The landscape of neg-words

In many languages sentences containing a negation and a negatively marked nominal expression are interpreted as containing a single negative element, and the same may hold for sentences containing multiple negatively marked nominal expressions. This is illustrated in (1) and (2) for Italian: although both sentences contain two negative expressions, the negative marker non ‘not’ and the negatively marked nominal expression nessuno ‘nobody’, (1-2) are interpreted as containing a single negative element. These sentences convey the meanings ‘Nobody called’ and ‘Nobody called anybody’, respectively, rather than the meanings ‘Nobody did not call’ and ‘Nobody called nobody’. This phenomenon of multiple negative expressions contributing a single negation is known as negative concord and languages exhibiting the phenomenon, such as Italian, as negative concord languages.

Negative concord behaviour exhibited in (1-2) becomes particularly puzzling when coupled with the observation that, to the extent that the negatively marked nominal expressions may occur on their own, they contribute a negative meaning. This is illustrated in (3). Accordingly, one would expect (1-2) to convey doubly negated meanings, contrary to fact. 

It is commonly agreed that this puzzling semantic behaviour has to do with the underlying properties of negatively marked nominal expressions in negative concord languages (we refer to such expressions henceforth as n-words after Laka 1990). Various theories have been proposed to explain this remarkable behaviour of n-words. According to some proposals (e.g., Zanuttini 1991, Haegeman 1995, Haegeman & Zanuttini 1996, De Swart & Sag 2002, Watanabe 2004, De Swart 2010), n-words are negative quantifiers (e.g., nessuno means something akin to ‘no person’), and some kind of semantic absorption mechanism accounts for why the n-word and negation in (1) and the two n-words in (2) are interpreted as contributing a single negation. On these proposals, the two n-words ‘fuse’ together at the level of interpretation. Such approaches, however, face the challenge of explaining why in (1) the presence of the negative marker is obligatory, as demonstrated again for Italian in (4) (whereas omitting negation is perfectly fine in a non-negative-concord language like English). Hence, without adopting further assumptions, n-words cannot be treated as negative quantifiers that ‘fuse’ together at the level of interpretation with another expression.

In light of this, other scholars (e.g., Ladusaw 1992, Brown 1999, Giannakidou 2000, Weiss 2002, Zeijlstra 2004, Penka 2010) have argued that n-words are semantically non-negative elements that for some reason always appear in the scope of a negation. In this sense, they resemble so-called Negative Polarity Items (NPIs) like English anybody and in weeks, whose distribution is also restricted (see e.g. Fauconnier 1976, Ladusaw 1979, Giannakidou 1999, Gajewski 2011, Chierchia 2013, Crnič 2014a, among many others; some authors argue that n-words are, in fact, a kind of NPIs, e.g., Giannakidou 2000, Chierchia 2013):