Martine van der Pluijm

142 Chapter 5 to four main categories to define the quality of language used by children and parents by coding their number of utterances in four categories: 1) language about content that is visible within the context, 2) decontextualized talk about content that is not visible in the context, 3) interactive talk, such as encouragements and praising and 4) other types of talk, such as procedural talk about how to do the activity and talk that is not related to the activity. Intercoder agreement The first author and two researchers coded the data. Both researchers, who were not involved in the research, were trained in coding in two sessions by the first author. A handout with examples was provided. After these sessions, the researcher and the assistants coded two scripts. Codes were compared until there was full agreement. Next, five transcripts, randomly selected from each of the two activities and the pretest and posttest, were coded by two researchers. Intercoder agreement was calculated as a percentage of agreement for each of the pairs of coders. The percentages were 80% for the quality of the interaction and 79.5% for the quality of language. These were considered to be adequate. Analyses First, we used repeated-measures ANOVA to analyze change from pretest to posttest in each of the variables (quality of interaction, quantity of language, and quality of language). Second, we used quality of delivery as a factor in the repeated measures analyses (comparison of parents with teachers who showed a higher quality of delivery versus those who showed a low quality of delivery). In some cases, the assumption of equality of error variances was violated, according to Levene’s test. In those cases, we used a nonparametric test (Mann-Whitney test). RESULTS STUDY 2 Table 5.5 shows means and standard deviations at pretest and posttest for each of the variables measured during the drawing and play activities. The observed behavior varies to a great extent (e.g., children who did not talk at all to children who talked a lot and more than their parent). The table also shows whether the differences between pre- and posttest are significant in the repeated measures analyses.