If one were to perform definitive science on this, what methods should be tested?
It seems there are three main variables: Applicable Skill, Teacher, and Student. If we follow rationally, this allows for a "is skilled" or "no skill" on all three. This means you need 4 teachers, with and without a teachable skill, and with and without the teaching skill, as well as two students, with and without the student skill.
This should occupy 16 dwarves total.
- Skilled Axedwarf, Skilled Teacher, Skilled Student
- Skilled Axedwarf, No Teacher, Skilled Student
- Skilled Axedwarf, Skilled Teacher, No Student
- Skilled Axedwarf, No Teacher, No Student
- No Axedwarf, Skilled Teacher, Skilled Student
- No Axedwarf, No Teacher, Skilled Student
- No Axedwarf, Skilled Teacher, No Student
- No Axedwarf, No Teacher, No Student
That is, (Axe+Teacher) and (Student) in two separate dwarves.
Proposed method is to embark with 6 dwarves and one laborer. The 6 will take some combination of the above skills, allowing 3 instances to be tested at one embark. Very quickly wall off, set barracks, and constant training for 1 year (perhaps from Summer to Summer to ensure that it's a solid year each) and compare the findings. Perform this for each of the 8 instances and compare data. Assuming the fortress isn't suddenly interrupted for some reason, this should allow for some fairly reliable results and solid evidence as to what matters where.