Multi-gap dependencies, such as ATB movement and parasitic gaps, constitute a puzzle for syntactic theory because they violate the strict one-to-one mapping between a filler and a gap. Over the last few decades, multiple `tools' have been proposed to model this unusual one-to-many relation, some of them unifying ATB and parasitic gaps under a single approach, others treating them as unrelated. These tools either aim to re-shape how we think about movement via multidominance or sideward movement, or apply independently established mechanisms, like, for example, ellipsis or operator movement. Despite their fundamental differences, nearly all tools remain to coexist, and it appears that the field prioritizes proposing new theories over identifying which of them has the best empirical coverage. The lack of clarity regarding the empirical facts themselves might be the main cause of this. On the one hand, different researchers focus on different types of diagnostics, ultimately yielding theories that empirically over- or undergenerate. On the other hand, there are certain diagnostics whose outcomes researchers outright disagree about. After illustrating these circumstances and motivating the need for more, but also more structured means of data collection in chapter 1, the aim of this thesis is to take a step towards `decluttering the toolkit' by adopting a more data-driven, experimental approach to the study of multi-gap dependencies. Instead of presenting different types of evidence in support of a single theory, this thesis deals with one particularly controversial piece of evidence and its implications for the plethora of existing theories. The core of the thesis investigates the contested observation that principle C reconstruction of nominal PP modifiers reveals an asymmetry between the initial and non-initial gap in ATB and parasitic gap constructions in extraposed adjunct clauses. In chapter 2, two experiments on principle C reconstruction of PP modifiers in German demonstrate that the effect is detectable in simple wh-dependencies, but not universally among speakers, items or even experimental setups. Then, in chapter 3, the limited cross-linguistic comparability of ATB movement and `English-type' parasitic gaps is motivated by an acceptability judgment experiment of German parasitic gaps in extraposed adjunct clauses, revealing that the majority of participants judge parasitic gaps to be marginal or entirely unacceptable. The investigation of principle C reconstruction in German multi-gap dependencies thus focuses exclusively on ATB movement, presenting three experiments that severely challenge the impact of principle C reconstruction on coreference resolution. The methodological insights from chapters 2 and 3 are then applied to the study of principle C reconstruction in English multi-gap dependencies in chapter 4. A multi-part experiment shows the limited but non-negligible relevance of the basic principle C reconstruction effect, the vanishing of its impact under ATB movement, and the strong resemblance of the coreference pattern observed in ATB movement compared to parasitic gaps. A follow-up experiment studying the effect of structural vs. linear distance on the coreference pattern in parasitic gaps in purpose clauses delivers inconclusive results, and I discuss methodological complications supported by post-hoc statistical analyses. Based on the findings from chapters 2-4, I argue throughout the thesis that the apparent asymmetry between the gaps with respect to principle C does not need to be accounted for in the syntactic derivation of ATB movement and parasitic gaps. What instead needs to be accounted for is the lack of robustness in principle C reconstruction not only with increased distance, but particularly with increased dependency complexity. In turn, the previously reported principle C reconstruction data do not constitute evidence for the movement path of the filler in multi-gap dependencies, and do not support asymmetrical derivations in which only the initial gap is created via movement. Chapter 5 recapitulates the findings presented in this thesis and what the orthogonality of principle C reconstruction implies for existing approaches to multi-gap dependencies. Finally, I discuss how we could approach reconstruction in multi-gap dependencies in future work, with a special focus on which types of reconstruction could lead to more firm conclusions about the derivation of ATB and parasitic gap dependencies.
Timea Szarvas (Thu,) studied this question.