learn.5tein.com Jared Stein's grad-school-community blog on teaching and learning.

Revisiting Bloom’s 2 Sigma Problem

Bloom, B. (1984). “The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring,” Educational Researcher, 13:6(4-16).

Bloom's 2 sigma problem confronts educators and researchers with the challenge of improving student performance/learning outcomes by 2 sigma based on a combination of 2 or 3 significant variables in instruction, learner, environment, or materials. This semester I am taking Jon Mott's 1 credit course on the subject, and look forward to finding many enlightening articles and sources, as well as lively and provocative discussion.

I've read and though about Bloom's 2 sigma problem before, but I think on this second read I actually got the point: It's not that 1-1 tutoring is so potent (it is, but this should be obvious, Oxbridge, apprenticeship models), but that Bloom and his students proved that it's possible to provoke remarkable improvements in the performance of the average student by altering just one or two variables. This suggests that our understanding of human potential may be misconceived, and that our standard practice of teaching and learning consistently fails to rise above mediocrity.

I've heard David Wiley say, why stop at 2 sigma? Why not 3 or 4? Why not indeed? And yet there are so many potentially significant variables in the Bloom study--or any other study that attempts to achieve similar results--that I am naturally cynical of finding a "break through". (If there had been one already, we would have heard of it, surely?)

A few questions I bring in:
Are the Bloom's students' results reliable? repeatable? at least one suggests its not, and without greater details from Bloom et al it's hard to reproduce the study.

What were the learning outcomes? How deep are they? How important overall to a student's progress?

What is it about 1-1 that is so useful? Focused and immediate feedback? Q &A? Social aspect? Behavioral?

Should we ignore the 1-1 possibility? Computers, AI have long been thought the possible solution for the human tutoring problem.

Does some 1-1 have a significant effect? Say, 1 hour per week? Could some 1-1 positively affect performance in other areas by (1) motivating, (2) modeling? Say each student in a classroom of 15 gets 30 minutes one-on-one a day in one subject?

How relevant is the 2 sigma problem today? Have our media communications--indeed our culture--changed so much in the past decade that the act of teaching and learning must first be redefined?

We are used to the idea of a bell shaped curve, of low and high achievers. Bloom's research tweaks that in favor of everyone's success. As a teacher what narratives do I tell myself to justify student failure?