Set boundaries determined by given data to the splits in the tree, such that in any inner node if its splitting value would be moved there, then the number of misclassified cases in this node would be one standard deviation over the actual misclassification.
softening.by.esd(fit, d)The soft tree
The data set
This is the same approach as C4.5 uses for "probabilistic splits"
Quinlan, J. Ross (1993), C4.5: programs for machine learning, San Francisco, CA, USA: Morgan Kaufmann Publishers Inc.