Results and Discussion

The purpose of this study was to develop and field-test an objective means for measuring infant--parent social interaction. First, operational definitions of infant, parent, and dyadic social interactive behaviors were developed and evaluated with the assistance of a national panel of experts. Evaluations suggested the interactive behavior definitions were valid. The operational definitions were incorporated into a systematic procedure for measuring aspects of infant--parent social interaction that minimizes the potential for observer bias. Preliminary reliability data with a population ranging in age from birth to 31 months and replication reliability data with a population ranging from 18 to 36 months developmental age suggest the instrument meets appropriate reliability criteria for future empirical and clinical use.

Dyadic Patterns

The instrument provides an objective means for examining patterns of infant-parent social interaction. The significance of specific infant, parent, and dyadic behaviors lies not in the frequency of their occurrence, but in the dyadic profiles that emerge and are demonstrated over time as interactive behaviors co-occur, re-occur, and co-vary (Mahoney, Fors, & Wood, 1990). Specifically, the frequency of parent directiveness is relatively meaningless as a singular behavioral index. However, in conjunction with patterns of interactive behaviors such as parent responsiveness, infant participation, infant signal clarity, and theme continuity, information about parent directiveness becomes much more useful. For example, parent directiveness that repeatedly occurs in conjunction with parent contingent responsivity and infant participation is likely to reflect parent scaffolding and support infant learning. However, an interactional pattern characterized by high levels of parent directiveness and low levels of infant participation or high levels of parent intrusiveness is likely to have a negative impact on child development.

Independently, the significance of infant initiation frequency is somewhat obscure. However, when patterns of infant initiation are analyzed for co-occurrence with parent facilitation, parent contingent responsivity, and dyadic theme continuity, the significance of the interaction patterns can be interpreted in a different light. Furthermore, interventionists may develop very different intervention strategies for infants who routinely demonstrate low-frequency infant initiation in conjunction with high infant signal clarity, high parent directiveness, and low theme continuity as opposed to intervention strategies for infants who demonstrate low levels of initiation even when paired with a highly responsive parent.

Assessment-Intervention Links

Multifaceted, objective, infant--parent interaction models that link assessment to individualized intervention strategies contingent on unique infant, parent, and dyadic variables and that are based on a strength-building model are needed (Barnard & Kelly, 1990; Farran et al., 1990; Rosenberg & Robinson, 1988). In addition, profiles of infant-parent social interaction across age-, cultural-, and risk-level groups are needed to serve as a reference in intervention decisions. The IPSIC was designed to be an objective, nonbiased assessment of infant--parent interaction that would be free from influence of infant age, risk level, and cultural influences and to serve as a link in the assessment-intervennon process.

The importance of using multiple assessment techniques is widely acknowledged. The IPSIC was designed to be used in conjunction with procedures assessing parents' qualitative descriptions of the infant parent interactive process. Specifically, parental identification of interpretable infant behaviors, parental interpretations of those behaviors, parental feelings of competence and enjoyment, and parental perceptions of social support and well-being are important factors to be considered. Consequently, measures of infant--parent interaction approach objectivity, and the subjective aspects of the assessment process stem from the parents' perspective rather than from third-party observers who are open to interpretative or cultural bias (Hopkins, 1983). In addition, naturalistic observation based on operationally defined interactive behaviors supplement the assessment process. The IPSIC also provides a means for measuring infant--parent interactive patterns that could be examined in conjunction with other aspects of the infant parent relationship such as attachment security, temperament, sense of well-being, and mastery motivation.

Measures of Reliability

In this study we emphasize the need to examine multiple measures of reliability and cast a shadow on the kappa as a primary agreement measure. In the experimental study one construct, infant intentional communicative acts, yielded an artificially inflated kappa. Very little variance occurred in the experimental population on this construct, yet a kappa of .6691 was obtained because the absence of disagreements on one side of the "confusion" matrix suggested a chance agreement probability of 0. A second construct, infant initiation, yielded a very similar pattern of agreement, yet disagreements on both sides of the agreement diagonal led to a kappa of .2748. These findings suggest the relative frequency of behaviors and the profile of agreements influenced reliability indices. Nonetheless, greater reliability with the older population was found.

Reliability continues to be an issue, especially for specific age and behavioral construct subgroups that yielded kappas less than .4 (Fleiss, 1981). The disadvantages of percentage of exact agreement as a measure of reliability are also underscored by the high percentages of exact agreement that were obtained for constructs having very little variance. Examination of descriptive statistics, in conjunction with percentage of exact agreement and kappas, provides more meaningful information than any one of these measures would have suggested.

