Why historians prefer primary to secondary sources

Historians usually prefer to use primary rather than secondary sources. Why? Can you think of an instance, however, where the reverse might be true? Discuss.