Stress realizations by different speaker groups

In this study, it was found that stress exerted a salient effect in terms of sibilant

realizations. In particular, larger sibilant contrasts were made in the stressed condition.

Such a result was compatible with the effect of prosodic strengthening on segmental

realizations, depicting that linguistic contrasts tend to be maximized or maintained in

prosodically stronger conditions (e.g., Cho, 2005; Cho & McQueen, 2005; Cole, et al.,

2007; de Jong, 1995). Nevertheless, despite the fact that larger contrasts were made in

the S3 condition, speaker group differences were observed. Results showed that the

realizations of stress actively interacted with the two social factors – region and gender.

The specific discrepancies are summarized in Table 5.1.

As can be seen, the stress effect was observed for all speaker groups except Taipei

males. S3 conditions were usually characterized by sibilants being realized towards a

more dental direction. Cross-regional comparisons further showed that both Kaohsiung

females and Kaohsiung males exhibited an even more obvious stress effect than Taipei

speakers. It should be noticed that although the effect of stress was generally less shown

in both Taipei female and male speakers, the weak effect of stress should be understood

differently. Taipei females usually made larger sibilant contrasts than Kaohsiung

females; therefore, the reason why they did not distinguish different stress levels might

be due to the fact that they had reached the ceiling effect. In other words, Taipei females

were very articulate even in the S2 condition. In this regard, the contrast between S2

and S3 was not easily revealed. On the other hand, Taipei males always had

comparatively more merging sibilant realizations. Therefore, their lack of stress effect

was more related to the fact that Taipei males generally did not distinguish their

retroflex and dental sibilants successfully, regardless of stress levels.

Table 5.1 The degree of sibilant contrast made by different speaker groups.

Speaker group Sibilant contrast Realizations ^a Taipei Male No stress effect

Taipei Female S3 > S2 D (more dental) Kaohsiung Male S3 > S2 R & D (more dental) Kaohsiung Female S3 > S2 D (more dental)

a. The descriptions in this column indicate how the larger sibilant contrast was made. D stands for dental sibilants, while R stands for retroflex sibilants. For instance, the description for Taipei Female means that the larger contrast was made in the S3 condition by realizing dental sibilants toward a more dental direction.

Our results further showed that stress realizations were speaker gropu-dependent.

Given the same stress levels, sibilant realizations differed among speaker groups. Table

5.2 summarizes the comparisons of sibilant contrast degrees with regard to the effects of

gender and region. As can be seen in Table 5.2(a), females generally made larger

sibilant contrasts than males. According to sociolinguistic theories, females are more

conservative speakers and tend to use more standard forms, and the gender effect found

in the present study showed similar patterns with previous sociolinguistic studies (M.-C.

Li, 1995; C. C. Lin, 1983; Y.-H. Lin, 1988). Across stress levels, Taipei females

consistently distinguished sibilants better by realizing retroflex sibilants in a more

retroflexed way than Taipei males. For Kaohsiung males, on the other hand, female

speakers showed larger sibilant contrasts, but both their retroflex and dental sibilants

were realized towards dentalization. Such a result could be attributed to the fundamental

physiological difference between females and males. It is known that females usually

have shorter vocal tracts than males; therefore, it is expectable that their sibilant

frequencies will altogether be higher. This explained why Kaohsiung females, though

generally having larger sibilant contrasts, had both their retroflex and dental sibilants

realized higher in frequencies than males.

Table 5.2 The comparisons of sibilant contrast degrees between (a) genders in the same region and (b) regions of the same gender.

(a) Region Stress Sibilant contrast Realizations ^a S2 Female > Male R (more retroflex)

D (more dental) Taipei

S3 Female > Male R (more retroflex) S2 Female > Male R & D (more dental) Kaohsiung

S3 No gender effect

(b) Gender Stress Sibilant contrast Realizations ^a S2 Taipei > Kaohsiung R (more retroflex) Female

S3 No region effect

S2 Kaohsiung > Taipei R (more retroflex) Male

S3 Kaohsiung > Taipei R (more retroflex)

a. The descriptions in this column indicate how the larger sibilant contrast was made. D stands for dental sibilants, while R stands for retroflex sibilants. For instance, the description for Taipei speakers in the S2 condition means that females made larger sibilant contrasts than males by realizing both retroflex sibilants and dental sibilants in more extreme directions.

