thai tone variants and the language teachers

THAI TONE VARIANTS AND THE LANGUAGE TEACHERS

Adrian Palmer

It is common knowledge to teachers of tone languages that the phonetic shapes of tones vary in different phonological environments. This paper, based upon a limited instrumental study of Thai, provides a technique by which a teacher can identify the en- vironmental conditions under which confusion is likely to occur. Two separate methods of charting are illustrated. One shows individual tone variation, and the other shows the effect of a specified tonal environment on all five tones. Both kinds of charts are useful in teaching Thai tones.

Language teachers,in teaching students to recognize and produce tones in tone languages, invariably hear their students make these complaints: “1 can’t hear the difference between tone A and tone B,” and “Tone A sounds different in this word than it does in that one.” The student is reacting to two facts about tones of which the teacher should be aware. These are:

1) A single tone can change in different environments, and 2) Tones which contrast in one environment may not contrast

This study offers three valuable aids to the teacher of tone languages. First, it illustrates a way of displaying tones on charts which show directly how a single tone can vary in different environments and how contrasting tones become neutralized in certain environments. Second, it suggests a procedure for instrumental analysis of tone which is applicable to any tone language. Third, it provides the teacher of Thai with a set of reference charts illustrating tone variation and contrast within the Thai tonal system.

In textbooks of Thai for speakers of English an attempt is often made to illustrate the relative pitch contour of syllables pronounced with each of the five Thai tones. Generally these pitch contours are given as lines on a scale which represents the pitch range of one’s normal speaking voice. The pitch contours given below are taken from the A.U.A. LANGUAGE CENTER THAI COURSE, Book 1, Prepared by J. Marvin Brown (1968). The drawings indicate the relative shapes and positions of the five tones. The label below each contour is the name by which the tone is identified in English.

in another environment.

287

288 LANGUAGE LEARNING, VOL. XM, NOS. 3 & 4

Table 1

Mid LOW Falling High Rising Tone name

symbol

A ” Tone -

The symbol below the label indicates how the tones will be marked above Thai syllables.

This is the only attempt made at illustrating the tone contours in this text. From this point on in the A.U.A. course, the student learns by imitation and practice with the help of numerous excel- lent drills. The effectiveness of the practice depends to a large part on the teacher’s ability to evaluate student performance and suggest appropriate corrective measures to those who make mis- takes.

From the drawings above the reader might draw the following conclusions:

1) The end point of the falling tone and the start of the rising

2) The high tone is essentially level except for a short (‘spike”

3) The rising tone rises to the pitch level of the high tone.

4) The low tone is level.

5) The tones always have the relative positions and shapes

The following study represents a restricted investigation of tone perturbations in specified environments, and it will allow us to evaluate these conclusions. It shows how the shape of the Thai tones is influenced by the tone of the preceding and following syllable. The experiment consisted of having a Thai speaker record

tone are the same.

near the end.

illustrated in the diagrams.

THAI TONE VARIANTS AND LANGUAGE TEACHERS 289

a set of 25 sentences of four syllables each.l The first and last words in the sentences were held constant. Both had mid tone. The middle two words were varied to give all possible sequences of tones. A change in the pitch contour of a particular tone from utterance to utterance can be attributed to the effect of the preceding or following tone.

To measure the pitch contour of a speaker’s voice a sound spectrograph can be used. The sound spectrograph is a machine which passes a taped utterance through a set of filters (or a single filter which is moved to correspond to a set of filters). If the filter’s bandpass is sufficiently wide, then the formants, which are bands of relatively high acoustic energy, will appear on the spectrogram. The noise present in sibilants and stop releases is also visible on the spectrogram. With this information it is usually possible to segment an utterance and identify the different speech sounds of which it is composed.

If the filter(s) in the spectrograph have a sufficiently narrow band-pass the individual harmonics of the fundamental frequency of the speaker’s voice can be distinguished on the spectrogram. The distance of any harmonic above the base line on the spectrogram is directly related to the fundamental frequency of the voice. Thus the narrow band spectrogram provides us with a permanent record of the pitch contours in an utterance.

Narrow band spectrograms are more helpful in distinguishing individual harmonics than formants. This makes segmentation of an utterance more difficult. However if the content of the utterance is known, and if it is short, it is possible to segment utterances directly from a narrow band spectrogram.

