Splitting a key into n key splits so that for some k (where k < n), any k key splits of the key can be used to form the key, but having any k-1 key splits provides no knowledge of the key value.
Source(s):
NIST SP 800-152