Crowley, Mary Lora NoffsingerThe goal of this project was to create a 40 minute long multiple-choice Instrument to assess an Individual's dominant level of thinking, as described by the van Hiele model of the development of geometric thinking, on the topic of quadrilaterals. The study was composed of four stages: (a) Item development, (b) pl lot testing, (c) field testing and (d) final testing. Initially 53 items were developed and reviewed by a panel of experts. The revised Items were then administered to 14 pl lot study subjects, and, subsequently, to 113 field test subjects, both groups ranging in academic background from sixth grade to university. Item analysis comparing these subjects' choices of level specific responses and their dominant van Hiele level, as determined through the Burger and Shaughnessy interview, resulted in the identification of 19 items for the final Instrument, the van Hiele Quadrilateral Test. For scoring purposes, the items on the test are considered as four subtests, with 4, 5, 6 and 4 Items corresponding to Levels 1, 2, 3 and 4, respectively. The items associated with Levels 2, 3 and 4 met alI the item analysis criteria. Two interpretation schemes were identified. The final instrument was administered to 50 subjects in ninth grade and 51 subjects in twelfth grade. Grade membership and performance on the Nova Scotia Achievement Mathematics Basic Concepts Test were compared to subtest performance and to the resulting mastery decisions. Chi squared statistics failed to support the independence of grade membership and van Hiele level. The correlation statistics, Φ, indicated that there was a weak correlation between grade level and mastery of Levels 1 and 2, with stronger statistics associated with Levels 3 and 4. Little of the total variance in mastery designations (η2y1x) was attributed to variance In grade level. Little to moderate variance (η2y1x) in performance on the Achievement Tests was attributed to variance in, the van Hiele level assignments. Two types of criterion-referenced reliability statistics, the agreement coefficient, Ρο, and Cohen's Kappa, K, were also determined. These indices suggest that the subtests do not yield consistent results for these subjects. Until reliability can be established, the Instrument ls not appropriate for determing van Hiele mastery levels. The implications of these findings and suggestions for further research are considered.enDissertation