Sweller, J., & Cooper, G. A. (1985). The Use of Worked Examples as a Substitute for Problem Solving in Learning Algebra. Cognition and Instruction, 2(1), 59–89

The first paper to suggest Worked example effect (AFAICT), by John Sweller and Graham Cooper. The paper predates Cognitive load theory but its claims are clearly starting to point in that direction.

Two main inspirations:

1. Thought and Choice in Chess - DeGroot and other early findings from the expertise literature which suggest that schema acquisition is central to expertise
2. Discovery learning findings which suggest that “search techniques retard knowledge acquisition rates” and “puzzle problem” findings which suggest that difficult problems seem to make it difficult for people to find solution schemas.

The authors suggest that students may be better off studying worked examples instead. They engage in five experiments to test this effect; controlling for study time, students who study worked examples take roughly half as long to solve similar problems on a post-test, and they make 1/5 as many errors. But the effect doesn’t seem to extend to problems which are varied on the post-test.

Q. High-level findings of experiment 5?
A. Algebra students who study worked examples take roughly half as long to solve similar problems on a post-test, and they make 1/5 as many errors. But the effect doesn’t seem to extend to problems which are varied on the post-test.

Q. Key limitation of findings?
A. The effect doesn’t seem to extend to post-test problems with even modest variations from the problems studied.