RESULTS
Statistical treatment of the data involved a series of analyses of covariance using pretest
scores as covariate each time. Hypothesis testing involved analyses based solely on the
aggregate scores for each kind of knowledge (simple, integrated, and total). Other exploratory
analyses were based on the three separate measures (multiple-choice tests, essay
questions, and concept maps), as well as the aggregate scores, for both the immediate and
delayed posttest results. These exploratory analyses were used to ensure that instrumentation
was not confounding results in this small-group study. Post hoc analyses using leastsignificant-
difference (LSD) pairwise comparisons were also examined for trends in the
data. In addition, contrasts comparing either the two groups that used talk with the writingonly
and control groups (TW T W C), or the talk-and-writing group with the other
three groups were also examined for both main treatment effects and gender–treatment
interactions (TW T W C). The first group contrast isolated the role of talk, either
alone or augmented by writing, for learning science, whereas the second contrast focused
instead on the synergistic effect of talk and writing.
Using group and gender as categorical variables, analysis of covariance (ANCOVA)
was employed to explore the aggregate scores. Separate analyses were conducted for the
immediate and delayed posttests. Table 1 shows the adjusted means and the standard
deviations for the aggregate scores by treatment for all students using the pretests scores
as covariate each time. An analysis of the adjusted means for the aggregate scores for all
students shows that the rank order of the talk-and-writing group and the talk-only group
were first and second, respectively. The writing-only and control groups ranked either third
or fourth for all knowledge measures at both the immediate and delayed posttests.
Table 2 shows the adjusted means and the standard deviations for aggregate scores by
gender and treatment with the pretest scores being used as covariate each time. The trends
across treatment groups were quite different for boys and girls. For boys, the talk-andwriting
and the writing-only groups generally ranked first and second, respectively, across
the different aggregate knowledge measures for both posttests. Writing thus appeared to
be more helpful than just talking for boys. In comparison, girls in the talk-only and the
talk-and-writing groups always ranked either first or second for all measures and time-oftests.
Talking thus appeared to be more helpful than just writing for girls.
Three null hypotheses guided this study. Hypothesis testing involved analyses based on
the aggregate scores for each kind of knowledge (simple, integrated, and total). Post hoc
analyses using LSD pairwise comparisons were also examined for trends in the data.