More recent research studies concentrate on how to assess students’ misconceptions more reliable, valid and useable. The most popular misconception test in physics education has been probable Force Concept Inventory (FCI). Since it was first published as Mechanics Diagnostic test, it has been most accepted and widely used test in the literature (Hake, 1998). However, it was criticized in the literature in terms of what really it measures, percentages of false positives (correct answer with wrong reason) and false negatives (wrong answer with correct reason), and inadequacy of alternatives of the items (Steinberg & Sabella, 1997). Debates in the first part had been focused on the mismatch between the proposed dimensions of the FCI and results of the factor analysis. Up to that time the authors had not supported their test with a factor analysis result. Debates in the second part could be more objective if the FCI had a mechanism of calculating percentages of false negatives and positives. For this purpose, two tier test items have been being used in the literature. In the first tier, concept or the problem is asked. In this sense the FCI items could be first tiers of a two tier test. In the second tier the reason or justification of the answer given to the first tier is asked. In order to discriminate misconception from lack of knowledge, some research have used third tier. In the third tier students are asked to what extent they are sure about their answers. If you want to use a three tier test to assess students’ conceptual understanding as an achievement test, you should give a true score to the students who choose the correct alternative in the first tier and the corresponding correct reason in the second tier and they must be sure about these answers. Likewise, if you want to use a three tier test to assess students’ misconceptions as a diagnostic test, you should give a score to the students who choose the misconception alternative in the first tier and the corresponding reason in the second tier and they must be still sure about these answers. In the light of the debated issues, the purpose of this study are i) to develop and validate a three- tier test to assess the misconceptions about heat and temperature, and ii) to assess Turkish bachelor, master, and PhD students’ misconceptions by a new uncommon coding style.

The participants of the study are the individuals who retrieved the homepage and filled the test. The test was filled by 5000 persons. 2500 of the participants chose that they hold elementary or high school diploma. These students’ responses were discussed in somewhere else (Eryılmaz, 2009). By interviews and essay type questions, we developed three-tier multiple choice test to assess the participants' misconceptions about heat and temperature. Heat and temperature of the same material desks with different sizes, and same size desks made of different materials were asked to be compared. Furthermore, the relationship between heat and temperature in these contexts was asked. Next, the final form of the test was published on a website. The website is a physics and physics education portal that has been being online on the Internet since 2003 and daily accessed by approximately 4,000 visitors.

This study showed the development and validation of the three-tier heat and temperature test. The study also revealed how to calculate correct and misconception scores from the test. The participants’ correct and misconception scores were calculated by using the first-tiers, first two-tiers and all three-tiers. These helped us show the superiority of using three-tier test to assess students’ achievement and/or misconception. These scores were supported by the suitable reliability and factor analyses. Results showed that the percentage of the misconception varied between 5% and 48% for bachelor students, between 4% and 49% for master students, and between 7% and 40% for PhD students depending on whether participants manifested the misconception in every situation or at least in one situation. The results were supported by reliability and factor analyses.