Full Idea
It seems that the fewer the instances, the more likely it is that there be a regularity, ..and if there are no cases at all, and no S is P, that is a regularity.
Gist of Idea
Regularities are more likely with few instances, and guaranteed with no instances!
Source
Stephen Mumford (Laws in Nature [2004], 03.3)
Book Reference
Mumford,Stephen: 'Laws in Nature' [Routledge 2006], p.39
A Reaction
[He attributes the second point to Molnar]