It is demonstrated that the proofs given in prominent and well-established weak generative capacity arguments for natural language are flawed, due to unexpected interpretations of strings. However, once unique representations of lexical semantic senses form part of such intersection-based proofs, the arguments stand.