Proceedings of Meetings on Acoustics

A. Coombs

Volume 19, 2013 http://acousticalsociety.org/

ICA 2013 Montreal Montreal, Canada 2 - 7 June 2013

Speech Communication Session 4pSCb: Production and Perception I: Beyond the Speech Segment (Poster Session)

4pSCb51. High tone processes in Ibibio Afton L. Coombs*

*Corresponding author's address: Linguistics, University of Southern California, Los Angeles, CA 90089, [email protected] Downdrift and downstep are processes which may cause lowering of high tone syllables. Downdrift is intonational, occurring at phrasal or utterance level, while downstep is a phonological process which acts from one tone-bearing unit to the next such that H tones lower successively. The relationship between these larger tonal lowering processes and individual tone units is complicated, however, by processes which may raise or preserve original high tone pitches. Ibibio, a Niger-Congo language spoken in southeastern Nigeria, is a terraced tone language with contrastive H and L tones. H tones in Ibibio experience automatic and non-automatic downstep, lowering both in sequences of high tones and around intervening lows. This study aims to determine those factors which counteract or overrule the downstep process. Average pitch readings were taken of entire syllables and compared with readings of other syllables within the same word. The main finding of this study is that while single words show acoustically measurable downtrends, they also show non-lowering and even raising of high tones, specifically in HHL contexts. This complicates how downtrends act across tone-bearing syllables, and may indicate that a high tone is raised in order to increase contrast with a following low.

Published by the Acoustical Society of America through the American Institute of Physics

HIGH AND LOW TONE IN IBIBIO

Downstep and downdrift are two causes of lowering in high-tone syllables in Ibibio (Urua 2001). The difference, however, if that downdrift is a phonetic process which may function across tones even on a phrasal level, while downstep is the acoustic sign of a contrastive tonological process with phonological effects (Urua 2001, Connell 2001). A downstepped high tone (!H) has a lower tone than previous H tones, but still much higher than L tones. A downstepped high tone can only stand word-internally after a H tone, and the sequence !HH never occurs across the language (Urua 2001). These !H tones create a ceiling above which other H tones cannot reach (Connell, 2001). Ibibio has been referred to as a classic terracing language (Connell 2001, Urua 2001) with highs that terrace by downstepping in succession or by downdrifting across the utterance, creating successively “terraced” levels. The data used for this paper are taken from Professor Pat Keating's files on the UCLA Phonetics Archive website, recorded in 1984 in Port Harcourt, Nigeria. The speaker's elicitations are labeled by word, gloss, speaker gender and number, and lettered repetition. They form a set of single-, two-, three-, and four-syllable words used for establishing phonemic contrasts in Ibibio. For this paper, however, they will be used to explore downtrends on high tones in Ibibio, specifically in factors which complicate and override these processes. Praat, developed by Paul Boersma and David Weeink, was used to segment, label, and measure the tones of the data. Average pitch readings were taken of entire segments on which high or low tones were marked. While tonal processes are difficult to study in such a limited set of data, these speakers do show consistent patterns in high tone variation which may influence downstep and downdrift, and it is the goal of this paper to explore and explain these patterns. While H tones with intervening L tones may undergo downstep (Connell 2001), it is not completely clear how H tones in succession may undergo a phrasal downdrift, with arguments for stable H tone (Urua 1996, Urua 2001) and downtrending H tone (Boys 1976 85). These data do show lowering on H tones with intervening L tones, but H tones in succession remain relatively stable and do not appear to undergo any lowering processes. In addition, non- word initial, pre-low H tones consistently rise in these data.

SUCCESSIVE LEVEL HIGH TONES

In these data, sequences of high tones here do not lower, but rather remain at a constant level pitch for the duration of the utterance. This is in contrast with H tones around an intervening L, which do lower. Table 1 shows lowering trends around a L tone for the purpose of comparison.

TABLE 1. Lowered H tones around L tone: H tones after intervening L are lower than previous H tones. Word Gloss Speaker Repetition H1 (Hz) H2 (Hz)

ḿbòppó maiden M2 a 175.85 149.20 b 165.32 141.95 ḿbòppó maiden M1 a 122.74 105.46 b 120.80 105.85

This lowering can also be observed on Figure 1, where the final H is distinctly lower than the previous H, dropping from 175.85 to 149.20 Hz. Note that these are true H tones, and the second is not a downstepped high (Urua 2004). The lower level then is not due to downstepping from the high tone, but rather from an intonational process triggered by the intervening L.

