Writing can be a tool for communicating and learning in content area subjects. This pretest-posttest quasi-experiment examined the effects of instruction in a content area writing framework on students’ text quality and ability to use writing to learn. It also examined the effects of possible moderator variables (gender, previous writing achievement) and mediator variables (genre knowledge, approach to writing). A multilevel analysis was conducted with students nested within classes. Instruction significantly increased argument genre knowledge and explanation text quality, but not argument text quality, explanation genre knowledge, or learning during writing. Gender predicted previous writing achievement and posttest argument text quality, but did not interact significantly with instruction. Previous writing achievement strongly affected several posttest measures, but did not interact significantly with instruction. A path analysis supported the theory that instruction affects genre knowledge, which affects text quality, which predicts learning during writing.