Syntactic priming as a window into the representational and experiential basis of syntactic processing in comprehension

• Author(s): Fernandes, Eunice G.
• Date: 2015
• Persistent ID: http://hdl.handle.net/10451/20606
• Origin: Repositório da Universidade de Lisboa
• Subject(s): Teses de doutoramento - 2015; Domínio/Área Científica::Humanidades::Línguas e Literaturas
• Document Type: Doctoral thesis
• Language: English
• Advisor(s): Costa, Maria Armanda, 1952-; Branigan, Holly Parima; Ignazio Coco, Moreno
• Contributor(s): Repositório da Universidade de Lisboa

Description

Tese de doutoramento, Ciência Cognitiva, Universidade de Lisboa, Faculdade de Ciências, Faculdade de Letras, Faculdade de Medicina, Faculdade de Psicologia, 2015

This thesis investigates syntactic processing during comprehension in its representational and experiential components, i.e., the mental linguistic representations we access and the way we use them when we comprehend sentences. Our main claim is that syntactic priming is a behavioural tool that can give important insights into both syntactic knowledge and performance. Syntactic priming, the tendency of speakers to repeat syntactic structure across otherwise unrelated utterances, was shown to be well suited to investigate the mental representation of verbs by Pickering and Branigan (1998), who showed that speakers repeated the previously processed variant of an alternation verb. They proposed the two possible argument structures had different representations at the syntactic level of its lexical entry, and that syntactic priming results from residual and transient activation of those representations. Substantial research produced corroborating evidence of such effects, typically assessed within-trial, i.e., in adjacent sentences. More recently, however, intriguing results showed that the magnitude of the effects was also a function of how much the alternative syntactic structures were experienced, for example, during an experiment. Jaeger and Snider (2007, 2013) demonstrated that priming was stronger in two apparent contradictory circumstances: participants were more likely to repeat a structure the more they had experienced it (cumulativity) but also the less that structure was expected (surprisal). The results support implicit learning accounts of priming. Explanations of syntactic priming remain under debate. In a series of experiments, we give new insights concerning both syntactic representations and the mechanisms associated with experiencing syntactic structures. In two self-paced reading experiments (Chapter 3) we test sentences containing locative alternation verbs, for which linguistics has provided solid, but contrastive, theories of lexical representation. Contrastingly, to our knowledge, it has not been tested in more than one syntactic priming study (Chang, Bock, and Goldberg, 2003). We demonstrate that the two variants of this syntactic alternation do not elicit the same magnitude of priming. Our results provide supporting evidence for theories postulating different lexical representations of the variants (e.g., Rappaport & Levin, 1988), and seem to contradict theories of a unique and similar representation (e.g., Borer, 2005). They contrast with previous research findings (Chang et al., 2003) and suggest frequency issues must be considered when investigating representations through experimental tasks. We test how the relative amount of experience with alternative syntactic structures affects syntactic priming in a visual-world paradigm task, where spoken sentences are understood situated in a visual context (Chapter 4). Here, we analyse anticipatory eye-movements to entities depicting possible arguments of a verb (e.g., Arai, van Gompel, & Scheepers, 2007), which index the comprehenders’ expectations about forthcoming linguistic input of the spoken sentence. We test how temporary ambiguity of relative clauses (RCs) (that can high- (HA) or low-attach (LA) to a complex NP, e.g., The father of the baby who will drink the beer / the baby bottle is tall) is resolved when participants are presented previously with sentences containing disambiguated HA or LA RCs. Our original manipulation is, however, of the amount of priming received within a trial: participants can be presented with one or two written primes before they receive the target. This design is intended to test the dynamics of priming at a short-term, departing from previous research that tested the dynamics of the effects along longer-term timeframes such as an experience. We report classical priming effects in the condition where participants are presented with a single prime. However, we find surprising effects of reversed priming when participants are exposed to two primes. That is, after experiencing two sentences with the same attachment type (e.g., LA), participants anticipate the alternative, non-primed type (HA). We speculate that one possible explanation of our findings is repetition suppression, the reduced neural activity upon stimuli repetition (Grill-Spector, Henson, & Martin, 2006). These results challenge the assumption of a positive and linear relationship between the amount of (short-term) experience and the strength of priming effects, and provide new evidence concerning short-term activation mechanisms underlying priming. We further explore the cognitive complexity underneath RCs’ processing by examining (Chapter 5) the relation between reading, viewing and working memory (WM) in the data collected in Chapter 4 (restricted to the condition where participants were presented with two primes). Here, we investigate pronoun resolution as indexed by looks, while hearing the pronoun who in The father of the baby who will drink…, to the depicted possible pronoun antecedents (e.g., to the FATHER (S1) under a HA interpretation). We predict fixations to S1 as a function of: i) second-pass reading time observed during reading the NP1 region in the prime sentences (e.g., LA, The helpers in The helpers of the baker who will [SG]…), and ii) individual WM scores. We demonstrate that high-capacity individuals anticipate more the (non-primed) visual referent when they reread more the corresponding NP (e.g., anticipate the visual referent FATHER when reread more often NP1, the helpers). We suggest that WM capacity allows individuals to maintain alternative syntactic analyses of the sentences while reading, and evaluate them upon a subsequent visual context. These findings provide support to constraint-based models of sentence processing (e.g., Just & Carpenter, 1992), and provide, for the first time, linking evidences of online measures during reading and online measures of subsequent image viewing (indexing the transfer of information from reading to viewing) and its mediation by working memory. Overall, this thesis reunites the fundamental questions of linguistic competence and performance through syntactic priming research. It highlights the potential of this research tool by providing new evidence relevant for both the old theoretical puzzle of argument realization and the topic of the cognitive mechanisms involved in language processing, of broad and current interest.