R help - Nov 2003 - rpart node size

After getting an error, I figured out that there is a
minimum node size for doing classification trees with
rpart.  Surely nodes with too few observations are not
reliable.  However, I have not found any references to
how one might set the min node size.  Knowing this
number will allow me to calculate how much data I need
for the tree to get reliable results (it's expensive
to run experiments).  Any thoughts?

Kais