Clause-medial stance markers in Cantonese ͳ Epistemic, evidential and attitudinal markers in clause-medial position in Cantonese Foong Ha YAP and Winnie Oi-Wan CHOR Hong Kong Polytechnic University and The Open University of Hong Kong Abstract This paper examines two types of clause-medial stance markers in Cantonese. 1 We trace the evo- lution of these stance markers from their lexical origins as complement-taking psych verbs and directional verbs. Similar to English I think parentheticals, the evolution of the Cantonese psych verbs (e.g. gok 3 dak 1 ‘feel/think’; paa 3 ‘fear’; m 4 zi 1 ‘don’t know’) involves ‘insubordination’ of the embedded complement clause. However, in Cantonese, subject ellipsis and topic construc- tions play an important role in the reanalysis of psych verbs into epistemic and negative attitudi- nal markers in clause-medial position. 2 Directional verbs in Cantonese (e.g. faan 1 ‘return’ and maai 4 ‘approach’), on the other hand, develop into attitudinal particles via verbal complementa- tion and a disjunctive strategy. Central to the development of both types of stance markers is the first person subject ngo 5 , whether explicitly or implicitly expressed, which makes the process of subjectification possible. Keywords: stance markers, utterance-medial (or clause-medial) markers, parentheticals, direc- tional particles, insubordination, subject ellipsis, topic constructions, verbal complementation, disjunctive strategy 1. Introduction In our daily conversations, we provide many linguistic cues for others to interpret our stance. For example, in terms of epistemic stance, we deploy a rich array of expres- sions to indicate our degree of certainty (or lack thereof) toward our own utterances or the utterances of others. In English, these epistemic strategies include modals (e.g. may), adverbials (e.g. probably), and epistemic phrases (e.g. I think) (Thompson & Mulac 1991; Kärkkäinen 2003; see also Wischer 2000 for a diachronic analysis of 1 We at times use the term ‘utterance-medial position’ instead of ‘clause-medial position’ in this paper. This is because, in the course of grammaticalization of some Cantonese stance markers, their position is sometimes ambiguous between clause-initial and clause-medial. As will be shown in our discussion of stance markers derived from complement-taking verbs in section 2, this is often related to the pres- ence of NPs whose syntactic status is ambiguous between a topic and subject. 2 The term ‘reanalysis’ is often used in grammaticalization theory to refer to shifts in readings arising from ambiguous contexts, while the term ‘reinterpretation’ is often used to refer to recategorization of functions and is often discussed in relation to polysemy rather than ambiguity. In this paper, we do not always make a strict, clear-cut distinction between these two terms given that both mechanisms are at work, often multiple times, in the grammaticalization of the ‘say’ construction from complement- taking predicate to quotative, evidential and other pragmatic uses. We wish to thank Elisabeth Leiss for valuable discussion on terminology here.