This study is to investigate whether self-assessment contains construct-irrelevant variances of gender and age, and whether self-assessment correlates with test scores. The data were collected from seventy-eight newly arrived international students at the English Language Center, Michigan State University, USA. Prior to class commencement, they were asked to self-assess their listening, speaking, and interactive skill. Then, a test of listening was administered; finally, they were assigned to perform task-based conversational activities.The statistical findings imply that, firstly, gender and age do not provide construct irrelevant variances to the validity of self-assessment and, secondly, self-assessment produces reliable scores.