Wug-Testing the "Tone Circle" in Taiwanese

As a widely attested phonological phenomenon in Chinese languages, tone sandhi refers
to tonal alternations conditioned by adjacent tones or the prosodic position in which a tone
occurs. For example, in Beijing Chinese, a dipping tone 213 becomes a rising tone 35 before
another 213. Due to a long history of language change and language contact as well as the
inherent volatility of tones, patterns of tone sandhi in Chinese languages are often extremely
complex and eschew workable synchronic analyses (Chen 2000, 2004).
Taiwanese tone sandhi presents a case in point. There is a five-tone inventory on CVV
and CVN (N=nasal) syllables in Taiwanese: 55, 51, 21, 24, 33. In non-prosodic-final positions,
four of these five tones are involved in a circular chain shift as in (1). The final syllable
preserves its tone. Moretone (2004) points out that this "tone circle" presents two outstanding
problems for an output-oriented theory of phonology such as OT: first, the sandhis are
phonetically arbitrary and the unfaithful outputs do not represent clear markedness gains (e.g., 21
51); second, the circular chain shift represents a series of cases of counterfeeding opacity,
which has no account in the traditional formulation of OT (Prince and Smolensky 1993).
In this study, we followed Hsieh (1970)'s methodology to test the productivity of
Taiwanese tone sandhi using novel forms artificially introduced to the language, also known as
"wug" forms (Berko 1958). Using SuperLab, we tested nine Taiwanese subjects, who were
asked to produce two monosyllables that they heard as disyllabic words. The tone on the first
syllable was one of the five tones in the inventory, while the tone on the second syllable was kept
constant to 33. Five groups of disyllabic words were tested: AO-AO (AO=actual occurring
morpheme) where the disyllable is also a real word, AO-AO' where the disyllable is non-
occurring, AO-AG (AG=accidental gap--legal syllable and tone but non-existent combination),
AG-AO, and AG-AG. For AO-AO, the frequencies of the disyllable and first syllable were
controlled for by using a spoken-word frequency corpus by Tsay and Myers (2005). For the
other groups, we controlled for the frequency of the segmental content of the first syllable, and
we also ensured that the disyllabic combinations were not real words with any tones on them.
The experiment yielded the following three results. (a) The subjects performed
extremely poorly on sandhi patterns in the tone circle in all wug words (AO-AO', AO-AG, AG-
AO, AG-AG), and the mistakes fell into two groups: the majority represented non-application,
but a handful had a falling sandhi tone (51 or 21). (b) The subjects did significantly better with
the sandhi that did not involve opacity--24 33. In the mistakes subjects made, 21--a falling
tone--was by far the preferred choice. (c) Unsurprisingly, the subjects performed the sandhis in
real words (AO-AO) correctly in virtually all cases.
These results are significant to phonological theory in the following respects. In line with
Hsieh's results and recent psycholinguistic works on phonological opacity (e.g., Sanders 2001,
Sumner 2003), our results cast further doubt on the productivity and learnability of opaque
phonological patterns. Furthermore, given that the falling tones in Taiwanese are only about half
in duration as the other tones (Lin 1988, Peng 1997) and that non-prosodic-final syllables are
shorter than the final syllable due to final lengthening, our results suggest that speakers are
sensitive to this type of non-contrastive phonetic information and can make use of it productively
in phonological patterning. This supports the relevant of phonetics to synchronic phonology
(e.g., Boersma 1998, