Participant maps were randomly coded and scored by a team of six content experts working in teams of two. Maps were scored on three subscales: overall map structure, number of propositions, and quality of the science content expressed in each proposition. Experts coded the maps separately, then negotiated their final score to consensuses. Results were entered into Excel, then transferred to SPSS for analysis.
A paired-samples t-test was conducted on data for each subscale. A statistically significant increase was found for the number of propositions pre (M=24.80, SD=13.643) to post (M=35.47, SD=4.642), t(15)=3.268, p=0.006 and for the quality of the science content in the propositions pre (M=43.27, SD=27.020) to post (M=66.27, SD=28.326), t(15)=2.694, p=0.017. The complexity of the maps was not found to change significantly from pre (M=4.00, SD=0.784) to post (M=3.93, SD=1.072), t(15)=0.268, p=0.793. The eta squared statistic indicated a large effect size for number of propositions (0.43) and science content of the propositions (0.34).