Jose Mestre, Thomas Thaden-Koch
7th International Conference of the Learning Sciences, Indiana University, Bloomington, IN, (2006)
Abstract
We will discuss an application of coordination class theory to a situation in which students were asked to
judge the realism of computer animations showing metal balls rolling on a pair of metal tracks. The animations
were based on a variant of a common physics classroom demonstration, sometimes called “High Road/Low Road.”
Given different sets of animations, students were asked to identify the “most realistic” one from the set. When
presented with animations of a single ball, most students focused on the presence or absence of realistic speed
changes. Adding a second ball (moving in an adjacent track) to the animations drastically changed the strategies
that many students used, to the extent that they identified as realistic a particular (unrealistic) motion with two balls
that they had ruled out minutes before with one ball. This motion featured a ball speeding up as it rolled uphill,
which the students found unrealistic in the one-ball context. When both balls were present, this same motion (with a
ball speeding up while rolling uphill) showed one ball speeding ahead of the other before they reached the ends of
their tracks together (a pattern that many students identified as a characteristic of realistic motion on these tracks).
We attempted a concrete (often phrase-by-phrase) mapping of student interview data onto the structures
postulated for coordination classes. Students often repeated similar descriptions of similar motions, which allowed
us to identify several common expectations of realistic motion, and some of the sources for those expectations; these
we mapped onto the causal net. The same repeated descriptions of well-specified stimuli allowed us to make
plausible inferences about the readout strategies students used in different circumstances.
Our analysis particularly highlights the effect that students’ knowledge can have on the observations they
make. The participants in the study were drawn from two different college courses. Students from an introductory
physics course were especially prone to the problem of accepting anomalous motion as realistic, as described above.
In contrast, students from an educational psychology class were virtually immune to the problem. Mapping
interview data onto coordination class structures facilitates an explanation in terms of the knowledge that students
brought to the task (causal net elements), and the sorts of observations (readout strategies) that they focused on
based on that knowledge. In this explanation, the readout strategies evoked were strongly dependent on the
knowledge that students brought to bear on the task, and on details of the context. The emphasis in other
applications of coordination class theory attends mostly to the use of readout strategies to gather information, on
which the causal net will operate. In our case, different causal net inferences led to very different ways of observing
the world—that is, the causal net affected readout strategies.
Another feature of this application of coordination classes is that virtually no attempt was made to identify
knowledge structures or concepts that could qualify as coordination classes. To the extent that it is useful, the case
provides evidence that coordination, and the language of coordination classes, can be fruitful even in cases where
“coordination classes” are of only peripheral interest. These issues are discussed in more detail in a recent article
extending our prior work (Thaden-Koch, Dufresne & Mestre, under review).