The locality conditions of head movement have long been a debate in movement theory. It is often suggested that head movement is exceptionally restricted compared to phrasal movement, which is blocked by intervening heads (i.e., the Head Movement Constraint). I argue against the presence of head-movement-specific locality conditions, and propose instead that both head movement and phrasal movement are subject to the same set of locality constraints, i.e., both of them are disrupted by featural interveners, instead of structural interveners. The empirical foundations come from an array of verb doubling constructions in Cantonese, where a copy of a verb appears in the (left or right) periphery. I observe that these constructions do not only obey island constraints and allow for cross-clausal dependencies, but they also exhibits the so-called Focus Intervention Effects, where intervening focused elements disrupts the dependencies between verbs. However, heads do not constitute a similar blocking effect in these constructions, constituting empirical challenges to the Head Movement Constraint. I suggest to assimilate these movement properties to phrasal movement such as wh-movement and focus movement.

Head movement lacks semantic effects in an array of cases, and appears to be substantially distinguished from phrasal movement. I present empirically challenges to this distinction by observing that quantificational heads such as aspectual verbs and modal verbs in Cantonese exhibit a high vs. low positional alternation under certain circumstances. I propose that they can undergo scope-shifting head movement, which redefines the quantificational or focus scope of the quantificational heads. The suggested movement shares a crucial property with Quantifier Raising: both of them are constrained by an economy condition on scope interpretation (i.e., Scope Economy).

Doubling effects of predicate clefts are attested in language after language. While it appears that verb movement is exceptional in allowing multiple realizations as compared to its phrasal counterpart, data in Cantonese suggest that this is at best a tendency, rather than a generalization. I observe that (i) not all cases of predicate clefts require verb doubling and that (ii) not all cases of phrasal movement disallow doubling. I propose to derive the different doubling patterns of head movement and phrasal movement through the different launching positions of these elements, instead of some particular properties of head movement. The proposal is implemented under a cyclic linearization/spell-out approach to syntactic structure, revealing the tension between minimizing movement copies and maximizing the easiness of the linearization process. Most importantly, the very mechanism of linearizing syntactic structure do not distinguish head movement from phrasal movement.

Whether or not raising of subjects/objects out of a CP (i.e., hyperraising constructions) is allowed divides languages into two classes. In my joint work with Ka-Fai Yip (Yale), we observe that the legitimacy of subject raising in both Cantonese and Vietnamese independently reveals a correlation with the requirement of indirect evidence of the embedding predicates. For example, a subset of attitude verbs such as ‘guess’ and ‘believe’ allow hyperraising constructions in these languages, but not other verbs such as ‘think’ and ‘know’. This language-internal variation calls for a more fine-grained analysis on the availability of hyperraising constructions. We propose that while CP is by default phase, it may be selectively “opened up” by a Agree relation with the attitude verbs that encodes indirect evidence. The proposal implicates that the phasehood of CP interacts with syntactic dependencies and locality conditions exhibit certain degrees of flexibility. (Joint work with Ka-Fai Yip)

While vP is often characterized as the first clause-internal phase, independent evidence shows that there is certain flexibility of the minimal size of this first phase. In my joint work with Victor Junnan Pan (CUHK), we establish further support for this idea based on the licensing conditions of VP movement and ellipsis in Mandarin and Cantonese. The core observations build on the inability of certain aspectual elements, such as progressive markers and aspectual verbs, to license VP fronting, VP postposing and VP ellipsis. These elements are in contrast with perfective markers, for example. To derive the distinction between different aspectual elements in this regard, we develop an account that assigns a privileged status to phases in licensing VP movement and ellipsis, and motivate a contextual approach to phasehood. Substantially, the aspectual phrase that hosts imperfective elements, when projected, replaces vP to be a phase, rendering the complement vP a non-phase (hence unmovable and unelidable). In other words, the size of the first phase is determined contextually. (Joint work with Victor Junan Pan)

