Reserach projects
PhD project:
Production, perception, and processing of focus in Turkish
© Chris Hellier, 2016
Communication is more than the mere transmission of information in the form of the words’ meanings. Language in coherent communication requires consideration of what was said previously and what communicative partners know or want to know at any given moment or, to be more precise, what the speaker assumes the partners know or aim to know (see Krifka & Musan, 2013). In other words, information in language is ‘packaged’ according to the communicative context and the informational states and needs of the interlocutors (see Chafe, 1976). Focus as one such information package contributes information to the knowledge of communicative partners, inducing alternatives in comprehension.
The main goal of this dissertation is to experimentally investigate how focus is realised, perceived, and processed by native Turkish speakers, independent of preconceived notions of positional restrictions. Specifically, we were interested in how the focus dimensions such as focus size (comparing narrow constituent and broad sentence focus), focus target (comparing narrow subject and narrow object focus) and focus type (comparing new-information and contrastive focus) affect Turkish focus realisation and, in turn, focus comprehension when speakers are provided syntactic freedom to position focus as they see fit. To provide data on these core goals, we presented three behavioural experiments during this dissertation: (i) a production task with trigger wh-questions and contextual animations manipulated to elicit the focus dimensions of interest, (ii) a timed acceptability judgment task in listening to the recorded answers in our production task, and (iii) a self-paced reading task to gather on-line processing data.
Master’s project:
Simulating Grammatical Encoding with Multi-Label Decision Tree Classifiers An Investigation into the Loci of Impairment in Turkish-Speaking Individuals with Agrammatism
The locus of impairment in agrammatism has been a widely debated issue in the field of aphasiology, with scholars like Bastiaanse and van Zonneveld (2004) arguing for the importance of the grammatical encoder in this regard. Most models of speech production assume grammatical encoding to be an integral part of syntax and morphosyntax formation. However, because grammatical encoding is a largely automatic and unconscious process, it cannot be directly studied through behavioural experiments. One can merely speculate as to its role based on theoretical assumptions and the method of exclusion. The present study investigates the role of lexical access and grammatical encoding by simulating a grammatical encoder in finite verb phrase production in Turkish. The Turkish finite verb paradigm has been chosen as it is both highly complex and regular, with the latter characteristic excluding the possibility of finite verbs being retrieved from the lexicon. In line with Bastiaanse and van Zonneveld’s proposal, it is hypothesized that damage to the lemma alone cannot result in agrammatic target patterns when verb inflexion, root production, and argument construction are considered. Rather, it is proposed that damage to the grammatical encoder itself leads to the target patterns postulated.
Based on Levelt’s (1989) model of speech production, three multi-label decision tree classifiers were constructed and trained using the Scikit-Learn environment (Pedregosa et al., 2011) in Python. These decision tree classifiers were either themselves damaged as systems, indicating damage to the grammatical encoder, or their inputs were damaged to varying degrees. The inputs were binary strings, constructed to resemble the lemma and the preverbal message. These strings were processed to outputs in the form of labels, indicating necessary argument positions, verb roots, and inflectional suffixes.
Bachelor’s project:
Towards a description of the autolalic future in Dzongkha - An interview-based qualitative study
Commissioned drawing of Tashichho Dzong (བཀྲ་ཤིས་ཆོས་རྫོང) by www.nuru.ch
The suffix -གེ་ནོ་ -geno marks in Dzongkha, the official national language of the Kingdom of Bhutan, a specific future form. Termed the autolalic future in its, to date, only linguistic description by van Driem (1998), its core function is the expression of a first-person subject’s (plural and singular) intent, solely used in self-directed, internal speech, and only assigning a first-person singular subject. Therefore, autolalic future constructions are never uttered and only appear written as direct citations in narratives. This description raises several questions: i. is the autolalic future primarily a tense marker, or rather an extended form of the adhortative suffix -གེ་ -ge, from which it is claimed to be derived, ii. how can the suffix express a first-person plural subject’s intent if it can only assign a first-person singular subject, and iii. how do Dzongkha learners acquire this suffix, if it is never uttered and only rarely written? To answer these questions, and to gain more insight into the autolalic future’s functionality, this study investigates this suffix utilizing qualitative, semi-structured interviews in English.
For this study, a total of eight participants (mean age 35.5, range 27-53 years) agreed and gave informed consent to take part in an interview. Due to the status of Dzongkha as a lingua franca, only one of these participants was a native speaker of Dzongkha, with the remaining six participants being L2 speakers who acquired Dzongkha in school. However, a chatroom with eleven active participants—nine of which were natives of Dzongkha—was maintained to gather additional grammaticality judgments and further evidence regarding inconclusive statements by the interviewees.