Questions, biases and ‘negation’: evidence from Scots varieties
Jamieson, Elyse Anne
In recent years, there has been considerable linguistic interest in ‘non-canonical questions’ (Dayal 2016). These constructions express some quantity of interrogative meaning, as well as some additional bias, or some additional request beyond pure information seeking. This category includes four constructions with ‘negation’, which look superficially similar but have received varying analyses in the literature: matrix biased questions; tag questions; polar rhetorical questions, and interrogative exclamatives. In this thesis, I address the question of how these constructions are related, and argue that the relationships between these four ‘non-canonical questions’ are closer than the existing literature suggests, building on and extending work from Sudo (2013) and Domaneschi et al. (2017) who establish a framework for using speakers’ epistemic beliefs and the biases provided by the evidential context to account for matrix biased questions. In order to develop this overarching argument, I focus primarily on two constructions that have received little attention in the literature: –int in Glasgow Scots and –n in the Shetland dialect of Scots. I adapt an acceptability judgment style methodology for dialect syntax research to incorporate recent work on non-canonical questions in experimental pragmatics in order to establish the distribution of these particles across constructions and belief/bias contexts. As part of this, I also qualitatively investigate how speakers of different ages in communities with different relationships to linguistic change interact with this common methodology for investigating linguistic variation, positing what I term ‘perceptual hyperdialectalism’ for the patterns of behaviour we see in an obsolescing variety. From the results of the Scots research, I show that –int is acceptable in a subset of the ‘non-canonical questions’, with –n available in the full set for older speakers but seemingly undergoing loss in one context and moving towards the distribution of Glasgow –int for younger speakers. This suggests that we should indeed treat the non-canonical question constructions as more closely related than the literature suggests. I then provide an analysis for the –int and –n constructions, arguing that they are check moves by establishing the pragmatic similarities between the constructions that permit the particles and then developing a semantic analysis for them within Ginzburg’s (2012) interactional semantics for dialogue. The final part of this analysis is to position the particles syntactically. I do so by employing a conversational domain in the syntax above CP, and by arguing for movement to this domain I build on existing literature that shows that discourse particles exhibit syntactic behaviour and should be analysed as such. My analysis of the Scots data has three main contributions: firstly, a full description and analysis of the previously understudied Scots particles; secondly, an analysis for standard English tag questions that clearly shows how they differ from the Scots constructions and is also able to deal with problematic data from the literature more accurately than other proposals; thirdly, new evidence that there needs to be more fine-grained distinctions made, based on beliefs and biases, between the types of conversational move that are made in e.g. tag questions and invariant particles, often grouped together as ‘confirmationals’ or ‘checks’. As well as the Scots data, I also address the relationship between the four non-canonical questions through two sets of constructions in standard English. Firstly, I present the results of an experiment which shows that although both matrix biased questions and tag questions are permitted in neutral and negative evidential contexts, speakers prefer biased questions in negative contexts, and tag questions in neutral ones. I show that this follows from the analysis I presented for the syntax and semantics of tag questions in standard English, and suggest that the results point towards a scalar model of beliefs and biases as the best way to understand the licensing of interrogative constructions. Finally, I look at polar rhetorical questions and rhetorical wh-questions. In the literature, these constructions receive the same analysis throughout. However, the Scots data indicates that the bias that is, or can be, expressed in polar rhetorical questions is not the same as the bias that is expressed in rhetorical wh-questions. I argue that polar rhetorical questions should receive the same analysis as matrix biased questions, following Romero’s (2015) falsum/verum approach to the construction, and show that this cannot hold for wh-questions. I then extend Kotek’s (2016) semantics for wh-questions to also include rhetorical wh-questions, showing how this can account for a number of properties that these constructions have (e.g. polarity flipped vs. non-flipped instances, NPI licensing, lack of pair-list readings and ‘generic’ interpretations).