This paper is concerned with the distribution and interpretation of yekan and its cognates. Syntactically they require negation, but semantically the sentences in which they occur are positive ones that make monotone increasing inferences possible. This syntax-semantics discrepancy can be best accounted for by showing that yekan and its cousins must be strictly c-commanded by metalinguistic negation at the surface structure and that the positive meaning of the sentences they are part of is derived from the cancellation of the pragmatic upper-bounding implicatum associated with them. These also enable us to explain why they do not occur in the environments where typical NPIs do and why only certain forms of negation license them.