In the first study I looked at locative phrase processing in an eye-tracking study and observed effects of phrase frequency and
relative entropy that are novel in the psycholinguistic literature on Mandarin Chinese. In addition, I found that fixation patterns are influenced not only by the lexical properties of the
current word, but - through parafoveal processing - also by those of upcoming words. The more information the current word contains, the longer and the more often it is fixated on. By contrast, a
greater amount of information in the upcoming word leads to shorter and fewer fixations on the current word. The pattern of results therefore suggests that locative phrase processing in Mandarin
Chinese is a highly dynamic process, in which readers actively seek to identify the locus of information and allocate resources to it.
In the second line of research, I have looked at the
performance of young and old Chinese-German bilinguals and young and old German monolinguals in a paired associate learning task. For German monolinguals and Chinese participants in Chinese the
typical pattern of data reported in the literature was observed, with decreased performance for older people. For the Chinese-German bilinguals in German, however, no age effect was present. This
argues against an interpretation of the age effect of monolinguals in terms of cognitive decline. By contrast, consistent with the predictions of discrimination learning models, it suggests that
the decreased performance of older monolinguals is a consequence of increased experience with the language. Such an interpretation is supported by the fact that bilinguals consistently
outperformed monolinguals Germans and performed better in their second language than in their native language.
I have been analyzing the results of a mega study in which I collected 30,830 naming latencies for two native speakers of Mandarin Chinese. In a principal components analysis (PCA) I found effects related to the frequency and visual complexity of (the characters in) a word, the number of homophones and homographs of both characters and the entropy over the character frequencies of two-character words. An initial analysis of the eye-movements recorded during the naming task showed comparable results. I am now exploring the possibility of predicting naming latencies through a naive discrimination learning model for Mandarin Chinese. If this attempt succeeds, it would demonstrate that the explanatory power of discrimination learning networks is not limited to alphabetical languages, but works for character-based languages as well.
I developed a lexical database for simplified Chinese, the Chinese Lexical Database (CLD). For each of the 30,645 one and two
character words for which she collected naming latencies in 2014, the CLD contains over 140 lexical variables. The lexical
variables in the CLD encode information related to the frequency, the visual complexity, the phonology, and the consistency of the orthography-to-phonology mapping for all words in the database, as well as for their component characters and semantic and phonological radicals. The CLD is by far the largest lexical resource for Chinese,
both in terms of the number of words in the database and the number of lexical
variables for each word.
All frequency-based measures in the CLD are based on frequency counts from a new corpus created by Ching Chu Sun and Cyrus Shaoul, which consists of websites written in simplified Chinese: the Simplified Chinese Corpus of Webpages (SCCoW). The SCCoW consists of 466,551,657 words and 773,697,216 characters. Frequency measures from the SCCoW outperform currently available frequency measures for Chinese in explaining naming latencies for 1 and 2 character words in Chinese.
The CLD will be made publicly available on a dedicated webpage, both in the form of a download of the full database and through an online search interface. The online search interface offers a custom selection of words, characters and radicals for which lexical variables should be shown. Furthermore, to allow for more targeted searches, levels of categorical variables and ranges of numerical variables can be restricted. The results of a search can be displayed in an interactive browser window or e-mailed to the user.
Supervisor: Dr. John Newman
In the second project, I described locative constructions in Southern Min, a language spoken in Taiwan and parts of China. I found that while locative scenes in Southern Min are expressed in a unique way, some descriptive characteristics are shared with other languages. This suggests that at least some of the structure of the semantic space underlying locative expressions may reflect general cognitive mechanisms that are similar for the speakers of different languages.
Supervisor: Dr. David Beck
The first project, involves the lexical class of adjectives in Igbo, a language spoken in Nigeria. Traditionally, it is believed that the lexical class of adjectives in Igbo consists of only 8 items (Welmers, 1974). Using Beck (2000 & 2002)'s theory of markedness and unidirectional flexibility, I established that this believe is incorrect and that the class of adjectives in Igbo should be extended to include at least 20 words.