While the study reported in this article is clearly very preliminary, some conclusions can be drawn from it. Perhaps the most important of these is that Logo programming can be fertile groundfor the investigation of specific problem-solving behaviors (McAllister, 1990). In particular, the clinical interview-style integration of talk-aloud protocols with transcriptions of students' programming solutions seemed to provide useful insights into the domain knowledge, cognitive styles, and problem-solving strategies employed by students participating in the study. Programming problems could probably be developed that would more systematically investigate these and other cognitive areas; continued use of such methodology might prove fruitful.

Second, the study provides evidence of differing cognitive styles in participating students' problem-solving behaviors. Because of the age of the students involved (situated as they were on the cusp between what has traditionally been labeled concrete and formal operational behaviors), it is impossible to either confirm or refute Turkle and Papert's (1990) contention that such cognitive styles represent mature approaches to the world and not, as Piaget (1971) maintained, developmental stages. That varying dimensions of cognitive styles emerged from this study, however, implies that the two views may not be mutually exclusive. Indeed, a third, knowledge-based explanation emerged from the patterns of cognitive styles observed. Clearly, relationships among all three sorts of explanatory structures deserve further investigation. Repeating the study with adults would be interesting. So too would be a longitudinal study that focused on changes in cognitive styles and related these to possible changes in problem-solving behaviors, and/or an exploration of links among students' cognitive styles, their domain knowledge, and their performances on traditional Piagetian tasks.

Finally, although the current study did not explore the relationship between instruction and the development of particular problem-solving strategies, previous work (Swan, 1989, 1990; Swan & Black, 1989) has demonstrated that at least some of the problem-solving strategies explored herein can be developed through explicit instruction and Logo programming practice. The current study suggests that at least a partial explanation for the success of the intervention explored in that previous work could lie.