A correlation research design (Campbell and Stanley 1966) was adopted by the study. The
participants were tested with the PSAT, DSKT, and RST for a total of three hours during a
one-week period. The testing activities were conducted as part of Earth science class time
and were administered by the class teacher who was trained in a one-hour test
administration session. The training session aimed at familiarizing the teacher with the
test administration procedures and test materials including the introduction of the tests, test
items, and testing procedures. This kind of process assures the collection and reliability of
the data. Students were administered PSAT for 1.5 h, followed by another 1.5 h of DSKT
and RST. Care was taken to ensure that the procedures for administrating these tests were as
similar as possible. The relationships between students’ problem-solving ability and
knowledge and reasoning skills were then determined for the total group of students using a
Pearson product-moment correlation method. We also tested the variance accounted for in
problem solving ability by domain specific knowledge and reasoning skill. Two
subcomponents of problem solving ability (DTA and CTA) were also considered in
separate regression models. The predictors of PSA including DTA and CTA used in the
stepwise regression model consisted of students’ DSK as an index of prior knowledge and
students’ RS as an index of reasoning skills. In addition, interviews with students were
conducted to supplement the understanding of students’ perceptions of how to have better
performance in problem solving activities. Nine students were randomly selected from the
valid test samples for semi-structured interview. The students’ interviews were audiorecorded
and the notes were taken with a focus on their perceptions and reflections on how
to solve the problems effectively.