Language change is a complex phenomenon influenced by both internal and external factors. While social factors like age and class have been extensively studied, the role of cognitive functions remains less understood. This study focuses on the merging of tone pairs in Macau Cantonese, a dialect exhibiting varying rates of tone merging. Cantonese, with its six lexical tones and three allotones, presents tone pairs (T2-T5, T3-T6, and T4-T6) that are undergoing or have completed merging processes. Prior research on Cantonese tone processing, primarily focusing on Guangzhou and Hong Kong, has yielded mixed results regarding the impact of cognitive factors, such as attention, memory, and executive function. Some studies have shown correlations between cognitive abilities and tone processing, but methodological limitations, such as inconsistent participant selection criteria and lack of focus on auditory modality, need to be addressed. This study aims to investigate the relationship between tone merging in Macau Cantonese and cognitive functions (attention, working memory, and executive function) in a younger generation of speakers, using auditory-only tasks to avoid confounds from visual processing. It examines the current merging status of the tone pairs, exploring whether the influence of cognitive functions is consistent across different merging stages and modalities (perception and production). The study will contribute to a broader understanding of how cognitive functions impact speech processing and provide insights into the origins of sound changes, and clarify the often-inconsistent relationship between speech perception and production in tone processing.

Previous research on language variation and change has largely focused on social factors, while the impact of cognitive functions is less understood. Studies on Cantonese tone processing have shown mixed results. Law et al. (2013) found that the ability to distinguish tones correlates with brain activity related to attentional switching and working memory. Ou et al. (2015) showed that reaction time and accuracy in speech perception and production were influenced by attentional switching and working memory capacity. However, methodological inconsistencies in these studies, such as varying participant selection criteria across tone pairs and failure to control for visual processing, limit the interpretability of the results. Ou and Law (2017) explored neural mechanisms, suggesting that attentional switching influences acoustic cue representation, thereby affecting speech perception and production, but their focus on only one tone pair restricts the generalizability of their findings. The Motor Theory and the DIVA model offer competing perspectives on the interaction between speech perception and production, with the former emphasizing a strong link and the latter suggesting that perception calibrates production. However, the inconsistent relationship between perception and production observed in previous studies on Cantonese necessitates further investigation.

This study recruited 44 native Macau Cantonese speakers aged 17-28. Participants underwent a production task involving reading aloud 24 words representing the six lexical tones using four CV roots. A perception task used a AX discrimination paradigm, presenting 32 tone pairs (including the three merging pairs) and 16 fillers, with participants judging same/different. Cognitive functions were assessed using the Test of Everyday Attention (TEA) for attention, the Digit Span Test for working memory (including forward, backward, and sequence subtests), and an auditory Stroop test for executive function. All cognitive tasks were auditory-only. Praat software was used to extract fundamental frequency (F0) data from the production task recordings, normalized using a semitone transformation. Growth curve analysis (GCA) was employed to analyze the F0 data at both group and individual levels, determining whether tone merging had occurred. For the perception task, discrimination rate and reaction time were analyzed. Statistical analyses included independent sample t-tests and Mann-Whitney U tests (where the assumption of normal distribution wasn't met), along with Pearson correlation analysis to investigate the relationship between cognitive function performance and tone processing.

At the group level, GCA revealed that T2-T5 tones were fully merged, while T3-T6 and T4-T6 were not. Individual-level analysis showed that all participants merged T2-T5, 22 merged T3-T6 in production and 18 in perception, and only 8 merged T4-T6 in production and 15 in perception. Independent sample t-tests confirmed these classifications. For T3-T6, the unmerged groups showed significantly higher discrimination rates and faster reaction times in perception, and a larger average F0 difference in production compared to the merged groups. Similar results were observed for T4-T6. Interestingly, the relationship between perception and production differed across tone pairs. For T3-T6, individuals who hadn't merged perceptually showed a significantly larger average F0 difference in production than those who had merged. No such difference was found for T4-T6. There were no significant differences in perception measures between merged and unmerged groups for production. For T3-T6, individuals who hadn't merged perceptually demonstrated significantly higher attention and working memory scores. For T4-T6, only attention scores were significantly higher in the unmerged group. No significant differences in executive function were found for either T3-T6 or T4-T6. Correlation analysis showed that for T3-T6 and T4-T6, discrimination rate positively correlated with attention and working memory, and reaction time negatively correlated with working memory. There were no significant correlations with executive function. For T2-T5, no correlations between cognitive function and tone processing were found.

This study's findings demonstrate that cognitive functions, particularly attention and working memory, play distinct roles in tone merging depending on the merging stage and the inherent characteristics of the tone pairs. The lack of correlation for T2-T5, already fully merged, suggests that cognitive demands related to these tones are no longer present. The correlations found for T3-T6 and T4-T6, at different stages of merging, suggest a relationship between cognitive effort and the stability of tone perception and production. The stronger correlations for T3-T6 in both perception and production, compared to T4-T6's correlations only in perception, indicate that cognitive resources are utilized more extensively when the merging is less complete. The inconsistent relationship between perception and production in the tone pairs is discussed in relation to individual cognitive abilities, suggesting a dynamic interplay between cognitive resources and tone representations over time. The results support the DIVA model, where accurate perception is essential for accurate production, and that cognitive functions may operate as a 'sliding window' during language change.

This study offers novel insights into the role of cognitive functions in tone merging. The varying degrees of correlation between cognitive abilities and tone processing across different merging stages highlight the dynamic nature of the relationship between cognitive resources and language change. Future research should focus on larger and more balanced samples, exploring neural mechanisms and the interplay between perception and production in more detail to refine our understanding of the cognitive underpinnings of language change.

The study's relatively small sample size, and the imbalance between merged and unmerged participants in some tone pairs, may limit the generalizability of the results. The limitations of the auditory Stroop test adapted for Cantonese might have affected the analysis of executive function. Further research using more sophisticated executive function tasks is recommended.