For change to occur in mathematics instruction, teachers need control of significantly different instructional sequences, evaluation schemes, and curriculum and to think beyond procedural views of mathematics. Two-week inservice and coursework separate from classroom experience are not sufficient to achieve these goals. Inservice and support must bring about conceptual understandings and parallel actual classroom implementation via extensive co-teaching and modeling. Sufficient time must be spent to integrate these new and desired understandings into teachers' routines. This paper describes implementations where such change took place and the model which led to joint ownership in project and outcomes. Valley Crest Project is a collaborative teacher/university researcher program intended to help change the teacher's instructional ideas and behaviors. The original instructional approach by the teacher was behavioristic and associationalist with an emphasis on correct answers. To change instruction, the project provided stages in which the teacher observed the university researcher, identified a problem area to work on, learned along with students as the university researcher taught, co-taught with the university researcher, and received long-term support after the project was over. The end result of this project was a transition of control to the teacher and a viable implementation of effective mathematical change at the school level. The teacher became indoctrinated into the culture of "real world" mathematics and was instrumental in disseminating this culture to the students and to other teachers. (Contains 16 references.) (Author/MDH)