On the internal structure of q-words

It is well known that the external syntax of wh-items is partly related to their internal composition. Rizzi (2002) proposes that the distinction between wh-words and wh-phrases is due to the fact that wh-words cannot be interpreted as Topic-like elements, while wh-phrases can. The same distinction is made by Grewendorf (2011) for German wh-items: while the element was ‘what’ is a pure operator, wh-phrases can be topic-like because they contain an overt nominal restrictor. In the literature on Romance languages, it has been pointed out several times that wh-words have a different distribution with respect to wh-phrases, so that they have been analyzed as having a different internal structure with respect to wh-phrases. Already Ambar (1988) claims that the distribution of wh-words and wh-phrases in Portuguese is different, Munaro (1999) notices that only bare wh-words can be left in situ in Bellunese, while wh-phrases have to be moved to the left of the clause, whether they are d-linked or not. To the best of my knowledge, it has never been pointed out that a similar di-chotomy between bare and non bare forms is found with quantifiers as well. In this article I will provide empirical evidence mainly coming from Italian varie-ties and Cimbrian, a German dialect spoken in the Italian Alps, that there exists a structural distinction in the internal makeup of QPs which distinguishes bare-QPs from quantified nominal expressions. This is so in French, Cimbrian but it is particularly clear when the distribution of bare quantifiers like tutto/tutti /everything/everybody/all’ in Old Italian is investigated in comparison with the distribution of quantified expressions containing the same Q followed by a whole DP. Old Italian shows that, while quantified expressions display the same distribution as definite DPs, bare Qs have a dedicated position. Old Italian also clearly shows that this cannot be due to a supposed structural “weakness” of the form, which might be a clitic or weak pronoun (in the sense of Cardinaletti and Starke (1999)), as it has often been proposed for the distribution of French bare Qs like rien, tout, tous etc. I will propose that this difference reflects a fundamental structural distinction in the type of internal projections activated inside the QP. Empirical evidence that this is so comes from Italian varieties: several Southern Italian dialects display a classifier-like noun paired to the quantifier, but which clearly shows that the internal structure of bare Qs is different.