By using a sample of 21 pairs of eleventh grade students (10 comprised mathematically-talented students), this study examined paired problem solving performance in terms of paired students' features concerning mathematical self-concept and cognitive empathy and found that collaborative problem solving performance was positively influenced by average mathematical self-concept for paired talented students. The talented pairs' bootstrapped data evidenced that this performance could be explained by a multiple liner regression model, where average mathematical self-concept for paired students and average cognitive empathy for paired students had zero or positive influence, whereas absolute mathematical self-concept distance for paired students and absolute cognitive empathy distance for paired students had zero or negative effect. Though not supported, the validity of that model was indicated by the average pairs' bootstrapped data.