The assessment of language endangerment requires accurate estimates of speaker populations, including information about the proficiency of different groups within those populations. Typically, this information is based on self-assessments, a methodology whose reliability is open to question. We outline an approach that seeks to improve the accuracy of self-assessment by exposing participants to a simple linguistic task before they render their judgments. The viability of the approach is evaluated with the help of a case study involving 81 partial speakers of Jejueo, a critically endangered Koreanic language.