Although it is popular to claim that smaller class size leads to better learning, especially in our public schools, I'm not aware of any empirical data to support this claim. In the US, the average class size in our public schools is 30, whereas in Japan it is 45--yet Japanese students have outperformed American students for decades. With regard to freshman composition, I know of only one published empirical study (Feldman, 1984, Class size and college students' evaluation of teachers and courses, Research in Higher Education, 20, 45-116), but it does not address the issue of class size and performance. There may be others, but if so, they are not well known. Given the number of variables associated with student performance--motivation, previous training, students' sex, SES, teacher expectations, teacher training and methodology, reading ability, intelligence, etc.--I'm not sure how one could go about even researching this question in a way that would control all the variables. Having stated what may be obvious, I would note that much anecdotal evidence supports the view that effective composition instruction entails establishing a mentoring apprenticeship with students. If this view is correct, and I believe it is, then the "ideal" class size would be much smaller than 12. In fact, a composition teacher would never have more than 4 or 5 students per year.

I haven't read research that supports that 12 students is the ideal, but it sounds like a good size to me! I imagine it would be quite difficult to prove since there are so many contextual variables. However, if you're looking for some background information on class size for first-year comp at other institutions, here's a good source, at CompPile:

Class sizes for First-Year Regular and Basic Writing Courses