Recently various specific and general cognitive diagnostic models(CDM) have been used in modeling data from language tests. CDMs have huge potential to provide learners and teachers with fine-grained diagnostic information about their learning and teaching beyond traditional total scores. Despite its advantages, current CDM applications in language testing are confined to analyses of large-sample data from national or testing agencies for educational decision making. For CDMs to be useful in practical classroom assessment contexts, they should be able to overcome constraints due to small sample conditions. Meanwhile the bayesian estimation procedures open a new path for modeling complex entities over the frequentist approaches: It does not depend on large-sample theory; it is flexible and robust in handling complicated models; it can incorporate extra information about the model into processing data; and it provides reasonable model-data fit statistics. However, bayesian approaches to CDMs are not well known to researchers in language testing as well as in SLA. This study introduces the bayesian approach to cognitive diagnostic modeling by applying the new method to a set of data from a listening comprehension test developed for measuring high school students’ listening ability. The study then compares the results with those from the conventional method and discusses its implications in modeling CDMs in language testing.