Hi Pete:
Thanks for both replies. The discrete features won't work because the scale I am working with is small - an experimental cube of soil 5m on a side with a void that is about 1 m in diameter, so the 1D and 2D discrete features will not accurately reproduce what is happening. I tried using a large K-value for the void, but the model showed the void as a preferential flow pathway with a higher water content than the surrounding soil. In reality, the void was just that -empty, with no water in it.
I have thought about ganging up a series of 2D discrete features along adjacent nodes to "build" a 3D feature, but have not tried it yet.