The study reported here uses a quasi-experimental design to compare the algebraic thinking of students after a 3-year, grades 3--5 early algebra intervention to students in more traditional (arithmetic-focused) classrooms. Data sources are students' responses to grades 3--5 written assessment items, designed by the project team to measure learning throughout the intervention, that focus on students' use of variables to represent generalizations across four situations: (1) algebraic expressions; (2) equations; (3) functional relationships; and (4) arithmetic generalizations. Results shows that intervention students performed significantly better than comparison students in both grades 3 and 4 post-tests. More interesting, however, is that intervention students were significantly more successful writing a function rule with variables than with words.