I'm having a hard time articulating a response to John Clement, because I
find myself both agreeing and disagreeing!
I certainly didn't mean to imply that all PER is a waste of time and/or
money. I believe that many useful results have been acheived, such as the
"momentum before energy" result you quoted. Many experiments can and should
be done.
But not every experiment that CAN be done SHOULD be done. Once you have
some good generalizations, you don't test every specific instance. I don't
need to unplug my computer to know that I would lose this email. Similarly,
I expect from my knowledge of learning that being self-consistent leads to
improved performance. I was willing to accept that I could extrapolate to
this specific instance without need of a new experiment. (Plus it would be
nearly impossible to segregate the students for 10 yr and shield them from
the hordes using a different convention to see if there was any long-term
advantage. You might actually find they do worse becuase they are forced to
learn a new convention to fit in with the crowd.)
Yes, having a variety of strategies helps. For example, when learning
kinematics, working with equations, graphs, verbal descriptions and and
actual experiments wit hmotion detectors all reinforce learning.
But not every alternate strategy is a good idea. I don't think that
introducing "slowness" = dt/dx and having students spend time solving
problems with this quantity would aid their learning. And I don't think
giving names to R, L, C, 1/R, 1/L, _AND_ 1/C would help. Pick a set of
three and concentrate on them.
Finally, I bet you would find a measurable gain. If nothing else, students
would do better on the parallel/series circuit problems.