Chapter 5 Discussion
5.1 Re-examination of Mandarin Tonal coarticulation
In terms of the both directions in Mandarin tonal coarticulation, Shen (1990) proposed that both carryover and anticipatory effects are equal, while Xu (1993, 1997) argued that carryover effect is more dominant than anticipatory effect. According to the results shown in the previous chapter, for all target tones, the main effect of Preceding tone is significant across all measured time points, which indicates that the carryover effect from preceding tones can extend to the whole syllable of all target tones. However, the main effect of Following tone is significant with more limited extent. Only Tone 2 and Tone 1 got influenced significantly across more than half of the syllable. That is, the extent of the anticipatory effect from following tones is smaller. Besides, the magnitude of the delta F0
variation caused by two directions is also different. By calculating the range between the maximum and the minimum value of delta F0, the variation caused by the preceding tone is at least 28 Hz while the variation by the following tone is no more than 20 Hz across four target tones. This implies that the range of delta F0 varied by carryover effect is greater than by anticipatory effect. To conclude from both aspects of temporal extent and delta F0
magnitude, the data in this study showed that carryover effect is much more prevailing than anticipatory effect in tonal coarticulation. This echoes the statement of Xu (1993, 1997) but not Shen’s (1990). The conflicting finding from Shen’s (1990) is probably resulted from the inadequacy of her experiment. As she measured only onset, offset and the turning point (if there is any turning point), the tonal variation across the whole syllable duration is unclear.
Therefore, it would be too rough to define that any difference of the target onset or offset is
59
affected exactly by carryover effect or by anticipatory effect. Moreover, Shen’s data was limited that it comes from only two speakers. If there was any speaker difference in recording, the conclusion will be more different from the way most people produced.
Tones as targets of tonal coarticulation were not consistent in previous studies. Lin &
Yan (1992) proposed that Tone 1, Tone 2 and Tone 3 have greater F0 variation under carryover effect while Tone 4 is more susceptible to anticipatory effect. Xu (1994) argued that Tone 1 and Tone 2 get more carryover effect because of the larger affected extent and Tone 1 gets most anticipatory effect. Xu (1997) also provided another statement that Tone 2 and Tone 4 are targets of anticipatory effect because their F0 get larger variation from following tones. The problem is that not only different tones were defined as targets, but the criteria to define also diverse. In this study, tones being affected by coarticulation are
examined through two aspects: the extent and the magnitude. Under carryover effect, four tones share the same affected extent as the effect extends to 10 measured time points on them equally. But four tones have different magnitude of F0 variation. By comparing the range between their maximum and minimum delta F0 in magnitude, the order of range is like Tone 4 > Tone 1 > Tone 3 > Tone 2 (The range of Tone 4 is about 62 Hz, of Tone 1 is about 42 Hz, of Tone 3 is about 33 Hz and of Tone 2 is about 29 Hz). This order shows that Tone 4 gets the largest F0 variation by carryover effect. In terms of anticipatory effect, four tones have different results on extent and magnitude. The extent of anticipatory effect is under the order that Tone 2 > Tone 1 > Tone 3 > Tone 4 (9 time points of Tone 2 are affected, 6 time points of Tone1, 4 time points of Tone 3 and 3 time points of Tone 4). And the
magnitude of anticipatory effect is under the order that Tone 2 > Tone 3 > Tone 4 > Tone 1(The range of Tone 2 is about 20 Hz, of Tone 3 is about 17 Hz, of Tone 4 is about 13 Hz and of Tone 1 is about 8 Hz). Therefore, Tone 2 gets varied the most both from the temporal
60
extent and the magnitude of anticipatory effect. In conclusion, the temporal extent of
carryover effect is equal across four tones, but the magnitude is largest on Tone 4. And Tone 2 is the most obvious target of anticipatory effect from both the temporal extent and the magnitude.
Tones as triggers of tonal coarticulation were discussed only by Shen (1990) and Xu (1997). Shen (1990) proposed that Tone 1 and Tone 2 trigger most carryover effects as the F0 of their following tones has the largest variation. Tone 4 and Tone 1 are anticipatory triggers as the F0 of their preceding tones varied the most. From Xu’s (1997) results he concluded that Tone 3 is the most significant trigger for anticipatory effect since it causes more variation on its preceding tones. As the tone which causes the largest F0 variation on its adjacent tones is taken as the trigger of tonal coarticulation, based on the results of post-hoc comparison in previous chapter, trigger tones at each time point for four target tones are listed in Table 28 and Table 29.
