The construction of bivariate Gaussian distributions by Puente and Kle
banoff,(1) as transformations of diffuse probability distributions via
space-filling fractal interpolating functions is reviewed. It is show
n that the ergodic theorem,(2) which allows a simple computer implemen
tation of the construction for some diffuse measures, leads in a great
many instances, to unexpected kaleidoscopic patterns of exotic nature
, which, when added, yield the bivariate Gaussian distribution. Exampl
es of such patterns, some resembling natural shapes, are presented.