The worldwide spread of democracy experienced in the past century and a half may be understood as a nonlinear (learning) process of innovation-diffusion. A Fisher-Pry test of this proposition was first reported by Modelski and Perry in a 1989 paper (published in this journal in 1991) on the basis of a data set that covered the period 1837-1986. A retest of the innovation-diffusion thesis has now been performed with basically the same methodology but on a refined data set and with data including the year 2000. It reaffirms the earlier result, and confirms that the 50% saturation point (flex-point) has been attained. It also reaffirms the earlier forecast that the 90% saturation level for democracy would not be reached until early in the 22nd century. The time constant (the time elapsing between 10% and 90% saturation) of this learning process is now estimated at 228 years. (C) 2002 Elsevier Science Inc. All rights reserved.