«Among the seven (not counting /er/) single vowels in standard Chinese [i, y z,, u, ɤ where z, ] stand for the high vowels following dentialveolars ...»
Er-suffixation in Chinese monophthongs:
phonological analysis and phonetic data
Tsan Huang 黄璨
University at Buffalo (SUNY)
Among the seven (not counting /er/) single vowels in standard Chinese [i, y z,, u, ɤ
where z, ] stand for the high vowels following dentialveolars and retroflexes), previous
studies have analyzed the surface forms of er-suffixed /u, ɤ, a/ as the original vowel
followed by either [ɹ] (Chao, 1968; Cheng, 1973; Pulleyblank, 1984; Lin, 1989; among others) or [ʅ] (i.e. [ ], Li, 1986), while the er-suffixed non-back vowels are said to surface as a concatenation of the original vowel or its glide counterpart (if there is an underlying vowel) and [əɹ]. However, Duanmu (2007) offers a different analysis, where [ɚ] is the final component in the surface forms of all seven monophthongs.
To evaluate these accounts, the present study investigated the production of the plain and er-suffixed monophthongs by eight native speakers from Beijing. Our data show (1) that there is no evidence for a separate [ɹ] coda in the surface forms, (2) that in addition to lowering of F3, er-suffixation has a centralizing effect on all monophthongs, and (3) that there is significant interaction between er-suffixation and vowel quality. The analysis of the Chinese er-suffixed monophthongs is thus revised as: [jɚ, ɥɚ, ɚ, ɚ; u˞; ɤ˞ɚ ˞ɚ]. Our
results also call for more careful phonetic study of similar processes cross-linguistically:
while there is evidence for a coda /r/ in Scottish English (Wells, 1982), in other languages the underlying rhotic segment or syll ble seems to be re lized only s “r-coloring” on the preceding vowel (and perhaps an [ɚ] off-glide). If this is true, then for cases such as Yanggu (Lin, 2004) one may not have to posit surface syllables with both a complex onset (e.g. [tl]) and an /r/ coda.
1. Introduction Chinese being impoverished in morphology, the rare phenomenon of er-suffixation (or erhuay 儿化) observed cross m ny di lects h s ttr cted linguists‟ ttention from e rly on (e.g. Wang, 1963; Chao, 1968; Cheng, 1973; T. Lin, 1982; Wang & He, 1983; Li, 1986; Y. Lin, 1989, 2004, 2007; Duanmu, 2007). Historically, the retroflex suffix /-ɚ/1 Not tion l conventions: 1) I sh ll lso use “/-ɚ/-suffix tion” nd “er-suffix tion” interch nge- 1 ably. (2) Where an exact phonetic transcription is not necessary, I shall use Pinyin with superscript tone description (e.g. /er35/ „son‟). The numbers indic te pitch heights on five-point scale (Chao, 1930), and Ch o‟s 1968) description of the tones i.e. tone 1 = 55 tone 2 = 35 tone 3 = 214, and tone 4 = 51) is adopted here. (3) Where an exact phonetic transcription is not necessary and for ease of typesetting, I may use “V+r”/“Vr”/“V-r” to indicate an er-suffixed vowel (e.g. [cha35.jir55 „te t ble‟). 4) V m y be used in short for „vowel‟.
HUANG: ER-SUFFIXATION IN CHINESEoriginated mainly from four morphemes (Chao, 1968; Li, 1986): (1) the diminutive /er/儿;
(2) /ri/日 “d y”; (3) the locative /li/里 “inside”; and (4) the perfective /le/了. (There are also a few isolated lexical items that employ rhoticity to convey specific meanings. See, e.g., T. Lin, 1982.) The diminutive /er/ 儿 is the most productive synchronically. Chao (1968) has a fairly detailed discussion on the types of morpheme or word to which it can attach and how it alters the original meanings of those morphemes or words in Beijing Mandarin. The /-ɚ/ suffix derived from /ri51/ 日 “d y” occurs in sm ll set of frequently occurring colloquial words and appears to be fossilized: e.g. in /jin55-ri51/ „tod y‟, /zuo35ri51/ „yesterd y‟ nd /ming35-ri51/ „tomorrow‟, /ri/ is deleted and the words surface as monosyllables with rhotic vowels (where /n/ in /jin55/ is also lost, while the engma in /ming35/ is deleted with nasality preserved on the vowel). The same can be said of the locative /li/ 里: e.g. /zhe51-li214/ „here‟ /na51-li214/ „there‟, /na214-li214/ „where‟, and /wu55li214/ „inside the house‟ (as in [jin51.wur55.zuo51.ba] „come in nd sit down‟). The perfective (/lə/了) surfaces as rhoticity in a few dialects (but not in Beijing Mandarin) (see Li, 1986).
