Xi Xiang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Xi Xiang has a total of 5 co-authors. Here are the top co-authors by number of shared publications:
- Bing Cheng
English Department & Shaanxi Key Laboratory of AI-Empowered Language and Culture Research, School of Foreign Studies, Xi’an Jiaotong University, China
3 shared publications - Xiaojuan Zhang
English Department & Shaanxi Key Laboratory of AI-Empowered Language and Culture Research, School of Foreign Studies, Xi’an Jiaotong University, China
3 shared publications - Yang Zhang
Department of Speech-Language-Hearing Sciences and Masonic Institute for the Developing Brain, University of Minnesota, Minneapolis
3 shared publications - Xiangrong Dai
English Department & Shaanxi Key Laboratory of AI-Empowered Language and Culture Research, School of Foreign Studies, Xi’an Jiaotong University, China
2 shared publications - Yu Zou
English Department & Shaanxi Key Laboratory of AI-Empowered Language and Culture Research, School of Foreign Studies, Xi’an Jiaotong University, China
1 shared publication
Latest Publications
- Open Access
- 27 October 2025
- Published by MDPI AG in Preprints.org
Modeling Individual Differences in Categorical Perception with a Bayesian Framework
Listeners vary in their perception of speech, falling along a continuum from categorical to continuous. We applied a Bayesian computational framework to model this individual difference in speech perception. We analyzed publicly available data (Honda et al., 2024) from 195 participants across four phonetic conditions using both two-alternative forced choice and visual analogue scale tasks. Our model characterizes each listener’s perception using two key parameters: perceptual warping (τ), the signal-to-noise ratio of phonetic encoding, and noise variance (\( \sigma_S^2 \)), a proxy to perceptual noise in experimental designs. Combining these two parameters revealed four perceptual profiles: Veridical (high τ, low \( \sigma_S^2 \)), Categorical (low τ, low \( \sigma_S^2 \)), Compensatory (low τ, high \( \sigma_S^2 \)), and Noisy (high τ, high \( \sigma_S^2 \)). These profiles predicted behavioral patterns coherently, while successfully distinguishing between listeners who would appear similar when characterized by behavioral measures alone. Critically, results revealed that profile distributions shifted dramatically based on phonetic conditions, with primary cues yielding a balanced mix of profiles and secondary cues producing distributions skewed heavily toward Veridical and Compensatory listeners (80%). Underscoring this flexibility, intraclass correlations for both τ and \( \sigma_S^2 \) were zero, with phonetic condition effects 30 times stronger for \( \sigma_S^2 \) (χ² = 803.91) than τ (χ² = 29.47). These findings challenge the traditional view of categorical perception as a fixed characteristic, demonstrating instead that it is a flexible, context-driven perceptual state.- 27 October 2025
- Published by MDPI AG in Preprints.org
Open AccessModeling Individual Differences in Categorical Perception with a Bayesian Framework
Listeners vary in their perception of speech, falling along a continuum from categorical to continuous. We applied a Bayesian computational framework to model this individual difference in speech perception. We analyzed publicly available data (Honda et al., 2024) from 195 participants across four phonetic conditions using both two-alternative forced choice and visual analogue scale tasks. Our model characterizes each listener’s perception using two key parameters: perceptual warping (τ), the signal-to-noise ratio of phonetic encoding, and noise variance (\( \sigma_S^2 \)), a proxy to perceptual noise in experimental designs. Combining these two parameters revealed four perceptual profiles: Veridical (high τ, low \( \sigma_S^2 \)), Categorical (low τ, low \( \sigma_S^2 \)), Compensatory (low τ, high \( \sigma_S^2 \)), and Noisy (high τ, high \( \sigma_S^2 \)). These profiles predicted behavioral patterns coherently, while successfully distinguishing between listeners who would appear similar when characterized by behavioral measures alone. Critically, results revealed that profile distributions shifted dramatically based on phonetic conditions, with primary cues yielding a balanced mix of profiles and secondary cues producing distributions skewed heavily toward Veridical and Compensatory listeners (80%). Underscoring this flexibility, intraclass correlations for both τ and \( \sigma_S^2 \) were zero, with phonetic condition effects 30 times stronger for \( \sigma_S^2 \) (χ² = 803.91) than τ (χ² = 29.47). These findings challenge the traditional view of categorical perception as a fixed characteristic, demonstrating instead that it is a flexible, context-driven perceptual state. - Open Access
- 24 October 2025
- Published by MDPI AG in Preprints.org
Root Mean Square Error as a Robust Index of Gradient Speech Perception
This study introduces the root mean square error (RMSE) as a new metric for quantifying gradient speech perception in visual analog scale (VAS) tasks. By measuring the deviation of individual responses from an ideal linear mapping between stimulus and percept, RMSE offers a theoretically transparent alternative to traditional metrics like slope, response consistency, and the quadratic coefficient. To validate these metrics, we first used simulated data representing five distinct perceptual response profiles: ideal gradient, categorical, random, midpoint-biased, and conservative. The results revealed that only RMSE correctly tracked the degree of true gradiency, increasing monotonically from the ideal gradient profile (RMSE = 5.48) to random responding (RMSE = 42.16). In contrast, traditional metrics failed critically; for example, slope misclassified non-gradient, midpoint-biased responding as highly gradient (slope = 0.24). When applied to published empirical VAS data, RMSE demonstrated strong convergent validity, correlating robustly with response consistency (r ranging from -0.44 to -0.89) while avoiding the ambiguities of other measures. Crucially, RMSE exhibited moderate-to-high cross-continuum stability (mean r = 0.51), indicating it captures a stable, trait-like perceptual style. By providing a more robust and interpretable measure, RMSE offers a clearer lens for investigating the continuous nature of phonetic categorization and individual differences in speech perception.
