How do people manage to refer to chocolate, despite knowing so little about it? Traditional semantic externalism gives a two-part answer, a negative claim that meanings are not determined inside speakers' heads, and a positive claim that meanings are fixed by external factors. This gets the semantics of ‘chocolate’ half right: the negative claim is correct, but the positive claim is not. There is nothing special about ‘chocolate’, and scientifically respectable natural-kind terms also fail to live up to the positive expectations of traditional externalism. However, kind-term indeterminacy is compatible with important advances associated with externalism's de re understanding of kind terms.