Consistency of students’ responses to matched pairs of
items has been used in the past to evaluate the ‘‘goodness’’
of the items. One can imagine, for example, that ambiguously
worded items may be interpreted differently by various students,
so that their answers are not an indication of their
science knowledge. On the other hand, inconsistent responses
to matched pairs can be a function of the respondents, rather
than of the items: for example, a long duration of a test, or the
conditions under which it is held, may cause some respondents
to answer carelessly. Furthermore, for some time it has been
recognised (Taber, 2000b, 2001) that students may at some
point be in a state of fluidity between two or more competing
explanatory conceptions. In this unstable cognitive situation it
is conceivable that small differences between the matched
pairs, or the circumstances at the time of responding to each,
could give rise to responses based on different conceptual
frameworks. Regardless of the reason, different responses to
paired statements lessen the reliability of the data as a basis for
making conclusions about participants’ understandings