My interpretation (fuzzy) ( :) ) is that if the experimenter puts a
pattern in the data, the caffiend (pace Kennita ;) ) is either able
to see *too many* possible patterns, or is less likely to share the
"expected" context and locks on to the "wrong" (in the opinion of the
experimenter) pattern.
But I don't really know. Coffee nerves keep me from making up my
mind. :)
MMB, at but not for OCC