This report describes efforts to measure teachers' knowledge of their own students' abilities in mathematics and offers preliminary findings. It provides a description of the sample, a description of the study design and its realization, and descriptive statistics for teacher judgment accuracy. The work described here was completed as part of a randomized controlled trial evaluating the implementation and impact of a teacher professional-development program called Cognitively Guided Instruction (CGI). The analytic sample for the spring 2014 wave of data collection involved grade 1 and grade 2 students (n = 504) and their mathematics teachers (n = 146). The analytic sample for the 2015 wave of data collection included a larger sample of students (n = 785) and their teachers (n = 200). During each wave of data collection, teachers were shown a set of problems from the interview and were told which students would be attempting to solve the problems. For each student and item, the teachers predicted whether the student would solve the item correctly. Overall, teachers accurately predicted their students' performance about 60% of the time. Overprediction was more common than underprediction. The authors anticipate data generated with the TAPSS instrument will be used to determine the effects of the teacher professional-development program on teacher judgment accuracy and the extent to which this type of knowledge is associated with student learning. The TAPSS instrument and the sample described here represent several advancements in the field of measuring teacher judgment accuracy, including the iterative development of a new instrument available to researchers and program evaluators. The authors expect that teacher judgment accuracy will ultimately yield important insight into the elusive link between teacher knowledge and student learning.