---
title: Perception and production of tonal coarticulation in Taiwan Mandarin and Taiwan Sothern Min
---

# Production
### Results
#### Anticipatory
##### Contours

###### Mandarin

###### Min

##### Comparison

Linear mixed model fit by REML. t-tests use Satterthwaite's method ['lmerModLmerTest']
Formula: y ~ x * lang * tone + (x | participant)
   Data: temp

REML criterion at convergence: 7137.1

Scaled residuals: 
    Min      1Q  Median      3Q     Max 
-7.2850 -0.5839 -0.0078  0.6154  5.9733 

Random effects:
 Groups      Name        Variance Std.Dev. Corr 
 participant (Intercept) 0.019116 0.13826       
             x           0.005751 0.07584  -0.56
 Residual                0.392081 0.62616       
Number of obs: 3688, groups:  participant, 22

Fixed effects:
                   Estimate Std. Error         df t value Pr(>|t|)    
(Intercept)      -9.423e-01  5.433e-02  5.375e+01 -17.343  < 2e-16 ***
x                 2.006e-01  5.827e-02  3.385e+02   3.442 0.000650 ***
langMin           2.306e-01  7.337e-02  3.705e+01   3.143 0.003282 ** 
tone33            6.872e-01  3.833e-02  3.646e+03  17.928  < 2e-16 ***
tone35            9.483e-01  4.584e-02  3.646e+03  20.688  < 2e-16 ***
tone51            4.359e-01  5.055e-02  3.642e+03   8.623  < 2e-16 ***
tone55            1.616e+00  4.827e-02  3.658e+03  33.470  < 2e-16 ***
x:langMin         1.029e-01  6.887e-02  1.445e+02   1.494 0.137337    
x:tone33          7.847e-03  5.787e-02  3.656e+03   0.136 0.892142    
x:tone35         -2.188e-01  6.042e-02  3.228e+03  -3.622 0.000297 ***
x:tone51         -1.543e-01  6.524e-02  3.554e+03  -2.365 0.018088 *  
x:tone55         -1.705e-01  6.381e-02  3.607e+03  -2.672 0.007583 ** 
langMin:tone51    2.534e-01  6.424e-02  3.648e+03   3.945 8.13e-05 ***
langMin:tone55    1.700e-01  6.184e-02  3.657e+03   2.750 0.005998 ** 
x:langMin:tone51  2.372e-01  8.160e-02  3.603e+03   2.906 0.003679 ** 
x:langMin:tone55 -1.669e-01  7.733e-02  3.631e+03  -2.158 0.031018 *  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

#### Carryover
##### Contours

###### Mandarin

###### Min

##### Comparison

Linear mixed model fit by REML. t-tests use Satterthwaite's method ['lmerModLmerTest']
Formula: y ~ x * lang * tone + (x | participant)
   Data: temp

REML criterion at convergence: 7960.6

Scaled residuals: 
    Min      1Q  Median      3Q     Max 
-7.4711 -0.5844 -0.0087  0.6133  6.1552 

Random effects:
 Groups      Name        Variance Std.Dev. Corr 
 participant (Intercept) 0.02020  0.14214       
             x           0.00489  0.06993  -0.58
 Residual                0.37383  0.61142       
Number of obs: 4220, groups:  participant, 25

Fixed effects:
                   Estimate Std. Error         df t value Pr(>|t|)    
(Intercept)      -1.030e+00  5.035e-02  5.649e+01 -20.464  < 2e-16 ***
x                 2.098e-01  4.894e-02  3.338e+02   4.286 2.38e-05 ***
langMin           3.215e-01  6.938e-02  3.953e+01   4.634 3.86e-05 ***
tone33            6.733e-01  3.541e-02  4.172e+03  19.012  < 2e-16 ***
tone35            1.027e+00  4.138e-02  4.133e+03  24.821  < 2e-16 ***
tone51            5.165e-01  4.537e-02  4.168e+03  11.384  < 2e-16 ***
tone55            1.703e+00  4.360e-02  4.187e+03  39.065  < 2e-16 ***
x:langMin         9.065e-02  5.922e-02  1.479e+02   1.531 0.127977    
x:tone33         -5.203e-04  5.420e-02  4.184e+03  -0.010 0.992342    
x:tone35         -2.192e-01  5.118e-02  3.726e+03  -4.284 1.88e-05 ***
x:tone51         -1.911e-01  5.611e-02  4.033e+03  -3.406 0.000666 ***
x:tone55         -1.862e-01  5.507e-02  4.099e+03  -3.382 0.000727 ***
langMin:tone51    1.642e-01  5.848e-02  4.177e+03   2.808 0.005013 ** 
langMin:tone55    6.184e-02  5.627e-02  4.186e+03   1.099 0.271857    
x:langMin:tone51  2.792e-01  7.231e-02  4.116e+03   3.860 0.000115 ***
x:langMin:tone55 -1.420e-01  6.730e-02  4.051e+03  -2.110 0.034891 *  
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

### Tonal spaces

[Mandarin tonal space](https://drive.google.com/file/d/16XFr_lLYDge8LIDWPvE2pr4g7HoVlBJY/view?usp=sharing)

[Min tonal space](https://drive.google.com/file/d/1yPcpadkBYqZnY88KK1hM7JISx_7FXeII/view?usp=sharing)

### Summary
* Both carryover and anticipatory effects are observed in both languages.
* Both carryover and anticipatory effects are assimilatory in both languages.
* The magnitudes of both effects are not significantly different.

# Perception
### Calculation (example)
#### Raw data

#### Logistic regression
- Two distances are measured:
- The .25 threshold distance
- The .75 threshold distance

### Results
### Overall

#### Mandarin (monolingual)

#### Mandarin (bilingual)

### Min

#### Monolingual group's Mandarin vs. bilingual group's Min

p=.0002***

#### Bilingual group paired comparison

p=.0441*

#### Summary
- This result is in accordance with our prediction: Southern Min speakers are less subject to perceptual compensation for tonal coarticulation.
- This discrepency is present even within the bilingual speakers. This may lead to discussion of bilingual cognition.

# Tonal sensitivity

### Calculation (example)

### Results

#### Raw
##### Mandarin (monolingual)

##### Mandarin (bilingual)

##### Min

#### Falling tone (21)

p=.0013**

#### Low level tone (51)

p=.0132*

### Summary
For the low tone, Mandarin has stricter tone boundaries than Min. For the falling tone, it is Min that has stricter tone boundaries.
The reason why Min has fuzzier boundaries for the low tone might be because of the fact that Min low tone becomes a sandhi falling tone preceded by other syllables. The subjects might therefore be more tolerant for the low-tone-to-falling-tone alternation.

### Discussion