why is the n df residual for between contrast sigma, but n subjects for within?

[kel-github]

R matches Psy output for contrast sigma (sqrt(MSE * sum (c^2/n)) when:

for between contrasts: n = residual df for within contrasts: n = n subs