The /-ɚ/ suffix can be combined with any rhyme in Beijing Mandarin other than the rhotic rhyme /er/, as in /er35/ „son‟, /er214/ „ear‟, etc. (see also Chao, 1968 and Li, 1986), producing many r-colored surface vowels (or erhuayin 儿 化 音 ). This paper focuses on how /-ɚ/-suffixation affects the seven monophthongs (i.e. single vowels; not counting the rhotic vowel /er/): i y z,, u, ɤ, a] (Cf. Chao, 1968). The fricative vowel z] only occurs after the dentialveolars /ts, tsh, s/, while the other fricative vowel [ ] is found only after the retroflex sibilants /tʂ, tʂh, ʂ, /. These monophthongs do not have the same phonological status in the sound inventory (see Duanmu, 2007 for a phonological analysis of the vowels). However, since the /ɚ/ suffix has different effects on these sounds, all seven of them will be examined and discussed separately.
2. Previous studies Different surface forms of /ɚ/-suffixed single vowels in Beijing Mandarin have been proposed by various researchers. Some are cited in (1) below.
(1) Surface forms proposed by various researchers for i y z,, u, ɤ, a] + /ɚ/
Notational differences aside, these analyses differ most noticeably in the surface forms of er-suffixed /i, y/: Li (1986) and Lin (1989) treat them as mostly unchanged in quality (with a schwa off-glide), whereas Pulleyblank (1984) has a rule that allows /i, y/ to lax into glide+schwa before /r/ (1984:56). Although Duanmu (2007) also has the high front vowel laxing rule, his surface rhymes derived from /i, y/ consist of a non-rhotic schwa as the nucleus and a rhotic schwa off-glide. In addition, Li (1986) uses [ʅ] (IPA [ ]) in his analysis. or Li (1986) claims, perhaps in agree-ment with Pulleyblank (1984) and Lin (1989), that in er-suffixed /u, a/ rhoticity is fused with the original vowels rather than being realized as an off-glide, whereas Duanmu (2007) has a rhotic schwa off-glide after the [u, a] nuclei. One notes further that for Pulleyblank (1984) and Lin (1989) er-suffixed /ɤ/ seems to also be analyzed as unchanged in vowel quality, whereas Duanmu (2007) again has a rhotic schwa off-glide after the nucleus.
The goal of the present study is to examine the surface monophthong rhymes acoustically and to compare them with their plain counterparts (i.e. surface monophthongs without the er-suffix) to see how rhoticity affects each of them. It is hoped that the results from this study will help us to better understand the er-suffixation process and to achieve more accurate phonological analysis.2
3.1 Recording Six (6) female and six (6) male speakers from the city of Beijing, average age 28, were recorded in a sound-attenuated booth, with a head-mounted Shure® SM-10A microphone and a Marantz® solid state recorder. The speakers read a semi-randomized list of 350×2 disyllabic words, balanced for rhyme and tone, with the second syllable containing the target plain rhyme or er-suffixed rhyme. The whole list consists of about 350 words and 38 sentences. Only the results for the single vowels from eight (four male and four female) speakers will be discussed here.
3.2 Data tagging and taking measurements After the recordings were transferred to a computer, they were segmented into words (and sentences), and saved as individual files using Praat (Boersma, 2001). The target vowels were then further segmented into the first 20-milisecond, last 20-milisecond and center portions and labeled with Praat (see Figure 1).
Measurements of the first four formants (i.e. F1, F2, F3 and F4) were taken from the center point of each of the three labeled segments in the target vowel (i.e. at 10ms from the vowel onset, mid point of vowel and 10ms from the vowel offset) with a Praat script (adapted from one originally written by Mietta Lennes). The LPC parameters were adjusted for each individual speaker. In general, five (5) formants under 5000Hz were 2 There were earlier acoustic studies (Wang & He, 1983; Lin, Zhou & Cai, 1998; Shi,
2003) on –r suffixation in Chinese, all of which used only one speaker.
333 HUANG: ER-SUFFIXATION IN CHINESE
specified for male speakers and five (5) formants under 5500Hz for female speakers. In an er-suffixed form, due to the presence of the rhoticity, formants may change drastically, with the frequency range for the first five formants lowered by 500 to 1000Hz. As can be seen in the right panel of Figure 1, the F3, normally around 3100Hz for this speaker, is lowered to below 1900Hz and an extra fifth formant is now visible under 5000Hz.
Accordingly, the LPC parameters were set to look for five (5) formants under 4500Hz for male and under 5000Hz for female in the er-suffixed forms. In some cases, further finertuned adjustments were made to get the correct measurements. The formant measurement data were then read into a text file and plotted using PlotFormant, a computer program developed by the late Peter Ladefoged of UCLA.
