The study reported here examined virtual manipulatives as an instructional treatment in 17 third- and fourth-grade classrooms. Students were randomly assigned to two treatment groups: texts and physical manipulatives (PM), and virtual manipulatives (VM). Results revealed no significant differences in achievement between the treatments. Additional results showed that objective ability predicted fraction achievement; virtual manipulative use can be modulated by test question type (e.g., symbolic vs. pictorial); percentage of class time using representations differed between VM and PM classrooms; and percentage of class time spent using representation types differed, potentially providing differential opportunities to learn.