5.2.1 Categorical assimilation of Cantonese tones
Given the different weighting of distinctive features that Cantonese and Mandarin assign to their tonal inventories (Gandour and Harshman1978, Gandour1983), it is predictable that native speakers of Mandarin may put more emphasis on tone contours when it comes to tone assimilation. In other words, Mandarin subjects are expected to be more sensitive to the contour feature than f0 height, and similarities or discrepancies in contour direction will have greater impact over f0 height in determining whether two tones can be assimilated as belonging to the same category.
Hence, Mandarin speakers are predicted to assimilate non-native tones to native tones primarily on the basis of their contour proximity to one another. In almost all literature, the High Level and High Rising tones are reported as the most easily recognized tones in Cantonese and can almost certainly be assimilated to the High Level tone and the Rising tone in Mandarin. In other words, they are said to be
“assimilated to a native category”.
Tone 1 (HL), Tone 3 (ML), and Tone 6 (LL). Within the High Level category,
Cantonese high-level, mid-level and low-level tones can be further differentiated with regard to their category goodness. Based on findings from the Labeling task, Tone 1 and Tone 3 are of similar nature to Mandarin-speaking subjects as level tones. A different preference of tonal labels can be found with regard to Tone 6. While subjects of the Putonghua group label Tone 6 as the other two level tones, Taiwan Mandarin subjects equally label the low level tone with Mandarin Tone 1 and Mandarin Tone 4, indicating their relatively higher sensitivity to f0 contrast and the existence of the extra tonal category.
Tone 2 (HR) and Tone 5 (LR). As for the Rising category, results from the Labeling
task show that Tone 2 and Tone 5 are similarly categorized as rising tones. However, since the rising slope of the High Rising tone is steeper than that of the Low Rising tone, High Rising is expected to be more prominent than Low Rising in category goodness.
Tone 4 (LF). A situation arises with regard to Cantonese Low Falling tone when
speculations in cross-language speech perception are made on the basis of theoretical models and phonological descriptions. Following the argument thus far, LF should be assimilated to the dipping Tone 3 in Mandarin on account of its similarity to the falling allotone of the Dipping tone. However, in some phonological description of Cantonese tones, LF is regarded as a level tone and Cantonese a language with four level tones. If such description is adopted in this discussion, then LF may very well be assigned to the native category where other level tones are assimilated. Labeling results show that Tone 4 are divisive among subjects of both groups. For Taiwan Mandarin subjects, Tone 4 can be clearly mapped into Mandarin Tone 3, indicating a dipping tone category for Taiwan Mandarin subjects other than the High Level and Rising categories of tones. However, for Putonghua subjects, it is difficult to label Tone 4 with any of the Mandarin tones. It seems that, because of the prominence of the rising component in Putonghua Mandarin Tone 3, Putonghua subjects do not share the inclination of their Taiwan Mandarin counterparts to label the low falling tone as Mandarin Tone 3. Therefore, depending which variety of Mandarin the subject speaks, LF may be assimilated either to a category of its own (the Mandarin Dipping tone) or to none of the particular categories in Mandarin.
The assimilation patterns of Cantonese tones to Mandarin ones can be summarized as follows:
Assimilated to High Level category: High Level tone and Mid-Level tone (good
exemplars), Low Level tones (acceptable exemplar);
Assimilated to Rising category: High Rising tone (good exemplar), Low Rising
tone (acceptable);
Assimilated to Dipping category: Low Falling tone (good exemplar for Taiwan
Mandarin subjects only)
Uncategorized speech: Low Falling tone (for Putonghua subjects only)
It should be noted that Mandarin Tone 4 may not involve in the non-native tonal assimilation under discussion here. Hence, one may propose that Tone 1 and Tone 3 are both categorized to Mandarin Tone 1, Tone 2 and Tone 5 to Mandarin Tone 2, and Tone 4 to Mandarin Tone 3 exclusively to Taiwan Mandarin subjects.
5.2.2 Non-native tonal contrasts on the basis of tonal assimilation
According to PAM, there will be great difficulty distinguishing tones within the same category with similar exemplar goodness and great ease distinguishing tones that belong to different categories. Based on the assimilation patterns of non-native tones to native tones that are inferred from results of the Labeling task, the assimilation patterns and discriminability for non-native contrasts can be speculated and examined:
Two-category assimilation (TC): Since Cantonese level tones and rising tones
are assimilated to different native categories (Mandarin High Level and Mandarin Rising), discrimination between the two sets of non-native tones is very easy and subjects are not likely to confuse the two sets with each other.