time point target tone
TP1 TP2 TP3 TP4 TP5 TP6 TP7 TP8 TP9 TP10
Tone 1 T3 T3 T1 T1 T1 T1 T1 T1 T1 T1
Tone 2 T3 T2 T2 T2 T2 T2 T2 T3 T3 T3
Tone 3 T4 T4 T4 T2 T2 T2 T2 T4 T4 T4
Tone 4 T4 T4 T4 T2 T2 T2 T2 T2 T2 T1
Table 28. The trigger tones of carryover effect
61 time point
target tone
TP1 TP2 TP3 TP4 TP5 TP6 TP7 TP8 TP9 TP10
Tone 1 n.s. n.s. n.s. n.s. T3 T3 T3 T2 T2 T2
Tone 2 T4 n.s. T3 T3 T3 T3 T3 T1 T1 T1
Tone 3 n.s. n.s. n.s. n.s. n.s. n.s. T3 T4 T4 T4 Tone 4 n.s. n.s. n.s. n.s. T3 T3 T3 n.s. n.s. n.s.
Table 29. The trigger tones of anticipatory effect
( The ‘n.s.’ refers that the anticipatory effect is not significant at that time point)
However, some problems will be found when these trigger tones are examined carefully. First, no regular pattern exists. Not only that four target tones are affected by different triggers, but even a certain target tone is affected by different triggers across ten time points. Also, if we re-inspect the data of post-hoc comparison, it will be found out that at some time points the F0 variation caused by the ‘second’ trigger is very similar to the ‘top’
trigger, which makes the trigger in tonal coarticulation become more complicated.
Therefore, it is problematic to define triggers in tonal coarticulation simply as the tones which caused largest variation on other tones. One possible way to resolve this is to narrow down the trigger to a smaller unit, which might provide more detailed information of tonal coarticulation. As Xu (1994) proposed that tonal coarticulation happens due to the tonal values adjacent to the target tones instead of the whole neighboring tones, another way is adopted to find possible triggers. That is, to exemplify those tonal values immediately next to the target tones.
For each target tone being affected by carryover effect, the results of F0 value varied the most at each time point are examined to define which preceding offset tends to be the
62
trigger. If the variation caused by the ‘second’ trigger is very similar to the ‘top’ trigger, both cases will be included. All of the results are presented in Table 30. To sum up, except for the final portion (TP8 to TP10) where the carryover effect is not the same on different target tones perhaps due to its far distance from the preceding trigger offset, the carryover effect is quite consistent on four target tones from TP1 to TP7. All target tones will firstly get lowered the most due to the preceding low offsets, and then from TP2, TP3 or TP4 will start to get raised the most due to the preceding high and mid offsets. Such different effects at different time points probably result from the nature of Mandarin tones. That is, instead of being level tones, Mandarin tones are contour tones which do not contain steady F0
contours. Therefore, the change on these contour tones will not contain steady patterns, either. Besides, as the preceding low offsets lower the target tones and the preceding high and mid offsets raise them, assimilation is found in carryover coarticulation. In conclusion, under carryover assimilation, the preceding low offsets significantly lower the target tones for the time points in front, while the raising effect from preceding high and mid offsets will overwhelm it afterwards.
63
Table 30. The preceding trigger offsets of carryover effect
target Tone 1 TP10 backward to TP5
F0 get raised the most by following mid onset (T2 and T3)
target Tone 2 TP7 backward to TP3
F0 get raised the most by following mid onsets (T3 and T2)
TP10 backward to TP8
F0 get lowered the most by following high onsets (T1 and T4)
target Tone 3 at TP7
target Tone 4 TP7 backward to TP5
F0 get raised the most by following mid onsets (T3 and T2)
Table 31. The following trigger onsets of anticipatory effect
64
The pattern of anticipatory effect from the following tone is very different from that of carryover effect. Table 31 shows the results of following trigger onsets of anticipatory effect by presenting the F0 value varied the most at each time point where the anticipatory effect is significant. To sum up, in every case where the anticipatory effect is significant on target tones, it is the following mid or high onsets trigger the tonal coarticulation. Not like the situation of carryover effect in which all variation is assimilation, though following mid onsets also raise the F0 of targets tones just like its raising effect in carryover coarticulation, following high onsets right behind the target Tone 2 and Tone 3 (tones with mid or low offsets) lowers the F0 of targets for three time points (from TP10 backward to TP8) instead of raising it. (Although target Tone 4 also involves a low offset, the anticipatory effect on Tone 4 is not significant from TP10 backward to TP8.) Therefore, dissimilation exists in anticipatory coarticulation, and it happens to the preceding tones with mid or low offsets.
When the preceding tone carries a high offset (Tone 1), it still undergoes the anticipatory assimilation from its following tonal onset.