Descriptive Statistics

Descriptive statistics for each interaction construct were calculated for a subset of 53 infants 2 1/2 through 6 1/2 months of age in the experimental group. Three constructs occurred infrequently and yielded little variance for this age group suggesting these infants may be developmentally immature for these behaviors: infant initiation, infant intentional communicative acts, and dyadic theme continuity. Other constructs reflected a wide frequency range and relatively large standard deviations. Frequencies beyond the plus or minus 1 or 2 standard deviation range may provide relevant interactional information, especially when patterns of co-occurrence are examined. Exceedingly low levels of parent response contingency or parent facilitation, as well as exceedingly high levels of parent intrusiveness may be risk factors. In addition, exceedingly high levels of parent directiveness co-occurring with the absence of infant participation might also be a risk. These descriptive statistics are presented in Table 3.

Examination of the percentages of agreement and kappas for the experimental and replication groups and descriptive statistics for the experimental group suggest individual interactional constructs may be uniquely related to infant age. For example, kappas reflect little variance for the constructs infant intentional communicative acts and infant initiation for the younger experimental population; however, the constructs yielded higher kappas and important information for the older children in the replication population. Likewise, although the construct infant signal clarity provided little information for the older population, it may provide valuable intervention information for infants who are younger and/or have special needs.

Limitations

Although the IPSIC was developed to provide an objective assessment of infant-parent social interaction, residuals of subjectivity are required for selected coding decisions. For example, discrimination between dyadically responsive and unrelated behaviors and between intentional and unintentional infant behaviors is required. In addition, co-occurrence of coded behaviors does not determine a causal link between the behaviors. These limitations appear to be inherent in the interaction assessment process.

Common objections to the use of molecular videotaped interaction analyses include the inaccessibility of expensive video technology, the time constraints required in analysis, the intrusive nature of the videotaping process, and the limited interaction time analyzed. The first concern has been addressed with increased availability of video technology. Second, the time required for a trained examiner to videotape, code, and analyze infant--parent interaction in the present study was found to be comparable to the amount of time typically required to complete other aspects of the early intervention assessment process such as assessment of infant development or assisting families to assess their concerns, priorities, and resources for promoting their infant's optimal development.

The next concerns, videotaping intrusiveness and the limited period of time analyzed, are legitimate and must be addressed. The need for professional sensitivity and a trusting relationship with families cannot be overemphasized. Repeated naturalistic and clinical observations based on the operational definitions developed by the IPSIC supplement and may provide more meaningful information than the videotaped analysis. The IPSIC was designed to be a single component of a multifaceted assessment process, and repeated interaction observations in multiple settings over time are recommended.

Future Research

Until objective descriptions of infant-parent interaction across age, cultural, and disability groups are available, infant-parent interaction intervention recommendations will remain speculative. In addition, outcome correlates of differing interactional patterns remain unclear. For interventionists to confidently make intervention recommendations, questions such as, How much facilitation is enough for this dyed? and How much intrusiveness is too much for this dyed? need answers.

The descriptive, operational definitions in the IPSIC suggest and provide a means for measuring interactive patterns and subtypes of interactive behaviors. For example, previous research suggests directiveness may have distinct subtypes. Intrusiveness is defined as a subset of directiveness that breaks infant attention or has a negative impact on infant behavior. In contrast, the occurrence of parent directiveness, in conjunction with infant participation, may provide the infant with a learning-mediated experience. Furthermore, the co-occurrence of directiveness and facilitation indicate parent directive behavior can facilitate infant attention. Further research is needed to describe the infant-parent interaction process across age, cultural, and disability groups, as well as to determine the outcomes associated with specific interactional patterns.

Additional validation and reliability information relative to the IPSIC is also needed. Of particular interest is the question of cultural bias. Would professionals from diverse cultural groups differ in their findings as they coded videotapes of dyads from dissimilar backgrounds? Would intervention recommendations differ depending on the cultural membership of the professional? The establishment of the IPSIC's reliability and validity is in early stages and the procedure is far from ideal; however, preliminary data suggest that the instrument approaches the goal of an objective means for measuring and describing infant-parent interaction.

Coding the presence or absence of each behavioral construct for each coding segment may not be necessary for clinical purposes. Clearly, a naturalistic, nonintrusive procedure for objectively describing interactive behaviors without inference or subjectivity would be ideal. Until such a procedure is available, the IPSIC appears to have unique clinical and empirical potential and may facilitate the future development of such a process.