Proceedings of Meetings on Acoustics, Vol. 19, 060232 (2013) Page 2 A. Coombs

250 ) z H (

h c t i P

50 0 0.748 Time (s)

FIGURE 1. Final H tone lowering: ḿbòppó 'maiden,' M1a. The second H is lowered by the intervening L tone.

In contrast, a phrase composed of H tones in sequence does not follow this pattern but rather maintains a level pitch (Table 2).

TABLE 2. Level H tones in succession: H tones without intervening lows remain relatively stable. Word Gloss Speaker Repetition H1-H2 avg. (Hz) H3 (Hz)

wéémé become light M2 a 136.15 136.54 b 134.55 134.96 míáká becoming rotten M1 a 138.13 139.81 b 133.01 127.16 míáká becoming rotten M3 a 161.70 160.02 b 160.68 163.57

In the data provided by these speakers, high tones with intervening lows exhibit a downtrend. Note that this is a distinct process from a downstep in successive high tones, as these H tones are not phonemically downstepped (Urua 2004). Rather, they are true H tones which show a downdrift. This downdrift, however, will not occur on H tones in certain other environments, and will even rise above the level of previous H tones.

PRE-LOW HIGH TONE RAISING

HHL words in the speaker's elicitations almost always raise the second high, rather than lowering it as expected in downtrends (Connell 2001) or keeping it at a level tone as expected when lacking a low tone (Urua 2001). This raising pattern does not appear on previously explored HHH words, in which the tones are generally stable or only slightly higher or lower than previous tones, so the raising can be attributed to the presence of the L tone.

Raised High Tones

HHL words exhibit neither the clearer terracing of LHH and HLH words nor the generally stable high tones of HHH words. Table 3 lists the pitches of the tonal segments in HHL words with the raised pre-low H2. The pitch

Proceedings of Meetings on Acoustics, Vol. 19, 060232 (2013) Page 3 A. Coombs

contour visibly rises on the second high tone before falling for the low tone (Figure 3). The final L tone shows an expected word-final falling contour (Kaufman 1968 25), which is typical of word-final but not word-internal L tones. The falling tones in these cases occur across the segments associated with the L tone, while a raised level H occurs across the segments associated with the pre-low H tone.

TABLE 3. Pre-low H raising: H2 before a L tone is raised to a higher level than the previous H tone. Word Gloss Speaker Repetition H1 (HZ) H2 (HZ) L H2 Rise (H2-H1) (Hz) ájéjìn grandchild M1 a 148.42 153.42 109.67 5.00

b 141.73 151.99 108.05 10.26 ájéjìn grandchild M2 a 131.76 133.62 99.93 1.86

b 126.11 130.28 99.93 4.17 ájéjìn grandchild M3 a 156.49 168.36 116.67 11.87

b 155.36 166.63 128.23 11.27 émánà life, M1 a 148.28 151.23 141.34 2.95 lifetime

b 141.46 148.80 113.45 7.34 émań à life, M3 a 161.69 171.39 107.47 9.70 lifetime

b 151.54 167.34 106.31 15.8

FIGURE 2. H2 raising: ájéjìn 'grandchild,' speaker M3a. The H tone rises above a previous H tone before the final L.

Speakers 1 and 3 show this raising consistently, while speaker 2 shows it inconsistently and with the smallest tone raising when it does occur. That the other two speakers show the trend consistently, however, testifies to some

Proceedings of Meetings on Acoustics, Vol. 19, 060232 (2013) Page 4 A. Coombs

raising process in their speech. This has been argued for by Urua (1996, 2001), and the data here support anticipatory raising.

The Shape of Raised High Tones

High tones generally maintain a level tone, while word-final L tones will assume a falling contour (Urua 1996). The shape of the pre-low H, however, has been described as taking an overall HL contour with the following L (Urua 2001). HHL and HHLL words differ in the data here, however, with the the HHL type taking a flat pre-low H and the HHHL type taking a falling contoured tone. The shape of the level raised H tone is pictured in Figure 3.

FIGURE 3. Level raised pre-low H: émánà , 'life, lifetime,' speaker M2a. The raised H tone does not show a falling contour, but rather the falling contour occupies the segments of the final L. This L shows an expected word-final fall.

