In previous sections we assumed that we were using one-dimensional models, and smoothing was easy. Working in two dimensions is nominally much more costly, but some tricks are available to make things easier. Here I tell you my favorite trick for smoothing in two dimensions. You can convolve with a two-dimensional (almost) Gaussian weighting function of any area for a cost of only sixteen additions per output point. (You might expect instead a cost proportional to the area.)