I see that one of early experiments showed that constant 18-day intervals were better than increasing intervals. So how does this prove the validity of SuperMemo?
This 1985 experiment does not prove validity of SuperMemo. Nor does it disprove it. In that respect, it is simply inconclusive. The experiment was an attempt to “guess” optimum increase in intervals and the guess appeared wrong. If the underlying hypothesis was that increasing intervals are always better than equally distributed intervals, then this experiment proved that hypothesis wrong. Obviously, if equal intervals are short enough or the increase in the length of intervals is too fast, equal-interval schedule may appear superior.