Anticipatory dissimilation was also found in Thai (Gandour et al., 1992c; Potisuk et al., 1996) and in Mandarin (Xu, 1994 & 1997). To investigate the phenomena of
anticipatory dissimilation, there should be two types included. For the first type, the
following lower F0 value raises the preceding higher F0 value rather than lowers it. And the second type is opposite that the following higher F0 value lowers the preceding lower F0
value rather than raises it, just as the results shown in this study. The anticipatory
dissimilation found in the studies of Gandour et al. (1992c), Potisuk et al. (1996), and Xu (1994 & 1997) belonged to the first type. Gandour et al. (1992c) discussed their results by explaining the adjustment of vocal folds. According to them, “The transition (of F0 range) requires complex adjustments of the vocal folds. Because of vocal fold dynamics, one may
65
speculate it is easier in some articulatory sense to move from an even higher F0 to an extremely low F0. This vocal fold adjustment is analogous to what happens when a
semi-trailer swings wide to make a sharp right or left turn. The extra wide turn facilitates the movement from a street going in one direction to a street cutting off at a 90 degree angle.
The anticipatory effects on the slope of the preceding falling tone are believed to follow as a consequence of the adjustments in height. From a given height to a fixed F0 onset, the slope must necessarily be steeper from a higher F0. Back to the semi-trailer analogy, the angle of the turn varies as a direct consequence of the wider swing around the corner.”
However, Xu (1994) argued that the theory of Gandour et al. was too wide that both low-to-high F0 transition and high-to-low F0 transition should be possible. That is,
according to Xu, “Not only should a high pitch target be raised by a following low pitch target, but also a low pitch target should be lowered by a following high pitch target.” Since the latter situation (high-to-low F0 transition) was not found in the results of Gandour et al.
(1992c) and of Xu (1994), Xu (1994) suggested that the explanation of Gandour et al. does not suit to the fact. Therefore, Xu (1994 & 1997) provided another way to describe the anticipatory dissimilation. According to him, the ‘High-Low’ tonal sequence is similar to the pattern of intonation declination in which the pitch also goes from high to low. In order not to confuse both tonal pattern and declination contour, the difference between the
‘High-Low’ sequence must be exaggerated. According to Xu (1994), “however, due to the physical limit of the lower threshold, this exaggeration is accomplished by fully
implementing the H target rather than by lowering L target.” Xu’s argument seems to make sense in some ways. However, even if it takes more efforts to reach a low pitch, it’s still possible that the lowering effect can exist with a minor magnitude that does not extend lower than one’s low threshold. The exaggeration should be possible to be completed by
66
varying the F0 of both targets. Moreover, since the second type of anticipatory dissimilation is found in this study, Xu’s argument that Gandour et al. (1992c) had unsuitable explanation is not tenable anymore. Thus, the ‘vocal fold adjustment’ explanation from Gandour et al.
(1992c) should be taken in again. That is, no matter the F0 transition is from high-to-low or from low-to-high, the F0 of the previous target will first raise or lower more to leave a greater range for easier transition to the next target F0.
To conclude the Exp. I, in this study it is verified that carryover effect is much more prevailing than anticipatory effect from both the temporal extent and the magnitude. Tone 4 is most susceptible target to carryover effect due to the affected magnitude (all four target tones share the same temporal extent of carryover effect), and Tone 2 is most susceptible target to anticipatory effect due to both the extent and the magnitude. In terms of the trigger in tonal coarticulation, instead of the certain tone, it is the adjacent tonal offset or onset that should be taken into consideration. Preceding low offset tends to trigger carryover
coarticulation on the beginning of its target tone, but after about two or three time points the effect from preceding high and mid offset will be more significant. Each kind of preceding offsets causes carryover assimilation. Anticipatory coarticulation is more complicated that it contains both assimilation and dissimilation. Similar to the raising effect in carryover coarticulation, following mid onset always raises its target tone, but following high onset tends to lower the mid and low offset of its preceding target tones. The anticipatory
dissimilation is possible to be explained by the ‘vocal fold adjustment’ theory of Gandour et al. (1994). Nevertheless, some subtle problems are involved in this study. Even though that the F0 region (High, Mid and Low) was divided well through detailed values averaged from ten speakers, that is, there should be no doubts on the division, the trigger onset and offset located in Mid region has High-region-liked performance that it always raises its adjacent
67
targets. If the Mid trigger does have the same influence as High trigger, but it does not cause anticipatory dissimilation just like High trigger did. The nature of the Mid trigger is still left without good explanations. Also, it could be strange that the anticipatory dissimilation found in this study is very different from the results of anther Mandarin study of Xu (1994
& 1997). Each of the study contains only one type of possible anticipatory dissimilation, while both types should be found theoretically. Whether there are other factors influencing the Mandarin anticipatory coarticulation which was not being considered should be
involved in the future related studies to solve this inconsistency.