No description available.

No description available.

When hearing speech, listeners begin recognizing words before reaching the end of the word. Therefore, early sounds impact spoken word recognition before sounds later in the word. In languages like English, most morphophonological alternations affect the ends of words, but in some languages, morphophonology can alter the early sounds of a word. Scottish Gaelic, an endangered language, has a pattern of 'initial consonant mutation' that changes initial consonants: Pog 'kiss' begins with [ph], but phog 'kissed' begins with [f]. This raises questions both of how listeners process words that might begin with a mutated consonant during spoken word recognition, and how listeners relate the mutated and unmutated forms to each other in the lexicon. We present three experiments to investigate these questions. A priming experiment shows that native speakers link the mutated and unmutated forms in the lexicon. A gating experiment shows that Gaelic listeners usually do not consider mutated forms as candidates during lexical recognition until there is enough evidence to force that interpretation. However, a phonetic identification experiment confirms that listeners can identify the mutated sounds correctly. Together, these experiments contribute to our understanding of how speakers represent and process a language with morphophonological alternations at word onset.

Kin Fai Ellick, Wong. / Thesis (M.Phil.)--Chinese University of Hong Kong, 1997. / Includes bibliographical references (leaves 74-83).

No description available.

Compounding is one of the most productive methods to construct words in different languages, e.g. joining the words “super” and “man” gives the compound word “superman”. For decades, researchers are interested to know how compound words are stored and retrieved in the lexicon. Different theories of lexical storage and retrieval of compound words were proposed to explain the compound word processing observed in both normal and abnormal adult subjects. However, little studies have attempted to apply these theories to explain the developmental pattern of storage and retrieval of compound words. To fill the gap, the major aim of the current study is to investigate the power of different theories of lexical storage and retrieval of compound words in explaining the typical and atypical development of compound word processing in Chinese children. Altogether, 20 grade 2 children, 22 grade 4 children, and 17 grade 6 children screened to have normal non-verbal intelligence and reading abilities were recruited from a local mainstream school. Three experiments were conducted to investigate the existence of the holistic representations of compound words and the representations of their constituent morphemes in the lexicon, and their involvement during the compound word retrieval processes across different grade levels. Results show that grade 4 and grade 6 children demonstrate significant whole-word frequency, morphological family size and semantic transparency effects in all three experiments, a pattern which resembles that observed in normal adult subjects. The grade 2 children, however, only demonstrate significant whole-word frequency effect but not the morphological family size and the semantic transparency effect. The results indicate that grade 4 and grade 6 children adopt the partial-decomposed approach of compound word storage and retrieval (e.g. Taft, 2003). As for the grade 2 children, it is hypothesized that their performances represent a developing stage of the partial-decomposed approach, where networks of morphological relations between family members were underdeveloped in their lexicon. Further investigation of the compatibility of the partial-decomposed approach in explaining the compound word storage and retrieval pattern resulted from atypical development was conducted. The three experiments mentioned above were administered on 16 poor readers (PR), 16 reading-level-matched (RL) peers and 16 chronological-agematched (CA) peers. Interestingly, the PR group’s performances resemble that of the RL and CA group in experiments of whole-word frequency and morphological family size but not in experiment of semantic transparency. The PR group’s performances can be explained by assuming a deficit in identifying shared semantic features between compound words and their constituents in the partial-decomposed approach. It is proposed that the PR group identifies frequently occurring morphemes as salient orthographic reading units without recognizing the shared semantic features between compound words and their constituents. In summary, results of the current study support the partial-decomposed approach of lexical storage and retrieval of compound words. The current study further proposes (i) a developing stage of the partial-decomposed approach to explain the compound word processing within an under-developed lexicon and (ii) a deviated partial-decomposed approach to explain the compound word processing of children with reading difficulties. / published_or_final_version / Speech and Hearing Sciences / Doctoral / Doctor of Philosophy

published_or_final_version / Psychology / Master / Master of Philosophy

A connectionist architecture is developed that can be used for modeling choice probabilities and reaction times in psychophysics and word recognition. The network architecture consists of a feed-forward network and a decoding module. Learning is by mean-variance back-propagation, an extension of the standard back-propagation learning algorithm. The new learning procedure is interpreted as a selective attention mechanism, and leads to a better model of learning in simple identification tasks than the standard back-propagation. Choice probabilities are modeled by the input/output relations of the network, and reaction times are modeled by the time taken for the network, particularly the decoding module, to achieve a stable state. The model is applied to both unidimensional and multidimensional identification tasks in psychophysics and to word recognition.

No description available.

Six experiments were conducted to address the role of phonological information in visual word recognition. A semantic decision task was used to ensure that word meanings were accessed. Experiments 1-4 showed that subjects make more false positive errors on homophone foils (e.g. living thing-FLEE) than on spelling controls (e.g. living thing-FLEX) only when both members of the homophone pair are uncommon and are similarly spelled. In Experiment 5, there was an increase in errors on low but not high frequency homophone category exemplars when they were preceded by a word related to the other member of the homophone pair (e.g. SHATTER-BRAKE). In Experiment 6, subjects produced longer decision latencies on homophone exemplars than on semantic controls only when they were low in frequency. These results indicate that, even in skilled readers, phonological information mediates the access of meaning for low frequency words, and that orthographic activation also contributes to the activation of their meanings.