An Alternative Derivation of Multiple Wh-Questions in German
Is there anything interesting for the analysis of multiple wh-questions in German that would relate to the conclusion that apparent interveners are nontopicalizable? I want to argue for a resounding yes. After all, if real, such nontopicalizability would be a striking property.
To begin with, the purported role of topicalizability looks rather obscure under a traditional analysis of multiple wh-questions in German, whose derivation would be analogous to that of multiple wh-questions in English: one wh-phrase fronts into the left periphery overtly, the other stays in situ (with ‘‘?P’’ = VP; see von Stechow and Sternefeld 1988).
(1)

Working from a rough structure like (1), I want to propose a radical alternative. After all, how could we tell from (1) that the intervener, an illegitimately topicalized nontopicalizable element, is indeed in a topic position? What is a topic position, anyway? Let me briefly lay out some background assumptions and sketch my analysis (drawing from the more detailed expositions in Grohmann 1998, 2000a, 2003a).
First of all, topicalization as understood here is an operation in the overt component; it is the result of movement (some form of the rule Move wh from Chomsky 1977), targeting a position like Spec,CP—or rather, within an articulated Comp, the position explicitly identified by Rizzi (1997) and others as Spec,TopP. That is, minimally (building on (1)), we are dealing with something like (2).
(2)

(2) is then our starting point for the structures at hand. The potential intervener—or rather, (non)topicalizable constituent—is signaled by XP: as a topic, it must move to Spec,TopP, and if it is nontopicaliz able, it cannot move there. Our first stab at formulating a ban thus yields (3).
(3)

Now the 64,000-dollar question is how we can ensure that the structures discussed above actually do involve some XP in a purported topic projection. In other words, can we show that what appears in between two wh-phrases in German is by necessity in a syntactic topic position? Working with standard assumptions as in (1), the answer is clearly no—and the whole endeavor is jeopardized. But with a simple suggestion first advanced in Grohmann 1998, we get a positive answer instead.
Assume that WH2 does not sit so low in the clause structure (‘‘in situ’’). Assume rather that, for some reason, WH2 has moved quite high, into the left periphery (i.e., our C domain), in the overt syntax. Assume further that this movement of WH2 also targets Spec,TopP. And finally, assume that in addition to WH2 and XP, WH1 occupies Spec,TopP and that fronted topics like to stick together.
We now have a potentially more interesting picture.
(4)

The line of the argument is clear: if both wh-phrases are topics and if they can sandwich only topics, then only topics may appear in between them. In other words, when we see any structure involving two wh-phrases, we know one thing for sure: whatever sits in between the two sits in Spec,TopP. Problem solved.
As nice as this argument may look, it carries a big if—and the big if consists at least of the number of assume thats, rephrased as follows:
(5)

What I have done in previous work is motivate (5a–c) with some variation. In the interest of space, I will confine myself to a brief exposition.
The specific details arguably play only a minor role, so that we can basically consider (6) to contain the most salient ingredients of the alternative analysis.
(6)

Obviously, the argumentation I have developed elsewhere does not involve some deus ex machina, but derives the desiderata on mostly empirical grounds (see the works cited). We can then just adopt (6) as make-belief in the absence of a more elaborate discussion in the present context. The single most important aspect of the derivation involved, however, is the topichood of both wh-phrases in a German multiple wh question
Given the explosion of CP in (6) qua articulated C layer as in (7b) or any version thereof, and the purported landing sites for the overtly A ¯ moved wh-phrases of (6) indicated in (8) (Grohmann 2003a, 114), any material appearing between the two wh-phrases must target TopP. This particular analysis (as developed in Grohmann 2000a, chap. 5) is a modified Q-based analysis of wh-question formation, what I call elsewhere the ‘‘Hagstrom-Bosˇkovic´ approach’’. More in line with what I have said above, and will say below, about (so far, purported) wh-topics, one might actually want to adopt an even further refined structure of the C domain, such as (7c). Again, I believe that the details of the proverbial devil, important though they certainly are, are not critical for present purposes.
(7)

(8)

Under this analysis, questions are licensed by a question morpheme [Q] that has to end up in C, in a position high enough to take scope over the entire structure. As Hagstrom (1998) lays out in detail, [Q] may move on its own, from a high base-generated position (here, F0), or it may tag along with the lowest wh-phrase (here, WH2), its alternative locus of generation. Bosˇkovic´ (2001) exploits this system to derive pair-list and single-pair readings, slightly revising Hagstrom’s original mechanism, which is further refined in Grohmann 2003a. As argued there, a derivation as sketched in (8) accounts not only for the topicalizability issues discussed so far, but also for the peculiar behavior of German multiple wh-questions with respect to the readings they allow (or require). For the latter account to work, nothing hinges on the exact (labels of the) positions, as long as the relational properties of (8) remain. In other words, we would not lose anything from the explanatory power of that analysis if the landing sites of WH1 and WH2 were in fact the same (type of) projection, Spec,TopP.
Note further that pursuing an alternative to the observed word order restrictions as hastily sketched above does not depend at all on the specific generalization, that is, on either universal-based intervention as in (3) or monotonicity-inspired topicalizability concerns— empirical shortcomings of either aside. It does not matter how we define inherent properties of illegitimate in-betweeners; what matters is that any constituent XP that wants to follow one wh-phrase and precede another (WH1,WH2) must be a topic. If it cannot be a topic, for reasons that go beyond this investigation, it cannot appear in-between, but must strictly follow.