The effect is especially obvious for rising tones, which are seldom confused with level tones. Another example is Tone 4 (LF) for Taiwan Mandarin subjects. Since the low falling tone is assimilated to an individual Mandarin
Dipping category, Taiwan Mandarin subjects have a high accuracy in the Identification task in comparison with Putonghua subjects.
Category-goodness difference (CG): Within the High Level category, the
contrast between Cantonese High Level and Low Level tones is more prominent than that between Cantonese High Level and Mid-Level tones.Therefore, subjects are more likely to confuse High Level with Mid-Level than with Low Level tones.
Single-category assimilation (SC): Since Cantonese High Level and Mid-Level
tones are both good exemplars of the High Level category, it may be challenging for native speakers of Mandarin to contrast Cantonese High Level with Mid-Level tones. Similarly, the distinguishing Cantonese High Rising and Low Rising tones proved to be equally difficult, especially for Putonghua subjects.Uncategorized versus categorized (UC): For Putonghua subjects, the distinction
of between Low Falling and other tones falls into this class, since the tone is not assimilated to any particular tonal category in Putonghua.
Results from the Identification task in part support predictions made on the basis of PAM. Taiwan Mandarin subjects do show a much better performance in the identification of Tone 4 than Putonghua subjects. Also, subjects of both groups have much difficulty identifying Tone 6, a tone that both groups find hard to give a definite label in the Labeling task. However, a closer look reveals that PAM may not account for all the results at hand and there are may be other factors coming into play.
For Single-category assimilation, distinguishing Tone 1 from Tone 3 and Tone 2 from Tone 5 seems challenging to Putonghua subjects, which, along with the findings above, conform to the prediction based on PAM. Since the Low Rising tone is often
confounded with the High Rising tone in literature (Francis et al. 2008), and the two tones are even reported to have a merger by native speakers in Hong Kong and Macau area (Fung & Wong 2010), it is predictable that contrasts between the two rising tones could be very challenging. However, this expected difficulty is not as found in Taiwan Mandarin subjects, who are also native speakers of a variety of Mandarin. Taiwan Mandarin subjects in general choose tones with lower tonal heights for Tone 3 (ML) and Tone 5 (LR), indicating sensitivity to f0 heights in tone perception. In fact, even if the identification of Tone 6 (LL) is in turmoil for subjects of both groups, based on confusions made by subjects of both groups, one can still find that Taiwan Mandarin subjects tend to choose answers with lower tonal heights while answers from Putonghua subjects distribute evenly among most of tones.
Therefore, for non-native tone perception, category assimilation may not be the only factor in effect. Similarly, the contrast between Low Falling and Low Level tones has shown to be challenging to Putonghua subjects even though the contrast is
uncategorized versus categorized (UC) and should be of relative ease. Since the two
tones are reported to have near-merger phenomenon recently in Hong Kong Cantonese (Fung et al. 2012), and on account of Mandarin speakers’ relative insensitivity to f0 contrasts, it is foreseeable that the two tones should be confounded with Low Rising, another tone with similar f0 onset height.
5.2.3 Tonal categories in production
Performance in the Vocalization task, however, does not follow the results of Identification at first sight. Subjects of both groups show good performance for Tone 1, Tone 2, and Tone 4 and poor performance for Tone 3, Tone 5, and Tone 6. In particular, Tone 3 seems to be the most challenging for both groups to pronounce and
Entering Tone 3 has the lowest accuracy among all entering tones. The result is at odds with findings from the Identification task, in which it is Tone 6 that poses greatest difficulty to subjects of both groups.
A closer look into tonal confusions in the Production phase shows that Putonghua subjects tend to pronounce Tone 3 lower as Tone 4, while Taiwan Mandarin subjects tend to pronounce it higher as Tone 1 or correctly as Tone 3. The relative sensitivity to f0 height seems to remain with Taiwan Mandarin subjects since it turns out that Putonghua subjects pronounce all tones with lower f0 heights, that is, Tone 4, Tone 5, and Tone 6, to a considerable extent as Tone 4.
It may be inferred that when it comes to imitation and production of non-native tones, subjects resort to their native tonal categories for pronunciation. For Mandarin-speaking subjects, it means one of the four tonal contours: high level, high rising, dipping (along with its allotone low falling), and high falling. Subjects may or may not have trouble perceiving height differences between High Level, Mid-Level, and Low Level tones and between High Rising and Low Rising tones acoustically, yet when it comes to vocalize a foreign tone, they resort to their native tonal system and produce all Cantonese tones as the similar level and rising tones in Mandarin.
Differences within the same categories that can be detected perceptually in the Identification task dissolve when subjects turn their attention to pronunciation. In the end, subjects attribute the non-native tone as either one of the tones within the repertoire of tones and allotones available in their native language, and pronounce it as the native tone anyway.