Perceptual Learning of Cantonese Lexical Tones by Tone and Non-Tone Language Speakers

Introduction

The study examines the perceptual learning of Cantonese lexical tones by two groups of listeners: native speakers of a tone language (Mandarin Chinese) and those of a non-tone language (English). Previous research indicates that acquiring a foreign language requires listeners to adjust their perception towards acoustic patterns they may have previously ignored or misinterpreted. Specifically, these listeners must accommodate new prosodic features that differ from their native language. The investigation focuses especially on tones, as Cantonese is a tonal language where meaning changes with variations in pitch.

Tone Identification in Languages

Cantonese and Tone Languages

In tonal languages, such as Cantonese, meaning is conveyed through changes in the fundamental frequency (f0) pattern. For instance, the Cantonese syllable [ji] can mean "cure" with a High Level tone (55) or "son" with a Low Falling tone (21). The current study highlights that while f0 (pitch contours) is crucial for identifying tones, in other languages like Mandarin, features like amplitude envelope and duration serve as secondary cues. However, for Cantonese, no consistent non-f0 cues for tone identification have been established.

Perception and Processing Differences

Research shows notable differences in how tone and non-tone language speakers process speech f0 both behaviorally and neurologically. Non-tone language speakers find it especially challenging to learn tone languages due to their lack of experience with f0 as a lexical feature. Two theories explain this difficulty:

Levels of Representation suggests that non-tone speakers struggle to relate tones to familiar linguistic categories because their native language lacks the requisite prosodic structures.
Category Assimilation argues that these speakers attempt to map foreign tones onto their existing intonational categories, resulting in perceptual confusion.

Methodology

The current study involved participants from a tone language background (Mandarin speakers) and a non-tone language background (English speakers). They underwent training to recognize Cantonese tones, after which their performance was measured through identification tasks and pairwise difference ratings. The aim was to shed light on how their native language experience informed their ability to perceive and categorize tones both before and after training.

Initial Results: Pretraining Perception

Initially, both groups exhibited comparable performance in tone identification, but significant differences in tone-specific challenges were apparent. For example, Mandarin speakers identified the Cantonese tones that closely matched their native tones more effectively, while English speakers had trouble with tonal contrasts unfamiliar to them. This result underscores that even baseline performance is shaped by the listeners' native linguistic categories.

Tone Inventory Analysis

Mandarin Tones

Mandarin features four main tones: High (55), Rising (25), Dipping (214), and Falling (51). These tones exhibit consistency in fluent speech, though exceptions occur with tone sandhi, affecting tonal realization based on adjacent tones. While some non-f0 features such as amplitude envelope contribute to Mandarin tone recognition, f0 cues remain the primary determinant.

Cantonese Tones

Cantonese is distinguished by its six tones, including High Level (55), High Rising (25), Mid-Level (33), Low Rising (23), Low Level (22), and Low Falling (21). Tone identification requires careful perception of f0 patterns, and difficulties experienced by cochlear implant users in tone recognition emphasize the importance of f0 as a sole cue in Cantonese.

Results and Training Effects

Both Mandarin and English speakers showed significant improvement in tone identification through the training process. While performance improvements were roughly equal between groups, differences emerged in specific tones where each group struggled. These performance changes post-training demonstrated how exposure and practice can enhance perceptual sensitivity to unfamiliar tones.

Confusion Patterns

Confusion matrices revealed distinctive patterns where Mandarin listeners frequently confused certain tones due to shared f0 characteristics, while English listeners made confusions based on their interpretations of intonational patterns. Training improved their ability to distinguish between tones, but the context (lexical versus intonational) influenced perceptual learning outcomes.

Conclusion

The study's findings illustrate that the perception and acquisition of lexical tones are intricately shaped by native language experience. The patterns of tone identification among both groups indicate that while both sets of listeners improved their performance through training, the underlying mechanisms driving their learning pathways vastly differ based on their respective linguistic backgrounds. Understanding these processes can inform teaching methodologies for second language acquisition, providing insight into how tone perception might be enhanced in learners of tonal languages.