I've been told by several GIS instructors that the pixel values of a Kernel Density Estimate raster output by the point KDE tool in ArcGIS 10.1 represents the probability of encountering a point, not the density of points at that location. I've been told that the KDE surface is normalized to a probability density surface by dividing the surface by the total number of points used to make it. My problem is that there is no reference to this normalization procedure in the ArcGIS help files at all, with the possible exception to a reference to the paper by Silverman (1986) from which the quadratic kernel function used in ArcGIS 10.1 is derived. Can any one tell me if this is the case? In addition, if this is the case, if you select a field value instead of just using the point location to build the KDE surface, does this normalization procedure still occur, and how would you interpret the surface?