The tonal shape then shows an initial rise to reach the tonal target on the first H tone syllable, a level raised second H tone, and a final falling L tone. The 'start up' delay on the initial tone has been previously documented, as has the pre-low raising of a high tone (Urua 2001, Urua 1996). The level shape of the H tone has been described as interacting with the following L to create a transitional HL contour (Urua 2001). The tonal picture here, though, shows this H acting as a level tone followed by a falling L, each distinctly anchored to their segmental sequences. In contrast, a HHLL tone sequence does not appear to have a level second high tone. The second H tone exhibits a falling contour (Fig. 4). The second H tone in fact begins higher than the first H tone, but drops to the level of the low as previously documented (Urua 2001).

Proceedings of Meetings on Acoustics, Vol. 19, 060232 (2013) Page 5 A. Coombs

FIGURE 4. Raised H2 falling contour: ɲ́érìjè ‘cornsilk,' M1a. The initial H segment shows a startup rise, followed by a falling contour on the raised H2. The first L tone exhibits a stable tone, while the word-final low has the expected falling contour.

The startup in initial H tone is to be expected, but the falling contour on the second high tone differs from the HHL type of pitch raising. While both type exhibit high tone raising before a low, the pre-low H of HHL is level, while the pre-low H of HHLL has a falling contour. The pre-low raising of the high tone can then occur either as a raised level tone, followed by a falling low, or as a raised falling tone, followed by a level low. The fall, then, does not spread completely across two syllabic segments. Rather, it may spread partially from one to the other in either direction. This analysis was made upon the segments to which the tones are anchored. The tonal information appears to anchor with some spreading, and as it is flattened onto the segmental sequence it may interact with or spread into neighboring segments. While final L tone lowering (to produce a contour) will occur, the contouring of the H tone may or may not. Two contours do not seem to occur in succession. The avoidance of two neighboring contour tones indicates a preference for a tone shift (appearing as a contour) of less than, but not equal to, two syllables. It appears that the contour will occur, but its domain is limited, and as a result the raised H may appear as either a contour or a level tone. Speculatively, since in Ibibio contour tones are not phonemic, avoiding an overly long contour could function to preserve contrast between the two syllable’s tones.

CONCLUSION

While words in this data set exhibit downtrends on high tones in some predictable ways, high tones may also predictably resist downtrends or even raise. Isolated sequences of H tones appear to maintain a level pitch. Furthermore, HH(H)L environments seem to show anticipatory raising. HHL environments should be tested further and with more speakers to ascertain whether the phenomenon happens consistently, and what dialectal differences exist. It seems, however, that some kind of anticipatory raising occurs with these speakers in contrast with other downtrends in the language. In addition, the raised H to L contour has a limited domain which impacts the shape, level or contour, of the raised H tone. With more data, the full picture of the contoured and level shapes in raised H tones will be investigated, as well as how downstepped H tones could be involved in pre-low raising.

ACKNOWLEDGMENTS

I would like to express my great appreciation to Dr. Christine Yu, who while a graduate student at UCLA researched tone languages as part of a cross-linguistic project for which I was a research assistant. Her knowledge and research experience were immensely instructive, and her enthusiasm and thoroughness are an inspiration to look

Proceedings of Meetings on Acoustics, Vol. 19, 060232 (2013) Page 6 A. Coombs

ever deeper into tonal processes. I would like to acknowledge the data maintained by the UCLA Phonetics Archive, which is at the heart of this study. I am also grateful to Dr. Russell Schuh for his presentations to our research group on African tonal systems in particular, and to Dr. Khalil Iskarous for his comments. Finally, I wish to thank my parents for their encouragement throughout my study.

REFERENCES

Boys, William Earl. (1979). Ibibio Phonology. (Ph.D. disseration, Ohio State University, 1979). Connell, Bruce. (2001). “Downdrift, Downstep, and Declination.” TAPS Proceedings (York University/SAOS). Kaufman, Elaine Marlowe. (1968). Ibibio Grammar. (Ph.D. dissertation, University of California, Berkelely, CA, 1968). Keating, Pat. (1984). “Ibibio.” UCLA Phonetics Lab Archive. http://archive.phonetics.ucla.edu/ (last viewed 19 Jan. 2013). Urua, Eno-Abasi. (1996). “A Phonetic Analysis of Ibibio Tones: A Preliminary Investigation.” (University of Uyo, Nigeria). Urua, Eno-Abasi. (2001). “The tone system of Ibibio.” TAPS Proceedings (York University/SAOS). Urua, Eno-Abasi, et al. (2004). Uyo Ibibio Dictionary. ABUILD Language Description Curriculum Materials #1. Course: Lexicography. Preprint draft V01 (University of Uyo, Akwa Ibom State, Nigeria).

Proceedings of Meetings on Acoustics, Vol. 19, 060232 (2013) Page 7