How to measure student learning?
Assessment tools
 Go beyond typical word problems
found at the end of chapters, and probe students' understanding and misconceptions,
rather than their ability to follow examples.
 Are typically given pre &
post instruction to give a better sense of what changed during the semester.
Normalized Gain
(G) 
One way to gauge students' improvement over the course of a semester.
Think of it as the the fraction of the concepts learned that were not already known at the beginning of the course.
Examples:
 A student with a prescore
of 20%, and a postscore of 60% would have a G of 0.50.
 A student with a prescore
of 50% and a postscore of 75% would also have a G of 0.50.


Dick Hake published one of the first
analysis of FCI results from colleges & high schools. One of his findings
was that "interactive" courses, no matter what techniques were used,
yielded higher gains than "traditional" courses.^{1, 2}
Despite the fact that most students in a class show improvements in their understanding of physics concepts, there are often students who struggle to grasp the fundamentals. Even with interactive teaching techniques often a fraction of the class is not able to perform at the expected level.
Within each course there are, of course, variations among the individual student gains. For example, here is an example histogram from
one of my recent Introduction to Mechanics courses that shows students' normalized gains on the Force Concept Inventory (FCI):
Despite the fact that the class average
gain was 0.48, several students had low gains (below 0.30). In fact, some students
had negative "gains"!
 R. Hake, "Interactiveengagement
vs traditional methods: A sixthousandstudent survey of mechanics test data
for introductory physics courses" American Journal of Physics
66, 6474 (1998)
 Some readers may remember a slightly different distribution of normalized gains that was presented in "Teaching physics: figuring out what works" by E. F. Redish and R. N. Steinberg, Physics Today 52, 2430 (1999). While the details are different, both distributions clearly show that traditional, passive courses result in less student learning.