V
Clinical Methods
19
A Scientific Perspective on Structured, Task-Based Interviews in Mathematics Education Research
Gerald A.Goldin
Rutgers, the State University of New Jersey
This chapter considers one methodological aspect of qualitative researchâthe use of structured, task-based interviews in observing and interpreting mathematical behavior. Several scientific issues and their implications are discussed briefly, including: (a) the reproducibility, comparability, and reliability of observations; (b) the generalizability of research findings; (c) the importance of mathematical content and structures; (d) the role of cognitive theory in designing and interpreting interviews; and, (e) the interplay among task and contextual variables. In evaluating task-based interview methods scientifically, I argue against some claims that have been advanced in the name of epistemological schools of thought ranging from radical positivism and behaviorism to radical constructivism, social constructivism, and postmodernism. Finally, some broadly applicable principles and techniques are proposed for improving the quality of task-based interview research.
The perspective offered here is that of a physical scientist as well as a mathematics educator who has been involved in empirical and theoretical research on mathematical problem solving for 25 years. I maintain that sound principles of scientific investigation, as developed and applied in modern science, should be applied to this endeavor too. This should never be done dogmatically or automatically. However, it should be done rigorously, paying careful attention to the reasoning behind the application of the methodological ideas of science. Although I have learned much from the research of others who hold different views, I remain entirely unconvinced by the arguments that are advanced occasionally, claiming that scientific methods of inquiry are inadequate for, or irrelevant to, the study of human psychosocial activities such as teaching and learning mathematics and mathematical problem solving. Because one purpose of this book is to consider quality standards for qualitative research methods, it is essential to consider the fundamental scientific issues.
My experiences with task-based interview methodology originated and evolved through a series of studies of individual mathematical problem solving by elementary school, high school, and college students and adults, conducted in collaboration with my students (Bodner & Goldin, 1991a, 1991b; DeBellis & Goldin, 1991; Goldin, 1985; Goldin & Landis, 1985, 1986; Goldin & Luger, 1975; Goidin & Waters, 1982; Luger, 1980; Waters, 1980). Most recently, members of a group of investigators that I led at Rutgers University have been analyzing and interpreting the results of a series of five task-based interviews in elementary school mathematics. We created these interviews as part of a longitudinal study of individual childrenâs mathematical development (Goldin, 1993a; Goldin, DeBellis, DeWindt-King, Passantino, & Zang, 1993). Between 1992 and 1994, structured interviews were conducted with an initial group of 22 third- and fourth-grade children, 19 of whom completed the full series. Partial results have been reported (DeBellis, 1996; DeBellis & Goldin, 1993, 1997; Goldin & Passantino, 1996; Zang, 1994, 1995). The development of interview scripts for this series was guided by the views described in this chapter; in turn, my views were influenced by insights gained during the study.
It is not my intention to describe the specifics of these studies here, but to focus on methodological suggestions and conclusions drawn in part from them. The chapter is organized as follows. The next section summarizes the meaning, importance, and limitations of task-based interview research in mathematics education. Here I try to explain the notion of structured interviews that are designed to investigate hypotheses using qualitative analyses of data, and offer some brief examples. The ideas presented carry forward and expand considerably on earlier deliberations about the measurement of mathematical problem solving outcomes (Cobb, 1986; Goldin, 1982, 1986; L. Hart, 1986; Lucas, et al., 1980) and the relation between cognitive theory and assessment (Goldin, 1992c). This is followed by a discussion of key scientific issues in connection with the methodology, and the case for explicit rejection of certain damaging conclusions derived from dismissive epistemological belief systems. The final section offers a preliminary set of broad, guiding principles and techniques for establishing and enhancing the quality of task-based interview research in the domain of mathematics.
