Does the NOL Effect Fade after 4 Levels of a Quantitative Attribute?

In the 1980s, Dick Wittink found that doubling the number of levels (from 2 to 4) for a quantitative attribute could significantly increase the attribute’s importance (holding the range of variation constant). In 2000, Marco H. presented a paper that suggested that the NOL effect faded after about three-four levels. We’ve done a couple new R-on-R projects recently to test that assertion.