resentations
While pointing out the lack of evidence for embodied conceptual metaphors, it is not my intention to suggest that metaphor comprehension is not grounded in any past bodily experiences. It is unprofitable for researchers to be limited by the current definition of simulation based on the conceptual metaphor theory. Instead, we should align our view of metaphor processing with the current understanding of grounded language processing, that is to define the simulation involved in metaphor processing simply as the partial re-enactment of sensory, motor, introspective, affective and other experiences related to the concept. In other words, when we use the phrase “a bright student”, we do not need to picture a student with a light-bulb flashing above their head; but we should still reactivate our physical, emotional and psychological experiences with a quick-witted student in order to achieve comprehension.
To propose a new definition of simulation in metaphor processing is in- deed the first challenge I will face in my thesis. Since simulation is complex and subconscious (Connell & Lynott, 2016), this definition needs to capture the ef- fort for simulation as a whole, allowing us to remain agnostic about the specific experiences reactivated. Throughout my thesis, this definition will be scrutinised and validated in both behavioural studies and EEG studies. Furthermore, I will study how embodied simulation can be coupled with linguistic distributional pat-
terns. Specifically, I will investigate the conditions under which each component is involved, and their interplay to create conceptual representations.
1.4.1 Thesis Outline
Thus, in my thesis, I will first propose an index of ease of simulation and generate a body of metaphoric sentences that vary independently on the measure of ease of simulation and the linguistic distributional frequency (Chapter 3). On the one hand, the ease of simulation measure will capture metaphor comprehension as a whole, measuring the effort required to arrive at a successful representation. On the other hand, linguistic distributional frequency will be defined in a minimal fashion, capturing the distributional patterns of metaphors’ constituent words in close proximity, within a narrow five-word window (e.g., Louwerse & Connell, 2011). This treatment of linguistic distributional frequency will advocate strongly for the efficacy of the linguistic information in following chapters.
Then I will investigate whether metaphor processing relies on both embod- ied simulation and linguistic distributional information, as the grounded approach suggests in the following chapters (Chapters 4-7). If both components play a role in metaphor processing, then the ease of simulation and linguistic distributional frequency should each contribute to the performance of metaphor processing with a unique portion. Furthermore in Chapter 4, I will test the hypothesis that the linguistic component can act as a shortcut when the task requires only shallow processing. Thus, I will contrast shallow and deep processing with two sentence processing tasks as Connell & Lynott (2013): that is, a shallower sensibility judge- ment task and a deeper interpretation generation task.
Studies in the following chapters (Chapter 5 and 7) will develop upon the same paradigm as the study in Chapter 4, in order to further examine the inter- play between ease of simulation and linguistic distributional frequency. Chapter 5 will continue to test the linguistic shortcut hypothesis and investigate the role linguistic distributional frequency plays under time constraint. If people use the distributional pattern of words as a shortcut to make judgments, the effect should be larger when the time resource is limited. Accordingly, the effect of embodied simulation will be smaller when the time resource is limited because to engage in embodied simulation is time-consuming.
Then in Chapters 6 and 7, I will search for the neurophysiological manifes- tations of the embodied and linguistic components using electroencephalography (EEG). Chapter 6 will identify these two components in literal language process- ing, and Chapter 7 will match them with metaphor processing. Using EEG will offer us greater insights into the online processing during metaphor comprehen- sion with high temporal resolution. However, the conventional practice among the EEG literature is problematic and prone to Type I error, for reasons discussed in the next chapter (Chapter 2). Therefore, apart from studying metaphor process- ing, another mission of my research in these chapters will be to develop a more rigorous way to analyse EEG data for psycholinguistic research.
Finally in the conclusions chapter (Chapter 8), I will summarise the theoret- ical and empirical advances in this thesis, specifically in relations to the linguistic shortcut hypothesis. I will discuss, first, how well current theories of grounded conceptual representations can be applied to the research on metaphor processing. Second, I will also discuss the implications of my thesis on grounded language pro-
cessing. Since starting point of my thesis is to cast metaphor processing under the big framework of language processing. My findings will increase our understand- ing of conceptual representations in general and point out a direction for future studies on grounded language processing.
CHAPTER
2
Using EEG to Study Metaphor Processing
2.1 Introduction
In my thesis, as discussed in the previous chapter, I will study the con- ceptual representation during metaphor with the specific question regarding the role of linguistic distributional patterns in mind. To test the linguistic shortcut hypothesis, it is important to validate a crucial condition for the linguistic compo- nent to become a shortcut: that is the linguistic component reaches the peak of its activation before the embodied component. In other words, it is necessary for the linguistic component to have a temporal advantage over the embodied component (though both components are activated simultaneously) for the linguistic compo- nent to become a valid guide for the processing of embodied simulation. Therefore, a crucial component of my research is the use of electroencephalography (EEG).
In order to study the grounded representation with EEG, it is important to identify and differentiate the effect of the embodied and linguistic components on the EEG waveforms. In previous research, the study of EEG waveforms primarily
relies on the identification of event-related potential (ERP) components, which are segments of EEG waveforms circumscribed to a period of time and an area on the scalp. Several ERP components have been found to be associated with semantic processing and even embodied conceptual representations (e.g., N400 and late positivity complex). However, as well be discussed in this review, the study of ERP components suffers from great problems, such as high researcher degrees of freedom and the use of ANOVA. Therefore, it is a mission of my research to seek a more rigorous way to study EEG, by limiting researcher degrees of freedom and using better statistical tools (i.e., linear mixed-effect model plus Bayes factor). In this chapter, I will first review available findings of the ERP markers of conceptual representation and argue why the current way to study EEG lacks the needed rigour. Finally, I will propose a better way to analyse EEG data, which will be used in studies of Chapters 6 and 7 in my thesis.