- 24 October 2025
- Published by MDPI AG in Preprints.org
Open AccessRoot Mean Square Error as a Robust Index of Gradient Speech Perception
This study introduces the root mean square error (RMSE) as a new metric for quantifying gradient speech perception in visual analog scale (VAS) tasks. By measuring the deviation of individual responses from an ideal linear mapping between stimulus and percept, RMSE offers a theoretically transparent alternative to traditional metrics like slope, response consistency, and the quadratic coefficient. To validate these metrics, we first used simulated data representing five distinct perceptual response profiles: ideal gradient, categorical, random, midpoint-biased, and conservative. The results revealed that only RMSE correctly tracked the degree of true gradiency, increasing monotonically from the ideal gradient profile (RMSE = 5.48) to random responding (RMSE = 42.16). In contrast, traditional metrics failed critically; for example, slope misclassified non-gradient, midpoint-biased responding as highly gradient (slope = 0.24). When applied to published empirical VAS data, RMSE demonstrated strong convergent validity, correlating robustly with response consistency (r ranging from -0.44 to -0.89) while avoiding the ambiguities of other measures. Crucially, RMSE exhibited moderate-to-high cross-continuum stability (mean r = 0.51), indicating it captures a stable, trait-like perceptual style. By providing a more robust and interpretable measure, RMSE offers a clearer lens for investigating the continuous nature of phonetic categorization and individual differences in speech perception.
- Open Access
- 23 October 2025
- Published by MDPI AG in Preprints.org
Beyond Categorical Perception: Gradient Lexical Tone Processing Revealed by Visual Analog Scale
Purpose: While the Visual Analog Scale (VAS) has revealed gradient perception in segmental speech sounds, its application to lexical tones, a critical yet understudied suprasegmental feature, has been absent. This study investigated lexical tone categorization using VAS, directly comparing it with traditional two-alternative forced-choice (2AFC). Method: Eighty-four native speakers categorized an 11-step F0 continuum from Mandarin Tone 1 to Tone 2 in both tasks. Four-parameter logistic functions yielded slope (categorization sharpness) and response variability. Within-category sensitivity (Δ) was quantified from VAS responses.Results: Paired Wilcoxon signed-rank tests showed significantly shallower slopes (p < .001, r = .76) and lower variability (p < .001, r = .87) in VAS versus 2AFC. One-sample t-tests confirmed listeners discriminated fine-grained differences within categories, with Δ reliably exceeding zero (left: M = 0.0335, t(270) = 8.89, p < .001; right: M = 0.0256, t(316) = 8.38, p < .001). Crucially, slope and response variability were weakly correlated in VAS (ρ = .27, p < .05) but strongly negatively correlated in 2AFC (ρ = -.67, p < .001). Moreover, response variability correlated significantly across tasks (ρ = .40, p < .001), while slopes did not. Conclusion: These findings provide the direct evidence for gradient perception at the suprasegmental level, further establishing VAS as a sensitive tool for uncovering the nature of speech categorization. The dissociation between task-dependent gradiency and stable response variability helps reconcile apparent conflicts in the categorical perception literature, suggesting that these conflicts may stem from methodological constraints rather than genuine theoretical disagreements.
- 23 October 2025
- Published by MDPI AG in Preprints.org
Open AccessBeyond Categorical Perception: Gradient Lexical Tone Processing Revealed by Visual Analog Scale
Purpose: While the Visual Analog Scale (VAS) has revealed gradient perception in segmental speech sounds, its application to lexical tones, a critical yet understudied suprasegmental feature, has been absent. This study investigated lexical tone categorization using VAS, directly comparing it with traditional two-alternative forced-choice (2AFC). Method: Eighty-four native speakers categorized an 11-step F0 continuum from Mandarin Tone 1 to Tone 2 in both tasks. Four-parameter logistic functions yielded slope (categorization sharpness) and response variability. Within-category sensitivity (Δ) was quantified from VAS responses.Results: Paired Wilcoxon signed-rank tests showed significantly shallower slopes (p < .001, r = .76) and lower variability (p < .001, r = .87) in VAS versus 2AFC. One-sample t-tests confirmed listeners discriminated fine-grained differences within categories, with Δ reliably exceeding zero (left: M = 0.0335, t(270) = 8.89, p < .001; right: M = 0.0256, t(316) = 8.38, p < .001). Crucially, slope and response variability were weakly correlated in VAS (ρ = .27, p < .05) but strongly negatively correlated in 2AFC (ρ = -.67, p < .001). Moreover, response variability correlated significantly across tasks (ρ = .40, p < .001), while slopes did not. Conclusion: These findings provide the direct evidence for gradient perception at the suprasegmental level, further establishing VAS as a sensitive tool for uncovering the nature of speech categorization. The dissociation between task-dependent gradiency and stable response variability helps reconcile apparent conflicts in the categorical perception literature, suggesting that these conflicts may stem from methodological constraints rather than genuine theoretical disagreements.