A method and system for estimating uncalibrated task performance are disclosed. The method includes providing a plurality of tasks to one or more examinees, wherein the plurality of tasks comprises a plurality of calibrated tasks and a plurality of uncalibrated tasks, wherein a task model corresponds to each task, wherein each task model comprises parameters relating to characteristics of the corresponding task, wherein each task model for an uncalibrated task further comprises one or more radicals, wherein a student model corresponds to each examinee, wherein each student model comprises one or more variables each relating to one or more of knowledge, skills and abilities of the corresponding examinee and parameters describing a relationship among the one or more variables; receiving a response for each task from each examinee; generating one or more observables for each response; and calibrating one or more student models and each uncalibrated task based on at least the task model corresponding to the task and the observables obtained from the responses.