There is long-standing interest in assessing students' progress throughout their education, which has increased since the implementation of the No Child Left Behind Act of 2001. One popular metric to assess progress is a vertical scale, which places tests at different educational levels but of similar content onto a single score scale.

Of interest in this dissertation was the extent to which the domain definition influences the vertical scale. This is important for several reasons. First, the content on which a vertical scale is based may become out-dated, or views of its importance may change. This may cause the vertical scale not to accurately reflect what educators believe students need to learn. Another reason educational stakeholders' interest in comparing results of state tests with those from other state and national tests. If the content on which the vertical scales are based varies between tests, meaningful comparisons of student achievement may be unattainable.

The purpose of this study is to compare the effect of varying the domain definition of science with the effect of using different psychometric models to create a vertical scale. Three tests of science were assembled that varied in the degree to which they assessed scientific inquiry skills, an area believed to be important for students in all grades. The tests were administered to students in grades 3 through 8, and vertical scales were constructed and compared for each test using the Hieronymus and 3PL IRT scaling methods.

Results suggest that domain definition had a larger effect on vertical scaling than did scaling method. This effect seems to be most pronounced for a test consisting solely of items that assess scientific inquiry rather than a mix of other sub-content areas (e.g., life, physical, and earth and space sciences).

Although the effect of domain definition was not pronounced, these results suggest that content can have an important influence on patterns of growth described by vertical scales. Greater attention to this influence has the potential to increase understanding of all aspects of vertical scaling methodology.