The syntax of question particles

Cross-linguistically, languages are largely head initial or head final. Most permit some disharmony, but Holmberg (2000) and Biberauer, Holmberg & Roberts (2012), among others, have argued that the structure shown in (1) is ruled out, where YP is X’s complement and ZP is Y’s complement: (1) *[XP [YP Y ZP] X] In structures such as (1), a head-final phrase immediately dominates a head-initial phrase, violating the so-called ‘Final-Over-Final Constraint’ (FOFC). Descriptively however, final question particles are readily found in languages with VO order, resulting in a structure that appears to violate FOFC. (2) illustrates this violation in Tetun (an Austronesian language of East/West Timor), and (3) shows the structure, with a final question particle ka immediately dominating a head-initial TP: (2) ó la bá sekola ká? 2S not go school or (Said to child playing:) ‘Didn't you go to school?’ (Van Klinken 1999: 212) (3) iii If ka constitutes the C head of CP, as is standardly assumed, the structure in (3) violates FOFC. I show, following Aldridge (2011), that these particles are best analysed as disjunctive elements, heading an elided clause: (4) [ConjP CP [Conj CP]] The particle is the head of the phrase, with the second CP as its complement and the first (pronounced) CP in Spec,ConjP. This solves the FOFC problem because the ‘particle’ is not final, and therefore the derivation does not include a head-final phrase dominating a head-initial phrase. Instead, the particle precedes its complement (which is not pronounced), and the clause that it follows (which is pronounced) is its specifier. I provide evidence for this position through typological investigation and theoretical analysis. In addition, the various proposals that have been put forward in the literature to avoid this FOFC-violation are considered, but are shown to be problematic in different respects. I discuss the idea that particles are not heads (Biberauer, Holmberg & Roberts 2012). However, they cannot be specifiers and an adjunction analysis fails to explain their properties, so it is unclear what they could be if not heads. Julien (2001), Lee (2005, 2008) and Simpson & Wu (2002) argue that final particles are derived by TP-movement to a Topic or Focus position. This is a promising explanation, but fails to derive the difference between final particles and other types. If the particle is syncategorematic (Biberauer, Holmberg & Roberts 2012), the fact that they appear in fixed positions is mysterious. Processing explanations of the data (Hawkins 2004, Philip 2012) go some way towards deriving the FOFC facts but do not, among other things, explain the high number of final particles in VO languages. The syntax of question particles is discussed in detail, and it is proposed that polar questions consist of two functional heads in combination: Force, giving a (main clause) question illocutionary force, and Polarity, giving a (neutral) iv question open polarity. A true polar question particle is therefore related to one or both of these heads: (5) With this background, the argument is defended in subsequent chapters that some particles cannot be true question elements in this sense and are instead instantiations of the disjunction. Cross-linguistic data demonstrate that final particles in VO languages differ from other types of question particle (initial particles, or final particles in OV languages) in very rarely marking embedded questions: they do so in only one language in the corpus. Homophony between the question particle and disjunction in many languages, combined with attested grammaticalisation paths, adds support to this claim. Furthermore, this analysis explains a number of properties of such particles in addition to their propensity to violate FOFC, including their frequent absence from negative questions, alternative questions and wh-questions. All of these are straightforward consequences of the particle being a disjunction. Finally, the analysis is applied to a particular language, Thai, as a case study, and it is compared with languages of the other types. It is shown that the disjunctive analysis is best able to explain the data and offer an elegant explanation of the FOFC facts.