Figure 1. Sound files were segmented and labeled with Praat.
Pinyin was used and tones indicated with numbers 1, 2, 3, and 4 (for T55, T35, T214 and T51, respectively). The letters „e‟ „m‟ nd „l‟ st nd for „e rly‟ „mid‟ nd „l te‟ respectively. The left panel shows the word /fei55.ji55/ „ irpl ne‟ cont ining a plain /i/ in the second syllable, and the right panel the word /cha35.jir55/ „te t ble‟ containing a rhotic /i/. The vertical lines mark the vowel onset, 20ms from the onset, 20ms from the vowel offset, and the end of vowel.
4. Results and Discussion
4.1 Beijing plain monophthongs Figure 2. Beijing plain monophthongs (female speakers). Formants were taken at the mid-points in vowels. The left panel shows the F1 vs. F2 plot, and the right one F1 vs. F3.
Figure 2 shows the steady state values of the seven monophthong vowels in Beijing Mandarin (female speakers), with the rhotic vowel er also plotted for comparison.
The left panel shows our familiar vowel space. The vowels are fairly well separated from each other, differing somewhat along either the F1 or F2 dimension or both, except for [ ] (labeled “zhi” in the chart) and z ] (labeled “zi”). Note th t “e” st nds for the b ck mid vowel [ɤ].) But in the right panel, a large difference between [ ] and z ] is seen along the F3 dimension. Note also that F3 in /y/ is noticeably lower than /i/, obviously to help maintain a perceptual distance from /i/; or one may say that F3 in /i/ is particularly high for the same purpose. The F3s in “zhi” (i.e. [ ]) nd “er” re consider bly lower due to the presence of rhoticity. The F3 values in other vowels are very similar and are around 3000Hz for the female Beijing Mandarin speakers.
4.2 Beijing rhotacized monophthongs Recall that formant measurements were taken at three points in the target vowel. In the “e rly” me surement data taken at 10ms after V onset (left panel of Figure 3), the vowel space already starts to shrink (Cf. Figure 2). Thus, the effects of rhoticity kick in from early on. Only the front and back high Vs maintain a separation both in F1 and F2 from the rest of the Vs.
Figure 3. Plot of “e rly” form nt me surements for Beijing rhotacized monophthongs (female speakers).
Note th t “er” denotes the inherent rhotic vowel while “e+r” st nds for /ɤr/.) In the plot of measurements taken from the mid points of the Vs (Figure 4), all vowels huddle together, except for /u-r/, which still maintains its height. Along the F3 dimension, it is obvious that rhoticity is present in all Vs. At 10ms from the vowel offset (Figure 5), things crowd even closer together, except, again, for [ur]. Along the F3 dimension, heavy rhoticity is observable in all Vs.
Figure 4. Plot of “mid” form nt me surements for Beijing rhot cized monophthongs (female speakers).
Figure 5. Plot of “l te” form nt me surements for Beijing rhot cized monophthongs (female speakers).
4.3 Comparison of plain vs. rhotacized monophothongs in Beijing Mandarin An n lysis of v ri nce ANOVA) w s run on the me surement d t with “gender of spe ker” s the between-subjects v ri ble nd “vowel qu lity” “presence (or absence) of er-suffix” nd “me surement point” s within-subject variables. ANOVA main effects were significant (p.001) for all variables. The four-way interaction of gender * vowel * er-suffix * measurement point was not significant for any of F1, F2 and F3, indicating that the female and male speakers behave very similarly – with the normal difference in frequency range across genders noted. This is not surprising, given that speakers of either gender can produce er-suffixed forms and be correctly understood by all fellow speakers of the language, who can always tell whether an er-suffixed form or a plain V was intended by the speaker. Thus, data patterns for F1, F2 and F3 from the mid point of the vowel, aggregated across genders, are presented here. As can be seen in the left-most panel, changes induced by the er-suffix in F1 (except for [u] and [u-r]) and F3 (except for [ ] and [ -r l beled “zhi”) are significant across the board. For /ɤ/ l beled “e”), the overlap of F2 measurements was only 5Hz. For [z] and [z-r] l beled “/zi/”), the overlap was 35Hz. So, these changes are marginal. Furthermore, as noted before, these changes under the effect of er-suffixation made the vowels less distinctive from each other.
Indeed, F2 (except for /u/) and F3 values are very similar in all er-suffixed vowels.
337 HUANG: ER-SUFFIXATION IN CHINESE
Figure 6. Comparison of formant values in Beijing Mandarin plain (circles with dashed lines) versus er-suffixed (triangles with solid lines) vowels.
(Ag in “e” “zhi” nd “zi” stand for [ɤ], [ ] and [z], respectively.) An asterisk marks a significant change. Formant values of the inherent rhotic vowel /er/ are also plotted as a reference.