How movement chains are pronounced is a central issue of any theory of linearization (i.e., the mechanism that maps structures onto linear strings). As an attempt to reveal the underlying properties of this mechanism, this talk zooms in onto two non-canonical realizations of movement chains, namely, (i) verb doubling (where gaps are fully pronounced), and (ii) discontinuous predicates (where gaps are partially pronounced). Drawing on evidence from a range of verb movement in Cantonese, I propose that the Economy condition serving to minimize copies (i.e., members a movement chain) compromise when confronted with (i) Syntactic Cyclicity and (ii) the Faithfulness condition on copy integrity. Non-canonical or Economy-violating patterns arise as a result of avoidance of violation of other hardwired principles.

Copy Deletion is held responsible for erasing extra copies created in movement chains, and partial (copy) deletion is proposed to derive discontinuous phrasal constituents such as split topics in German and left-branch extraction in Serbo-Croatian. In principle, nothing a priori rules out the possibility of partial deletion applying to head chains. In my joint work with Ka-Fai Yip and Sheila Shu-Laam Chan, we fill the theoretical gap through an in-depth study of discontinuous predicates in Cantonese, where a (disyllabic) predicate is “split” into two parts by elements such as {verbal suffixes} and wh-expressions. We propose that discontinuous predicates in Cantonese are resulted from a combination of a syllable deletion rule induced by affixes in the language and a subsequent application of partial deletion. The proposal implicates that partial deletion is not restricted phrases but applies also to heads. It also reveals that partial deletion applies instead of (full) copy deletion so as to avoid a violation of a faithfulness condition when determining copy pronunciation. (Joint work with Ka-Fai Yip, Sheila Shu-Laam Chan, and Carmen Tang)

Right dislocation posits various puzzles in the theory of linearization. Based on evidence from Cantonese right dislocation, I propose to derive the surface word order in right dislocation through multiple leftward movements. The first movement step involves A’-movement (defocalization) that may target both heads and phrases, followed by a subsequent movement of TP to the specifier position of the projection headed by sentence-final projections. The proposal incorporates unexplained data in previous accounts on right dislocation in Cantonese, and provides further support to head movement to specifier position. I also look at the information structural status of right dislocated elements in Cantonese and Mongolian (based on my fieldwork on Alasha Mongolian.

In languages without morphological marking of tense, it is a debate whether they are tensed languages with covert morphology or genuinely tenseless languages. In my joint work with Maria Luisa Zubizarreta and Roumyana Panchenva, we pursue an account that invokes no use of tense to derive temporal reference in Cantonese. We propose that temporal reference is expressed via evaluation time shift, a mechanism adopted to explain narrative present in English. We argue that the notion of (neo-Reichenbachian) tense is unnecessary in Cantonese, and falls short of capturing temporal reference in this language. The proposal implicates that tense is not a universal semantic notion. (Joint work with Roumyana Pancheva and Maria Luisa Zubizarreta)

The relative strength of necessity modals (e.g., must vs. should) receives a lot of attention in the literature. A question arises as to whether a similar distinction is found in possibility modal. I suggest that the expression you-liyou ‘have-reason’ in Mandarin exemplifies such a possibility. The expression is logically stronger than the possibility modal keyi ‘may/can’ and it stands in dual relation with yinggai ‘should’. The proposal exhausts the four possibilities entailed by two parameters in modal discussions: force and strength.

A recent line of research focuses on the modal meaning in the nominal domain. In a number of languages, certain grammatical markings on indefinites convey an ignorance component typically oriented to the speaker, opening up the discussion of epistemic modality in the nominal domain. I conduct a case study of epistemic indefinites in Cantonese, which arguably take the form of m-zi ‘not-know’ + wh-expressions. I argue that the associating ignorance component is a presupposition (instead of a conversational implicature) encoded by mzi, which is grammaticalized as a choice function. The proposal enriches the landscape of epistemic indefinites and sheds light on the interpretation of wh-expressions in Cantonese.