Year: 2020
Type: dissertation
Abstract: This dissertation investigates how speech sounds may change as a function of varying working memory load during speech. While cognitive factors have long been suspected to influence speech sounds, being responsible for unexplained variability in speech prosody, the hypothesized connection between cognition and speech production has yet to be systematically theorized with empirical data. Despite the widely accepted assumption in psycholinguistics that working memory involves speech production for temporary maintenance and active manipulation of speech information, the sound planning process remains an open question. Production models not only acknowledge that they only partially account for the phonological-phonetic encoding processes, but production models also adopt direct retrieval of memorized articulatory routines from long-term memory that may allow speakers to bypass the working memory process. Moreover, a comprehensive review from a memory perspective requires direct evidence and reexamination of the psycholinguistic assumption, suggesting that there is no working memory used in speech production. Furthermore, if the key lies between active ongoing planning and automatic routined response, then native speakers and nonnative speakers of a language may reveal different production patterns. These speaker types and their production patterns have not yet been directly compared in the literature. \n\tTo address this gap, this study asks whether working memory is involved in English speech production and whether the involvement differs between native (first-language, L1) and nonnative (second-language, L2) speakers of English. To explore how cognitive processing overload derails typical speech sounds, in this study I experimentally manipulated the amount and the type of cognitive load in working memory by having participants engage in multitasking while speaking. Native speakers of American English and Korean EFL learners of English produced a set of syntactically complex sentences under two working memory load conditions and two control conditions. In the load conditions, speakers engaged in an additional task that taxed either verbal or spatial working memory while speaking. In the control no-load conditions, they solved an equation before speaking. \n\tThe results show different working memory influences on L1 vs. L2 speech production. Both the L1 and the L2 speech had more errors and were faster during multitasking. However, the effect of load type was different. The L1 speech remained similarly intact, regardless of whether the task was verbal or spatial. By contrast, the L2 speech was impaired only by the verbal task; speech produced during a concurrent verbal task became more erroneous, faster, less variable in word durations, and less distinctive in vowel quality. The spatial task did not impact L2 prosody. \n\tI interpret these results as suggesting that different cognitive processes underly L1 and L2 speech production. This dissertation proposes a tentative model for L1 and L2 speech production with direct reference to working memory versus long-term memory. L1 speech production is usually automatic and spontaneous via retrieval from long-term memory and not impacted by (verbal) working memory. Thanks to sufficient exposure to and practice of form-sound units, speech production directly accesses long-term memory and instantly executes the articulatory gestural routines that are already stored along with the lexical forms. As L1 speech production draws from long-term memory, working memory is not overloaded by a (verbal or spatial) working memory task, and speech becomes only more erroneous and faster due to divided attention within a set time frame. By contrast, L2 speech sounds mainly emerge on the go via phonological-phonetic encoding in (verbal) working memory. Due to limited language experience in L2, articulatory templates are often not fully developed or readily retrievable for automatic execution. As L2 speech production draws from verbal working memory, a concurrent verbal task, but not a spatial task, overloads the capacity-limited verbal working memory component. While a general effect of divided attention can be observed just as in most multitasking, a verbal task disrupts speech more significantly than a non-verbal task. \n\tThe study contributes direct evidence regarding the relationship between working memory and speech production. I examine some speculated effects of cognitive load on prosodic variations and move toward resolving the controversy over the encoding and the retrieval processes by referring to language experience. Beyond the scope of this dissertation and as a guidepost for research, we need to investigate how speech production processes may develop as a function of L1 language experience and L2 proficiency. \n \nKeywords: working memory, working memory capacity, speech production, prosody, phonological-phonetic encoding, retrieval, L1 versus L2 \nStudent Number: 2006-30916
Author: 이옥영
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W3124286083