Full Idea
If kinds are based on similarity, this has the Imperfect Community problem. Red round, red wooden and round wooden things all resemble one another somehow. There may be nothing outside the set resembling them, so it meets the definition of kind.
Gist of Idea
You can't base kinds just on resemblance, because chains of resemblance are a muddle
Source
Willard Quine (Natural Kinds [1969], p.120)
Book Reference
Quine,Willard: 'Ontological Relativity and Other Essays' [Columbia 1969], p.120
A Reaction
[ref. to Goodman 'Structure' 2nd 163- , which attacks Carnap on this] This suggests an invocation of Wittgenstein's family resemblance, which won't be much help for natural kinds.