Implicit in discussions of learning curves in organizations (and explicit in most) is the idea that focused, uninterrupted learning is best. Learning curves (which go down and to the right, please note) are descended because of doing the same thing over and over.
That may not be quite the case. In “Learning by Doing Something Else: Variation, Relatedness and the Learning Curve” [PDF], Melissa A. Schilling, Patricia Vidal, Robert E. Ployhard and Alexandre Marangoni say that what’s important learning to learn is important. If you learn on related materials, you will have steeper learning curve (read: quicker down the learning curve to performance parity) on other things.
In other words, you can learn to learn.
Variation related to the subject helps the learning rate, according the authors. Variation that is unrelated doesn’t help.
At the individual level, some psychologists have noted that an individual is unlikely to make a significant contribution to an area until they have at least a decade of intense study in a particular domain of knowledge (Hayes 1989). Simon and Chase (1973) quantified this expertise by studying chess grand masters and other experts, concluding that individuals need approximately 50,000 “chunks” of information related to a narrowly defined problem domain prior to making a fruitful discovery. [pp. 44]
(Citations are for J. R. Hayes’s The Complete Problem Solver (1989) and H. Simon and W.G. Chase, “Skill in chess”, American Scientist, 61:394-403.)
That’s a lot of learning. 50,000 chunks, taken at 10 per day, will take you almost 14 years to learn. And that’s with no weekends off. You simply must Related variation actually helps getting down the learning curve.
The findings indicate that groups working under conditions of related variation — that is, working on different but similar types of problems over time — learned at a significantly faster rate, on average, than did teams that either worked under conditions of specialization or unrelated variation…. Finally, though not hypothesized, the results suggest that teams with greater variability in their scores per hour performed slightly better overall. [pp. 52]
This isn’t a final word by any means: their study reeks of college student participants. But it certainly raises some very interesting questions that deserve further study. It explains some of why consultants seem so much more able to learn than their equals who are not consultants. Consultants work on related but different projects — similar projects at different clients have substantial differences, even with the same project domain — while these things are much more stable for employees of equal ability.
And it only took me 18 months to find it. And because it appeared in Management Science, there’s the usual sucking up to Linda Argote.
CITED: “Learning by Doing Something Else: Variation, Relatedness and the Learning Curve” by Melissa A. Schilling, Patricia Vidal, Robert E. Ployhard & Alexandre Marangoni. Management Science, Jan 2003 49(1): 39-56.
Note: Learning curves go down and to the right. You want the learning curve to be as steep as possible, since you get back up to high performance quicker. Just in case you haven’t seen this before.
Image Credit: Photo by Delano, Jack. Young worker at C&NW railroad, 40th Street shops. Chicago, IL, 1942. Via Library of Congress collection at Flickr.