The effect of region is shown in Table 5.2(b). As can be seen, the speaker groups

that made greater sibilant distinctions all had retroflex sibilants being realized more

retroflexed. Cross-regional comparisons revealed different patterns. For female speakers,

Taipei females distinguished sibilants to a greater extent than Kaohsiung females in the

S2 condition, whereas for male speakers, it was Kaohsiung males that made greater

sibilant distinction, regardless of stress levels. This inconsistency was not expected. Our

original prediction for the region effect was that since Kaohsiung speakers speak Min

more frequently than Taipei speakers, if the frequency of using Min is indeed effective,

the merging degree should be more salient for Kaohsiung speakers. While this was

generally true for female speakers, male speakers did not follow this pattern.

Examinations of male speakers’ metadata showed that possibly there was another

confounding factor involved, i.e., Min proficiency. As shown in Table 3.2, while female

speakers in the two regions had relatively comparable Min proficiency, Taipei males

shown that Min proficiency have an effect on the realizations of retroflex sibilants. For

example, Lin (1983) found that native Taiwanese speakers, proficient in Min, are more

likely to mispronounce retroflex sibilants than native Mandarin speakers, non-proficient

in Min. In other words, the more proficiently one speaks Min, the greater degree of

sibilant merging would be expected. In the present study, since Taipei males showed

more merging patterns than Kaohsiung males, it thus suggested that the effect of Min

proficiency had overridden the effect of region, which is correlated with the frequency

of using Min.

The sibilant merging patterns of Taipei males could be further illustrated by our

substitution data. Substitution demonstrated the extreme cases for sibilant merging. In

particular, it was found that the substitution of dental sibilants for retroflex ones,

corresponded with the total deretroflexion process depicted in previous studies, was

especially frequent for Taipei males. Such a result was compatible with the analysis of

canonically realized sibilant tokens, in which Taipei males showed the most sibilant

merging patterns, and in certain conditions, retroflex and dental sibilants were not even

distinguished. Taken together, it could be concluded that for Taipei male speakers,

sibilant merging was actively in progress. While most of previous studies usually

roughly referred to this sibilant merging phenomenon as deretroflexion, our study

further identified different realizations of sibilant merging. That is, for Taipei male

speakers, not only did retroflex sibilants showed the tendency to become more similar

to dental ones, but the extreme cases of total deretroflexion were seen from time to time.

While deretroflexion is generally the norm observed in Taiwan Mandarin, it was

interesting to see that the other direction of substitution, i.e., the substitution of retroflex

sibilants for dental ones, was also found in our data. Particularly for Kaohsiung females,

they had a number of instances of such substitution. Since this substitution direction

was contradictory to what we generally observed in Taiwan Mandarin, we needed to

explain its occurrence. Chung (2006) thought of such a phenomenon as hypercorrection.

Given the fact that retroflexion is commonly considered to be prestigious, speakers

would incorrectly substitute them for dental sibilants. Taking this into accounts, we

should be able to see why Kaohsiung females hypercorrected so often. According to

Labov (1990), in general, female speakers are more linguistically insecure, and thus

they adopt the prestigious forms more often than male speakers. If we consider Taipei

dialect as the standard dialect of Taiwan Mandarin, Kaohsiung speakers, especially

females, would be eager to modify their speech in approach of the standard dialect.

Therefore, it was not surprising that more hypercorrections were found in the speech of

Kaohsiung females. Their attempts to adopt the standard form were also revealed in our

canonically realized sibilant data. As discussed previously, while Kaohsiung females

made smaller contrasts than Taipei females at the default stress level, this regional

difference was however eliminated in the prosodically prominent condition. Taken

together, it could be seen that Kaohsiung females were so sensitive to the standard

performance that they were eager to approach the standard form. Such eagerness was

further demonstrated by their mistaken usage of retroflexion in their speech.