This study used eye-tracking technology to investigate students' visual attention while taking the Force Concept Inventory (FCI) in a web-based interface. Eighty nine university students were randomly selected into a pre-test group and a post-test group. Students took the 30-question FCI on a computer equipped with an eye-tracker. There were seven weeks of instruction between the pre- and post-test data collection. Students' performance on the FCI improved significantly from pre-test to post-test. Meanwhile, the eye-tracking results reveal that the time students spent on taking the FCI test was not affected by student performance and did not change from pre-test to post-test. Analysis of students' attention to answer choices shows that on the pre-test students primarily focused on the naive choices and ignored the expert choices. On the post-test, although students had shifted their primary attention to the expert choices, they still kept a high level of attention to the naive choices, indicating significant conceptual mixing and competition during problem solving. Outcomes of this study provide new insights on students' conceptual development in learning physics.

The Force Concept Inventory (FCI) is a 30-question multiple-choice assessment that has been a building block for much of the physics education research done today. In practice, there are often concerns regarding the length of the test and possible test-retest effects. Since many studies in the literature use the mean score of the FCI as the primary variable, it would be useful then to have different shorter tests that can produce FCI-equivalent scores while providing the benefits of being quicker to administer and overcoming the test-retest effects. In this study, we divide the 1995 version of the FCI into two half-length tests; each contains a different subset of the original FCI questions. The two new tests are shorter, still cover the same set of concepts, and produce mean scores equivalent to those of the FCI. Using a large quantitative data set collected at a large midwestern university, we statistically compare the assessment features of the two half-length tests and the full-length FCI. The results show that the mean error of equivalent scores between any two of the three tests is within 3%. Scores from all tests are well correlated. Based on the analysis, it appears that the two half-length tests can be a viable option for score based assessment that need to administer tests quickly or need to measure short-term gains where using identical pre- and post-test questions is a concern.