On the Processing of Single-character Words in Chinese Reading: Evidence from Eye Movements

中文閱讀中單字詞的加工研究:來自眼動的證據

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date24 Aug 2020

Abstract

In modern Chinese, most characters manifest their capacity to stand as single-character words (SCW) and account for a large proportion of lexical tokens in daily use. From a universal linguistic perspective, their wordhood is beyond all doubt as they meet the definition of word and can pass most wordhood tests available. In psychological studies of reading, however, there has been a view that they are not necessarily segmented as words and often processed in a way as if they were sublexical morphemes (SLMs) and instead, multi-character words seem to be the focus of word segmentation and cognitive relevant processing. Given the essential difference between characters as SCWs and those as SLMs, that the former are independent and the latter subordinate, the cognitive processing of the two has to be different. This dissertation is specifically aimed at investigating if and how the way of our processing of SCWs really differs from that of SLMs and multi-character words in a real scenario of reading. To achieve this goal, three studies by empirical experimentation have been conducted for this thesis research.

The first study consists of two experiments that were performed for the purpose of examining whether SCWs are processed in the same way as non-word morphemes. In the first experiment via natural reading, no evidence was found for any significant difference in reading performance between real and subjective words (i.e. phrases) of two characters. But with spaces inserted into these target strings as word boundary markers, the second experiment showed that a good performance in terms of total fixation duration could be maintained on targets of two SCWs but not on those of two SLMs, providing evidence suggesting that an SCW is processed differently from a single-character morpheme in Chinese reading. This difference has been observed for the very first time in this research. It is covert in natural reading mostly due to the prosodic effect that tends to override the contrast between SCWs and SLMs. With the inserted space to disrupt prosodic structure and to induce readers to process characters as SCWs, the wordhood effect appears.

Unlike the first study to examine wordhood effect in a special scenario of reading with extra spaces inserted, the second study is aimed at revealing how a word-level property of characters modulates natural sentence reading. Wordhood probability of a character, estimated by the quotient of its character frequency divided by its word frequency to measure how likely a character is to be a word, was selected as the word-level property for examination. A large-scale eye movement corpus of Chinese reading under development at City University of Hong Kong was adopted as real data to fulfill this goal. Characters from three string patterns under five conditions were chosen as targets. By a series of mixed-effects regression analysis, it is found that wordhood probability of character has facilitative effect on reading characters as either SCWs or SLMs, be they in foveal or parafoveal vision. This further suggests that characters are firstly activated as SCWs in reading, irrespective of their wordhood in a particular context; yet the SCWs are not segmented. The wordhood probability effect of a morphemic character also appears to be modulated by the processing difficulty of the whole word in question. Besides, the effect in parafoveal vision may affect saccade target selection.

The third study investigates how and when SCWs are segmented in natural reading. An experiment was carried out to compare eye movements on embedded SCWs in two conditions, plausible vs. implausible in terms of the preceding context. In the plausible condition, both an embedded SCW and its matrix two-character word (TCW) fit the preceding context while in the implausible condition, the matrix word fits but the embedded SCW does not. The result shows that the whole matrix TCW region is more likely to be skipped in the plausible condition whereas in the implausible condition, the last fixation before the target region was longer when the region was skipped than when it is fixated. The difference of skipping rate between the two conditions is ascribed to the fact that the early parafoveal segmentation of SCWs occurs more often when an embedded SCW fits its preceding context. The duration difference caused by the contrast of skipping in the two conditions is attributed to the use of context information to avoid incorrect segmentation that would cost extra effort. SCWs manifest a special case that word segmentation can be accomplished in parafoveal vision, which has to presume that boundaries of SCWs are already visually detected.

In summary, this dissertation presents our original research to examine several important characteristics of SCWs that have interesting effects on Chinese reading, demonstrating that SCWs play an indispensable role in psychology of reading as well as in linguistics. More future studies are still required in order to incorporate the reading of SCWs into existing reading models.