991) are associated with positive social outcomes. Although the significant curvilinear association between noninterruptive simultaneous speech coordination and sensitivity is theoretically plausible, it is


537

essential to note that any interpretations drawn from this and any finding of this study must be considered with caution. For audio data to be analyzed by the AVTA system, auditory feedback between mother and infant microphones had to be eliminated. Thus mother and infant were required to interact in a face-to-face situation that prohibited holding during recording. Also, the Ainsworth sensitivity measure (Ainsworth & Bell, 1969) was created for use in the home. We modified the measure to rate laboratory interaction sessions. To the best of our knowledge, such an adaptation to Ainsworth’s measure has never before been implemented. Thus the extent to which our finding would replicate under more natural mother–infant interaction conditions remains uncertain. The broad conceptualization of sensitivity reflected in Ainsworth’s scales served as the basis for the post-hoc analyses employed in this study. Included in Ainsworth’s measure are items that tap the quality of mother–infant interaction, as well as scales that rate the frequency, or amount, of interaction offered by mother. Intuitively, it does not seem to follow that more stimulation or interaction indicates more sensitive parenting. Our initial suspicion that this was not the case was somewhat supported by the lack of strong agreement between the newly created frequency and quality domains. The fact that the relations between CIT and quality of interaction were quite similar (in size of effect and shape of the trend) to the findings of CIT and sensitivity overall and theoretically consistent with previous findings—while regression of the frequency domain on the coefficients of CIT failed to yield significance—reveals the potential benefit of tapping sensitivity at a more specific behavioral level and informs us that high quality, not high frequency, of maternal interventions is associated with optimal levels of mother–infant vocal timing. Additionally, although the post-hoc exploration of interaction effects of CIT and infant sex on sensitivity yielded no significant results, the role of gender as a moderator in this relation should be explored in subsequent analyses. In the present analysis mothers engaged female infants with more interaction than male infants. Further, studies have revealed that the nature of sensitive parenting is different for male versus female infants. Taken together, these findings provide suggestive evidence that maternal sensitivity is indeed a complex construct, requiring only modest levels of maternal involvement in some instances, but perhaps high levels of other types of involvement, in other instances. What surfaces from this research is a further confirmation of the emergent notion that sensitive mothering includes many facets, each of which predicts infant social outcomes in different ways. Perhaps sensitivity is best conceived of in van den Boom’s (1997, p. 593) terms: “It would be more fruitful to think of sensitivity not as a parenting dimension that exists apart from other dimensions, but rather, as permeating all interactive behavior. The sensitive parent must package her social interactive behavior in

538


such a way in the interaction flow that it will promote rather than interrupt the exchange.” The results of our study reveal that future research endeavors that are more narrow in scope and open to investigation of both linear and curvilinear trends should aid the field in further defining the complexities of sensitive parenting.

REFERENCES Ainsworth, M. D. S., & Bell, S. M. (1969). Some contemporary patterns of mother-infant interaction in the feeding situation. In A. Ambrose (Ed.), Stimulation in early infancy (pp. 133–170). New York: Academic Press. Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns of attachment. Hillsdale, NJ: Erlbaum. Ashley, A. M., Feldstein, S., Hoffhines, V. L., & White, K. (1997). Mother-toddler attachment and coordinated interpersonal timing during play interactions: A preliminary analysis. Poster presented at the Biennial Meeting of the Society for Research in Child Development, Washington, DC. Bakeman, R., Adamson, L. B., Brown, J. V., & Eldridge, M. (1989). Can early interaction predict? How and how much? In M. Bornstein & N. Krasnegor (Eds.), Stability and continuity in mental development (pp. 235–248). Hillsdale, NJ: Erlbaum. Belsky, J., Rovine, M. J., & Taylor; D. G. (1984a). The Pennsylvania Infant and Family Development Project III: The origins of individual differences in infant-mother attachment— Maternal and infant contributions. Child Development, 55, 718–728. Belsky, J., Taylor, M. J., & Rovine, M. (1984b) The Pennsylvania Infant and Family Development Project II: Development of reciprocal interaction in the mother-infant dyad. Child Development, 55, 706–717. Bowlby, J. (1969). Attachment and loss: Vol. 1. Attachment. New York: Basic Books. Cohen, J. (1998). Set correlation and contingency tables. Applied Psychological Measurement, 12, 425–434. Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum Associates. De Wolff, M. S., & van IJzendoorn, M. H. (1997). Sensitivity and attachment: A meta-analysis on parental antecedents of infant attachment. Child Development, 68, 571–591. Feldstein, S., Jaffe, J., Beebe, B. et al. (1993). Coordinated interpersonal timing in adult-infant vocal interactions: A cross site replication. Infant Behavior and Development, 16, 455–470. Feldstein, S., & Welkowitz, J. (1987). A chronography of conversation: In defense of an objective approach. In A. W. Seigman & S. Feldstein (Eds.), Nonverbal behavior and communication (2nd ed.). Hillsdale, NJ: Erlbaum. Goldsmith, H. H., & Alansky, J. A. (1987). Maternal and infant predictors of attachment: A metaanalytic review. Journal of Consulting and Clinical Psychology, 55, 805–816. Haggard, E. A. (1958). Intraclass correlation and the analysis of variance. New York: Dryden. Isabella, R. A., & Belsky, J. (1991). Interactional synchrony and the origins of infant-mother attachment: A replication study. Child Development, 62, 373–384. Isabella, R. A., Belsky, J., & von Eye, A. (1989). The origins of infant mother attachment: An examination of interactional synchrony during the infant’s first year. Developmental Psychology, 25, 12–21. Jaffe, J., Beebe, B., Feldstein, S., Crown, C. L., & Jasnow, M. D. (2001). Rhythms of dialogue in infancy. Monographs of the Society for Research in Child Development, Serial No. 265, 66, No. 2.


539

Jaffe, J., & Feldstein, S. (1970). Rhythms of dialogue. New York: Academic Press. Main, M., Kaplan, N., & Cassidy, J. (1985). Security in infancy, childhood, and adulthood: A move to the level of representation. Monographs of the Society for Research in Child Development, 50(1–2). Ostrom, C. (1978). Time-series analysis: Regression techniques. Beverly Hills, CA: Sage University Papers. Schaffer, H. R. (1977). Studies in mother-infant interaction. New York: Academic Press. Shaywitz, S. E. (1982). The Yale Neuropsychoeducational Assessment Scales. Schizophrenia Bulletin, 8, 360–424. Thomas, A., & Chess, S. (1977). Temperament and development. New York, NY: Brunner/Mazel. Van den Boom, D. C. (1997). Sensitivity and attachment: Next steps for developmentalists. Child Development, 64, 592–294. Warner, R. (1998). Rhythm in social interactions. In J. E. McGrath (Ed.), The social psychology of time: New perspectives (pp. 63–88). Beverly Hills, CA: Sage Publications.