This study is part of a longitudinal study of 22 elementary school children observed over a 3-year time span in the classroom and in individual task-based interviews. Through the examination of videotaped structured clinical interviews of Marcia, the first in 1992 when she was 9 years old (4th-grade), this study seeks to draw inferences about her internal strategic problem representations while engaged in problem-solving activity, and to analyze how these representations developed over a 2-year time span. An analysis of Marcia's task-based interviews shows that strategies are accessible to her, and even more importantly, that they do not necessarily need to be "taught." The research reported here shows that Marcia, when left to her own devices, not only invented strategies and representations to aid her in finding a solution to the problem, but she often did so in noncanonical ways. (Author)