When discussing statistical power of a paradigm, the values 0.80, 0.85, or 0.90 are often referred to as acceptable or good levels.

Is there a default citation which refers to these values?

I understand that Cohen's book is often cited but he doesn't technically make a case for a certain power criteria.