A list of sentences was prepared, each consisting of four mono- syllabic words. The middle two words were varied to give all two-tone combinations. The first and last words of the sentences were held constant. The sentences were put into random order and written in the Thai script. A native speaker of Thai read the list twice, Recordings were made at the University of Michigan Language Laboratory on an Ampex 350 full track tape recorder with a Neumann U-67 condenser microphone in a soundproof recording room. The adult male informant was asked to read the utterances clearly and at normal speed. Narrow band spectrograms of each utterance were made on a BTL-2 Model-D sound spectrograph at the University of Michigan. The machine was not calibrated before usage.

High shaping was used.

‘See Appendix 1 . I would like to thank Hom Klayanonda for recording the Thai utterances and express my appreciation to John Peterson for his many helpful comments.

290 LANGUAGE LEARNING, VOL. WL, NOS. 3 & 4

Measurements Measurements of the fundamental frequency of the speaker’s

voice during his production of the sentences were taken from narrow band spectrograms. The tenth harmonic was located and marked on the spectrogram. A template was used which related the distance above the base line on the spectrograms to frequency in cycles per second. By dividing the frequency of the tenth harmonic by ten, the fundamental frequency of the speaker’s voice was obtained.

Normalization The fundamental frequency of the speaker’s voice varied be-

tween 100 and 150 cycles per second (cps). It was decided to round off the end points of each pitch contour to the nearest 10 cps,which was as close as one could read the spectrograms with consistency (repeatability).

Transitions between tones were ignored. In utterances 6- 15 the third word began with a voiced sound. When this was the case there was often a rapid transition from the end of the preceding tone to the beginning of the following tone which could be seen on the spectrograms. This transition was very short compared to the length of the pitch contour on the vocalic nucleus of the syllable. The transition was not visible when the syllable began with a voice- less sound. The transition, when visible, was predictable as a rapid change from the end pitch of the preceding tone and the in- itiating pitch of the following tone.

Consistency In only 7 out of 25 repeti-

tions was there a measurable difference of 10 cps (maximum) in the initiation or terminus of the pitch contour. For 15 of the remaining pairs the pitch measurements were the same. For three pairs the difference between pitch measurements was on the order of 5 cps and difficult to measure consistently. Thus in no case would the variation between repetitions of the same utterance ac- count for a difference of more than one point on the six-point scale which is used below in charting pitch contours.

Each sentence was recorded twice.

Presentation of results The measurements of fundamental voice frequency are pre-

sented in charts which make it easy to observe two phenomena: 1) The variation in a single tone which is due to the effect of

the tone of the preceding or following syllable.


2) The effect of the preceding or following tone on how all five tones contrast (whether there is neutralization between contrasting tones in certain tone environments).

In the following charts, the term “initialy* refers to the second word in the frame sentence, and c‘final” refers to the third word. In sentence 15, rZw h& rhz dz’i, only the second and third words are of interest. Thus hza is “initial” with respect to &, and & is 6gfinal” with respect t o h h . In this sentence it is possible to observe the effect of an “initial” rising tone on a “final” high tone, and vice versa.

Thus each tone is shown as it is affected by the five tones which follow it.

Charts 1-5 each show a single tone in initial position.

The key to the interpretation of charts 1-10 is as follows:

.......... = affected by mid tone -*-.-.-*-. = affected by low tone

---------- = affected by falling tone = affected by high tone

XXXXXXXMM = affected by rising tone

For convenience, all the charts a re reproduced in Appendix 2. On all charts the base line represents the lowest fundamental frequency encountered (100 cps), and the top line represents the high- est (150 cps). The subdivisions marked on the left sides of the charts represent 10 cps increments. In the body of this paper four of the charts are discussed in some detail. Of Charts 1-5 which each show one tone in initial position, a typical one is Chart 4.

