We studied the perceived effects of pair programming (PP) compared to solo programming in a large scale, industrial software development context. We surveyed developers (N=28) regarding effects of PP on learning, quality, effort, schedule, and human factors. Our findings support earlier results from studies done with students, or professionals doing small tasks. The positive effects of PP were largest for learning, schedule adherence of tasks, getting to know other developers, and team spirit. A small but clearly positive effect was perceived for various quality aspects, discipline in following work practices, and enjoyment of work. The improvement of estimation accuracy was almost negligible. The amount of refactoring did not change. On the negative side, the development effort for individual features was higher. In the beginning of the adoption, the exhaustiveness of work was perceived higher, but over time it decreased to the level of solo programming.