TASK-BASED INTERVIEWS
Structured, task-based interviews for the study of mathematical behavior involve minimally a subject (the problem solver) and an interviewer (the clinician), interacting in relation to one or more tasks (questions, problems, or activities) introduced to the subject by the clinician in a preplanned way. The latter component justifies the term task-based, so that the subjectsâ interactions are not merely with the interviewers, but with the task environments. Group interviews with two or more subjects fall also within the purview of this discussion, leading to the need to expand our interpretations of some of the ideas.
Normally, provision is made for observing and recording for later analysis what takes place during the interview, through audio- and/or videotaping, observersâ notes, and the subjectâs work. Explicit provision is made too for contingencies that may occur as the interview proceeds, possibly by means of branching sequences of heuristic questions, hints, related problems in sequence, retrospective questions, or other interventions by the clinician. It is this explicit provision for contingencies, together with the attention to the sequence and structures of the tasks, that distinguishes the âstructuredâ interviews discussed here from âunstructuredâ interviews, which may be limited to âfreeâ problem solving (where no substantial assistance that would facilitate a solution is given by the clinician to the subject) or to the handling of contingencies on an improvisational basis. By analyzing verbal and nonverbal behavior or interactions, the researcher hopes to make inferences about the mathematical thinking, learning, and/or problem solving of the subjects. From these inferences, we hope to deepen our understanding of various aspects of mathematics education. We may aim to test one or more explicit hypotheses, using qualitative analyses of the data; we may seek merely to obtain descriptive reports about the subjectsâ learning and/or problem solving; or we may hope to achieve an intermediate goal, such as refining or elaborating a conjecture.
Of course, the design of structured task-based interviews needs to take into account their research purposes. These may include (for example) exploratory investigation; refinement of observation, description, inference, or analysis techniques; development of constructs and conjectures; investigation or testing of advance hypotheses; and/or inquiry into the applicability of a model of teaching, learning, or problem solving. In addition the design is affected by the complexity of the phenomena in the system being investigated.
Task-based interviews can serve as research instruments for making systematic observations in the psychology of learning mathematics and solving mathematical problems. They also can be adapted as assessment tools for describing the subjectâs knowledge and/or improving the practice of mathematics education (cf. R.B.Davis, 1984). The value of taskbased interviews for either of these purposes lies in the fact that they provide a structured mathematical environment that, to some extent, can be controlled. Mathematical tasks can be adjusted in wording, content, setting, sequence, and structure, based on express criteria and the outcomes of prior research. Interview contingencies can be decided explicitly and modified when appropriate. In comparison with conventional, paper-and-pencil test-based methods, taskbased interviews make it possible to focus research attention more directly on the subjectsâ processes of addressing mathematical tasks, rather than just on the patterns of correct and incorrect answers in the results they produce. Thus, there is the possibility of delving into a variety of important topics more deeply than is possible by other experimental meansâtopics such as complex cognitions associated with learning mathematics, mechanisms of mathematical exploration and problem solving, relationships between problem solving and learning, relationships between affect and cognition, and so forth. A few examples may illustrate some of these ideas and their evolution.
During the 1950s and 1960s, many researchers investigated the use of strategies by problem solvers. These studies were consistent with the prevailing behavioral focus in psychology, considering a âstrategyâ to be essentially a pattern in behavior. Strategy scores were defined, based on the kinds of discrete choices made by subjects during problem solving. For example, Bruner, Goodnow, and Austin (1956) distinguished various sorts of âfocusingâ and âscanningâ strategies in conjunctive concept identification tasks, whereas Dienes and Jeeves (1965, 1970) found âoperator,â âpattern,â and âmemoryâ strategies in card tasks that had the underlying structure of a mathematical group. The nature of the interviews was to pose problems where the spectrum of choices available at each point (i.e., the set of possible âbehaviorsâ) was limitedâfor instance (depending on the task) to trying an exemplar, making a guess or conjecture, or choosing a card. The tasks and questions were highly structured in order to circumscribe the outcomes. Then, certain kinds of hypotheses could be investigated quantitatively: ways in which strategy scores might depend on task variables, subject var...