The Wh/Clitic Connection
There is a further link one can make between wh-elements and some notion of topicality. In this final section, I will discuss how the phenomenon of clitic doubling in Bulgarian, as well as other languages, correlates with fronted wh-phrases, or rather vice versa. All in all, we can observe a close connection between topical wh-phrases and clitics—or, alluding to the work of Boeckx and Stjepanovic´ (2005), there is more to the ‘‘wh/clitic connection’’ than meets the eye.
Citko and Grohmann (2001) advance the bold thesis (expanded in Grohmann 2000a, 2003a) that German is like Bulgarian. The main parallel, they argue, is that both languages have the obligatory fronting requirement on all wh-phrases in the overt component, presented above. In an intriguing recent paper, Jaeger (2004) presents a curious set of data from Bulgarian that is much more in line with the view on Ger man multiple wh-questions espoused here. Moreover, the analysis Jaeger suggests bears striking similarities to the one presented here. Let me summarize the data and suggest an interpretation that fits nicely into the present set of assumptions, drawing also from joint work with Cedric Boeckx.
The parallelism between German and Bulgarian multiple wh-question formation hypothesized elsewhere faces the problem that Bulgarian interrogative syntax has a well-known property that German lacks: all wh items are obligatorily fronted en bloc or adjacent to one another (Rudin 1988); that is to say, fronted wh-phrases may not be split up. Rudin already offered some data pertaining to parentheticals (see Bosˇkovic´ 2002 for recent discussion). More relevant to the present issues, another type of constituent that may occur in the left periphery and is presumably the result of movement may not intrude either: topics. Richards (1997, 2001) presents data involving the interaction between topics and fronted wh-elements which suggest first, that the two are not incompatible, and second, that topics must precede, rather than follow, the wh-constituent(s).
(1)

(2)

The same effects can be found in Romanian (Ileana Comorovski, pers. comm.). What makes Romanian interesting is a fact first observed by Dobrovie-Sorin (1990). This language allows clitic doubling, and as it turns out wh-phrases may be doubled by a clitic just in case they are D linked—in fact, D-linked wh-phrases must be doubled in this way, as the contrast in (3) shows. (The particle pe occurs only in D-linked questions, for reasons that do not bear on the present issue.)
(3)

Non-D-linked wh-words, on the other hand, do not allow doubling at all.
(4)

Picking up on the discussion in Billings and Rudin 1998, Jaeger (2004) now examines a curious property of (colloquial) Bulgarian: clitic doubling is allowed not only with overtly marked D-linked wh-elements, but with bare wh-words as well. Consider these contrasting structures:
(5)

An important initial observation is that the doubled wh-phrase precedes all others—in particular, a subject wh-phrase (see Jaeger 2004 for discussion). This fits in well with the proposal developed above for German: clitic-doubled wh-phrases seem to underlie different syntactic operations than nondoubled ones. The particular operation one could envision here, with the proposal from German in mind, is topic movement. One result of the many Rizzi-inspired works on the left periphery, I take it, is that they have identified a crosslinguistic tendency that topics precede wh items if the two co-occur. The data in (1)–(2) suggest that this holds at least for Bulgarian.
And indeed, Jaeger explores an account along the same lines. He introduces a definition of topicality that I will not recreate. Suffice it to say, the two approaches are compatible in many respects. One discourse function of doubled wh-phrases in (colloquial) Bulgarian is illustrated by the following context and the question relating to it:
(6)

To the best of my knowledge, the interpretation of the question would be analogous to something one could reconstruct in English as ‘For each of our friends, who painted him?’ or simply ‘Which friend was painted by which artist?’. (Jaeger offers the latter as the sole translation, reflecting native speakers’ intuitions about the interpretation.)
Another property of clitic-doubled multiple wh-questions in (colloquial) Bulgarian is that the doubled wh-phrase may also precede a D linked subject.
(7)

By now, it should come as no surprise that D-linked wh-phrases may also be doubled, as shown in (8). Here, one function of the doubling clitic seems to be disambiguation, since, as the pair in (9) shows, D linked wh-phrases may front in either order (in which case the interpretation is, of course, unambiguous, as suggested already in (8); see Jaeger 2003, 2004).
(8)

(9)

I leave the discussion of Bulgarian at this point. Clearly, much more can be—and needs to be—said, but that will have to wait for future research. The goal of this section was to draw a connection between wh elements that are in one way or another linked to the discourse (wh topics) and another discourse-heavy element of grammar, clitics. This wh/clitic connection has been noted in different contexts and might be worth exploring further.