C h a r t 4

Initial High

Chart 4 indicates that initial high tone is relatively unaffected by changes in the following tone. The initiation and terminating frequency does not vary more than 10 cps and all tones show a

29 2 LANGUAGE LEARNING, VOL. WL, NOS. 3 & 4

rising pitch contour. High tone when followed by either high or falling tone has only one contour. It begins at 130 cps and rises to 140 cps. When high tone is followed by rising o r mid tone it also has only one contour, beginning at 130 cps and rising to 150 cps. The point of initiation of these two contours is the same (130 cps). They terminate 10 cps apart. When high tone is followed by low tone its initiation is 10 cps higher than in the previ- ous cases. One can conclude that these differences a re insignifi- cant since 10 cps is almost within the range of e r ror one has in reading the spectrograms and normalization. Also, as noted before, there was as much as a 10 cps variation among tones between repetitions of a single utterance.

Thus, each tone is shown as it is affected by the five tones that precede it. In these charts the greatest variation among tones in final position is shown in the high tone.

Charts 6-10 each show one tone in final position.

This is illustrated in Chart 9.

Chart 9

Final High

Chart 9 shows that the high tone is affected by the preceding tones s o much that it changes direction. ceded by either mid, low, or falling tones, it has a rising slope (which it also has in initial position). However, when preceded by high or rising tone, it has a falling contour. This falling contour can be noticed when listening to the speech of Thais, though it ap- pears to be pointed out to students only rarely. The variation in point of initiation is as much as 20 cps, and the variation in point of termination is 30 cps. These values are outside the range of tone variation found in repetitions of a single utterance. This find- ing suggests that the student should be informed of the variation so he does not become confused when he hears a high tone with a falling pitch contour.

By examining Charts 1-9 in the Appendix, one can determine whether o r not the conclusions on p. 288 are correct. These charts confirm the first conclusion that the termination frequency of a

When high tone is pre-


falling tone is essentially the same as the onset frequency of a rising tone. They suggest that high tone might be more accurately diagramed as a slight r ise at a high pitch level (except when following high or rising tones as noted). Conclusion 3 is generally confirmed. Conclusion 4 should be modified to state that the low tone generally has a slight falling contour at a low pitch level.

The tones do not always have the same relative positions and shapes indicated in Table.1.

. However, there is remarkable consistency in the shapes of the tones. Thus, on the basis of these data, it would be justifiable for a teacher to present the tones to the student as generally un- changing and note only that the final high tone shows some variation, The revised tone diagrams the teacher could use are the

Conclusion 5 is shown to be untrue.

following :

Table 3

, \ , Mid Low Falling High Rising Tone

name Tone -

symbol

A ”

The remaining charts each illustrate all five tones in contrast. Each chart shows this contrast in a particular tone environment. In Charts 11-15 the five tones are shown as they contrast in final position. Thus each chart illustrates the effect of a particular preceding tone on all five tones. In Charts 16-20 the five tones are shown as they contrast in initial position. Thus each of these charts illustrates the effect of a particular following tone on all five tones. One chart from each set will be presented and examined in some detail in the body of the paper. This will be followed by some general conclusions drawn from an inspection of the remaining charts in Appendix 2. The key to the interpretation of Charts 11-20 is as follows:

294 LANGUAGE LEARNING, VOL. WL, NOS. 3 & 4

.......... = mid tone _._.-.-.-. = low tone ---------- = falling tone

= high tone

xxxxxxxxxx = rising tone

Charts 11-15 each show all five tones in final position. An inspection of Chart 12, for example, shows how the five tones contrast following low tone.

Chart 12

Low -

In this environment mid and low tone are identical, s o the contrast has been completely neutralized. In addition, high and rising tones are almost the same, showing only a 10 cps difference in point of initiation. Since this difference lies within the range of variation between repetitions of a single utterance, one could expect that this contrast would not always be preserved.

Other cases of neutralization, or near neutralization, in the final position are illustrated in the remaining charts of this set (11-15) in the Appendix. Chart 11 shows that following mid tone there is near neutralization of mid and low tones (only a 10 cps difference in the pitch of the tone termination). Chart 13 shows that following falling tone, the neutralization of mid and low tone and of high and rising tone is similar to that shown in Chart 12. Chart 14 shows that following high tone, mid tone is lower than low tone, although only by 10 cps. It would also be possible in this environment to confuse high tone with low tone. Chart 15 shows that following rising tone, mid and high tone might be confused as their initiation differs only by 10 cps.

Charts 16-20 show how all five tones contrast in initial position. Chart 16 is typical of this group.


