Evidence for Multiple A -Movement
I want to consider additional evidence supporting the assumption that even the lower wh-phrase WH2 moves overtly quite high up in the structure. Leaving aside the details of how topicality and interrogativity are to be married (section 10.4), the proposed step must be an instance of what used to be called A ¯-movement; the landing site, whether Spec,TopP or any of the other candidates, is arguably an A ¯ position (terminology I continue to use for purely expository reasons).
If all wh-elements do indeed A ¯-move overtly into an articulated C domain in German, we would expect them to behave syntactically in ways that go beyond word order restrictions or intervention effects, such as A ¯ binding. As it turns out, such effects are notoriously difficult to show for German. But interestingly, we can find some—and while these do not necessarily suggest A ¯-movement proper, like wh-movement in English, they do suggest some sort of movement, even (and especially) of the lower wh-element. The following discussion is taken from Grohmann 2003a, but the data are considerably enriched: whereas the arguments provided there suggested that even the lower wh-element, WH2, moves somewhere, there is now nearly conclusive evidence that it moves into the C domain, as an A ¯-element.
Consider parasitic gaps, for example, which are licensed by overtly A ¯ moved elements only—at least in English (pace Nissenbaum 2000; see, e.g., Hornstein 2001 and Nunes 2004 for recent discussion and ample references about this core property of parasitic gaps and others). As it turns out, the lower wh-element in German may license a parasitic gap ( pg), thus at first glance suggesting overt A ¯-movement.
(1)

Before we can use parasitic gaps as a bona fide argument in favor of overt displacement of the lower wh-element into the C domain, it has to be shown that German indeed has parasitic gaps analogous to those of English. But since wh-phrases do not scramble in German (unlike, say, in Japanese; see Grewendorf and Sabel 1999), the landing site of the lower wh-phrase cannot be the typical scrambling position. In effect, the licensing of parasitic gaps by the lower wh-phrase can indeed be integrated into the analysis that the landing site is within the C domain, and even as things stand now concerning the status of parasitic gaps in German, data such as (38a–b) do not provide a counterargument to my proposal—quite the opposite, in fact, as the ensuing discussion shows.
Note first that the A ¯-licenser of the parasitic gap has to c-command it in the overt syntax (Webelhuth 1992); thus, the adjunct clause containing the parasitic gap cannot scramble over the licenser (or whatever analysis one wants to apply to German word order, clause structure, and phenomena like scrambling).
(2)

With this in mind, consider the following contrasts:
(3)

Regarding the mixed A/A ¯-character noted by Webelhuth, Mahajan (1990) offers an interesting solution for German: prior to the step of A ¯-movement licensing the parasitic gap, the licenser undergoes A movement (from the landing site of which it can license anaphors, one of the issues of contention). I adopt this basic line of analysis, which, interestingly, should also work for anyone’s take on headedness of the German VP (see, e.g., Sabel 1996, 37n26). In other words, regardless of the particular derivational operations and structural positions involved, the parasitic gap is eventually licensed from an A ¯-position and contrasts with lower positions (scrambled noninterrogatives or multiple wh-phrases as traditionally conceived). This approach makes a number of testable predictions.
One concerns the contrasts in (3). The grammaticality of (3a) would be expected since we have seen that an adverb like sorg fa¨ltig ‘carefully’ must follow all wh-phrases. This said, the ungrammaticality of (3b) is equally expected since the adjunct containing the parasitic gap is scrambled over its licenser, the lower wh-phrase was ‘what’. Interestingly, now, (3c) is bad: as the parasitic gap seems to be correctly positioned with respect to its licenser, one might hypothesize that what rules this structure out is the relative ordering of the adverb and the lower wh-phrase, for the by now well-known reasons.
A similar line of reasoning can be constructed for a different type of ad verb. Frey (2000) identifies the adverb vermutlich ‘presumably’, among others, as a marker for sentence topics: what stands to its left must be a topic. Consider some relevant data:
(4)

Both ordering relations between adverb and lower wh-phrase are possible—at first glance unexpectedly under current assumptions, since, as a topic marker, vermutlich ‘presumably’ should follow was ‘what’. (One property of topic markers is that they do not allow the topic to follow.) Admittedly, Frey and I disagree on the implication that topic markers (and his discussion of sentence topics in general) carry for the syntax. He identifies the notion ‘‘sentence topic’’ with a Mittelfeld position, that is, within the German middle field (here, roughly IP-internal). But it seems to me that unless the analysis hastily sketched above is fleshed out in full detail, we can remain agnostic about exactly which projection in the structure is identified here. Rather, I would like to focus in on the contrast between (4) and (5).
(5)

While the lower wh-phrase may follow the topic-indicating adverb, it may not license a parasitic gap; this it can only do from the preadverbial position. (If the sentence receives any interpretation, the adverb in (5a) can only be understood as modifying was ‘what’.)
This suggests that (4) contains two quite different structures, one in which the wh-phrase has moved to a left-peripheral topic position, and one in which it has not. So far, I have not considered the latter possibility at all. In fact, in previous work I have always argued that multiple wh fronting is obligatory in German. With data such as the above at hand, however, this conclusion might have to be revised. At this point, I do not see any difficulties with such a revision; on the contrary, it might come in handy. Consider, for example, the following examples:
(6)

(7)

Drawing on Frey’s work again, we see that the structures in (6)–(7) contain two relevant elements: the expression zum Glu¨ck ‘fortunately’ serves a topic-marking purpose similar to vermutlich ‘presumably’, while the expression [NP jedenfalls ‘at least’] must refer to some aforementioned NP; that is, it must be a topic (and as such cannot follow the topic marker, as in (7d)). The clear contrast in the complex examples here tells us one thing for certain: the lower wh-phrase is in a higher position in the (a) examples than in the others. This conclusion suffices for the purpose of this section: suggestive evidence that the lower wh-phrase moves overtly into a high left-peripheral position with A ¯-character exhibiting properties reminiscent of topics.