In physics education, a number of studies have developed assessments of teachers' knowledge of student understanding (KSU) of specific physics concepts with modified versions of existing concept inventories, in which teachers were asked to predict the popular incorrect answers from students. The results provide useful but indirect information to make inferences about teachers' knowledge of the misconceptions that students may be using in answering the questions. To improve the assessment of teachers' KSU, a new instrument is developed using a three-tier item design. The items were adapted from 17 questions from the Force Concept Inventory on force and motion. Each item was designed in three tiers, with tier 1 asking for teachers' own answers to the question to test their content knowledge, tier 2 asking for teachers' predictions of popular students' incorrect answers, and tier 3 asking for teachers' explanations of students' incorrect answers in an open-ended form. The three-tier design captures teachers' content knowledge, predictions, and explanations in a single item to allow explicit measures of teachers' own content knowledge and their KSU on students' misconceptions. The instrument was validated with preservice physics teachers, who were master-level graduate students in a normal university in China. The assessment results also suggest that the preservice teachers' KSU of force and motion was only moderately developed, and their content knowledge was uncorrelated with their KSU. In addition, a four-level progression scale of KSU was also developed, which categorized the preservice teachers into five proficiency groups.