The production and perception of English lexical stress by Hong Kong Cantonese learners of English

Abstract

Fewer perception studies on suprasegmental features as compared to segmental ones, fewer studies targeting Cantonese ESL learners as compared to Mandarin ones, as well as previous findings on tone-stress relationship and F0 significance led to the need for this study.

This study aimed at identifying the dominant cue, among F0, intensity and duration, in English lexical stress production and perception by Hong Kong Cantonese learners of English (CS), the difference between CS and native English speakers (NS) and that between highly proficient and less proficient CS (HCS and LCS).

Four experiments were carried out. The first three experiments involved the production of (a) seven disyllabic English noun-verb pairs, (b) 11 polysyllabic English words, and (c) six disyllabic/trisyllabic English donor words and their corresponding Cantonese loanwords respectively. Experiment 4 involved the perception of the disyllabic English noun-verb pair of “contract” produced by a female NS speaker, with either syllable undergoing F0, intensity or duration manipulation.

Experiments 1-3 were further divided into acoustic and perceptual studies. The acoustic studies involved 70 CS (with 11 HCS, 11 LCS and 48 with moderate English proficiency) and 14 NS participants producing tokens analysed with Praat, whose F0 (Hz), intensity (dB) and duration (ms) measurements were statistically analysed and compared. The perceptual studies involved 13 NS participants listening to CS tokens and performing noun-verb distinction in Experiment 1, rating acceptability in Experiments 1 and 2, and rating similarity between the English donor words and corresponding Cantonese loanwords in Experiment 3. Thirty CS listeners served as controls for Experiments 1 and 3, and 42 CS for Experiment 2. Experiment 4 was a perception test involving these 42 CS listeners performing noun-verb distinction, whose accuracies were compared with seven NS control listeners’.

The findings revealed CS’ F0 dependence but intensity negligence in both English lexical stress production and perception, suggesting a possible perception-production link and L1 influence. HCS’s overuse of F0, and better use of intensity and duration than LCS suggested the effect of L2 proficiencies.

The results supported Motor Theory (MT), postulating a specialised module between production and perception, rather than General Auditory and Learning Approaches (GA) and Direct Realist Theory (DRT). Two-Category Assimilation (TC) and Uncategorised versus Categorised (UC) in Perceptual Assimilation Model (PAM) explained the assimilation of English lexical stress and non-stress contrast while Speech Learning Model (SLM) and Native Language Model (NLM) explained the warped perceptual space.

Furthermore, HCS’s heavy reliance on F0 and LCS’ little awareness of intensity and duration suggested explicit teaching/audio-visual training to redirect attention to neglected dimensions, which may support Attention-to-Dimension Models (A2D). Future studies should include the vowel quality cue, a larger pool of participants, more appropriate tokens, and include investigation of vowel duration’s relationship with F0 and speech rate.

It is anticipated that this study can inspire English teachers or language educators for the development of a more effective English lexical stress teaching regimen and serve as a basis for future studies on CS’s acquisition of English prosody such as speech rhythm, sentential stress and intonation.