The units identified by the Relational Coding Scheme represent different patterns of mutual adjustment

between partners and therefore the interaction corresponds to a sequence of episodes defined by an action of a partner followed by an selleck compound opportunity to act for the other. To take an oral conversation as an example, one partner talks and at the same time provides the other with a variety of opportunities to reply. The partner can reply in a way that follows on from the other’s content, at the same time bringing into the conversation something new; so, their communicative episode can be considered to be coregulated in a reciprocal manner. According to the coding system, the coregulation forms we observe in a communicative process vary from unengaged to unilateral to asymmetrical to symmetrical coregulation, and breakdowns in communication can also occur (see Table 1 for the operational definitions). For the purpose of the

present study, the symmetrical code was divided into three subcodes—affect, Selleck NVP-LDE225 action, and language, respectively—so, the original scheme has been partly modified (see Table 1). Coding was done continuously from the video by two independent coders and the coregulation states were identified by segmenting joint activity into units, lasting at least 3 sec, corresponding to the above categories. The onset time of each code was also recorded. From the coding records, durations of each category were computed and used as measures for the analysis. Because of slight variations in the session length, the raw durations in each session were transformed into proportions according to the duration of that session (proportional durations). Proportions of categories of less than .5% were excluded from the data analysis. Interobserver reliability was calculated on 30% of the entire data set. To be specific, 30% of sessions were randomly sampled for each dyad from each of the following

three age periods: 44–64, 65–88, and 85–104 weeks (Bakeman & Gottman, 1986, p. 77). Kappa assessments were based on whether two independent Astemizole coders agreed about the category coded in each second. Across all categories, the average kappa was not less than 80% in each of the three periods. Hierarchical random effects modeling (Goldstein, 1995, 2003; Snijders & Bosker, 1999) was used to test the advanced hypotheses. MLwiN statistical software was used to implement all the models (Goldstein et al., 1998). In the present study, data were collected on a two-level hierarchy (Rasbash, Steele, Browne, & Prosser, 2005), with the dyads at the higher level (level 2) and the set of measurement occasions (i.e., the infant’s age in weeks) for each dyad at the lower level (level 1).