Wh-mo outside the Neg-c-command Domain

Abstract

An indefinite expression such as dare-mo and nani-mo in Japanese requires the sentential negation (Neg) and gives rise to an interpretation with universal negation. In analogy to English any-, they have generally been treated as Negative Polarity Items (NPIs) and have been assumed to be in the c-command domain of Neg at LF, following Klima (1964).

Italian and Spanish have another kind of indefinites involving negation, e.g., nessuno in Italian, which is called Negative Concord Item (NCI). It has a sentential negative force, but induces a single negation even when it is ‘doubled’ with Neg. While the canceling of negative force is assumed to induce a single negation, it has not been specified which of the two negative forces gets canceled (Haegeman and Zanuttini (1998), among others). Recently, Watanabe (2004), assuming that dare-mo and nani-mo are NCIs, has argued that the negative force of Neg rather than that of NCI should be canceled in ‘negative doubling’ cases, concluding that NCIs are inherently negative, on the basis of the ellipsis data involving dare-mo/nani-mo.

In this paper I show that dare-mo/nani-mo cannot be c-commanded by Neg at LF and that they do not have a negative force, either, based on their interactions with XP-sika, another Neg-sensitive element in Japanese, and I conclude that neither the NPI-analysis nor the NCI-analysis is appropriate. I also point out problems with Watanabe’s analysis of ellipsis, and argue that it cannot be maintained. Referring to dare-mo/nani-mo with or without a co-occurring case-marked N as wh-mo, and assuming that the entire phrase (e.g., gakusei-ga daremo) functions as an argument, I argue, based on how wh-mo interacts with XP-sika, that wh-mo has a universal force and therefore must c-command Neg at LF to induce universal negation.