Chart 16

Mid

Chart 16 shows how all five tones appear when they are followed by mid tone. Here the tones are well spread out and there appear to be no cases where tones would be confused. The tones hold closely to their averaged contours on Table 3. This seems to be also the case with the remaining initial environments shown in Charts 17-20 (AppendE). There a re no instances of neutralization or near neutralization such as that which is evident in Charts 11-15. At least 20 cps separates either the initiation or termination of any two tones on these charts. Thus, the initial environment is a more stable one for tones than the final environment.

Advice to the teacher Though these charts suggest that there will be real problems

in distinguishing certain tones in particular environments (see Charts 11-15), such might not always be the case. The Thai speaker may speak more slowly in order to preserve maximum distinctness between tones. At first students should be taught to produce the tones while preserving maximum distinctness. The obscuring of tonal distinctions comes naturally enough to most students and should not be encouraged. Students, however, should not be asked to perceive distinctions between tones where distinctions do not exist .

A certain amount of deliberate distortion of the tones in the early stages of instruction leads to a later maintenance of distinctions where it is desirable. This is particularly true in the case of mid tone. From Charts 1 and 6 one can see that mid tone generally has a slight fall in pitch. If students are allowed to produce mid tone with this fall, they often end up making a tone which is mistaken for falling tone. From Charts 11-20 one can see that mid tone should rarely be confused with falling tone (possibly only when followed by high tone). Thus the teachers should insist on the production of an absolutely level mid tone in the early stages of instruction so that it does not later degenerate into a falling

296 LANGUAGE LEARNING, VOL. Mx, NOS. 3 & 4

tone. With this pedagogical goal in mind, the contour for mid tone in Table 3 might be changed to indicate that it is level, although this is not really true.

The charts included in the Appendix a re particularly useful. With a little practice the teacher can quickly locate a chart which will enable him to decide if a student’s complaint of difficulty in discriminating between two tones is based upon a real case of tone ambiguity or whether the student is failing to hear a contrast which is actually present. The teacher needs only to determine the tone environment in which the suggested ambiguity is occurring, locate this environment on a chart, and examine the tone contours.

Future studies This study does not investigate how tones are produced in

checked syllables, nor a re the additional phonetic features of creaky voice or glottalization described in relation to the tone system. Neutralization of tone is unstressed syllables with short vowels is also a phenomenon of which the teacher must be aware, but which has not been studied instrumentally.

REFERENCE

Brown, J. Marvin. 1968. A.U.A. Language Center Thai Course. Bangkok, Thailand: American University Alumni Association Language Center.


Appendix 1

1. r a w daa kay d I i

2. r a w rnIi kay d?i

3. r a w chiy k$ d i i

4. r a w hAy k$ d z i

5. r a w hga k$ d i i

6. r a w dAa w a a d z i

7. r a w Gi Ga d F i

8. r a w chgy &a d i i

9. raw hay i a d i i

10. r a w hga &a d i i

11. r a w &a rda d i i

12. r a w di ru'a d i i

13. r a w chiy rda d I i

14. r a w hay rda d i i

15. r a w hga rda d i i

16. r a w &a phaa d i i

17. r a w di phaa d i i

18. r& chiy phaa d i i

19. r a w hay phaa d i i

20. r a w h l a phaa d l i

21. r a w &a &a d i i

22. r a w r n i i fga dTi

23. r'aw chiy fga d i i

24. r a w h%y fga d i i

25. r a w hga €5, d?i

W e cursed the good chicken.

W e have a good chicken.

We used a good chicken.

We gave away a good chicken.

We looked f o r a good chicken.

W e cursed the good cow.

We have a good cow.

We used a good cow.

W e gave away a good cow.

We looked f o r a good COW.

We cursed the good fence.

We have a good fence.

We used a good fence.

We gave away a good fence.

W e looked f o r a good fence.

W e cursed the good cloth.

We have good cloth.

W e used good cloth.

We gave away good c lo th .

We looked f o r good c lo th .

We cursed the good w a l l .

W e have a good wall.

We used a good wall.

We gave away a good wall.

We looked €or a good w a l l .

298

N

.5 -2 a, a a c

LANGUAGE LEARNING, VOL. WL, NOS. 3 & 4

8 .rl

? &

X *

.rl

Y E

8 .rl

X ua

3 .rl

R

thai tone variants and the language teachers

Documents