A widely assumed limitation on the distribution of negative polarity items (NPIs) is that they must be c-commanded by a trigger. Scrutiny of this limitation reveals, however, that c-command by a trigger is neither a necessary nor a sufficient condition on the distribution of NPIs. The failure of ccommand to serve as the basis for an account of the syntactic distribution of NPIs is taken here as the impetus to pursue an alternative approach, one in terms of scope and linear order instead of c-command. A particular concept of scope is established that serves as the key notion for characterizing the syntactic relationship between NPIs and their triggers. Scope is defined in a dependency grammar (DG) theory of syntax in terms of scope domains. The catena unit plays an important role. Given the notions of scope put forward, it is possible to discern two conditions on the syntactic distribution of NPIs.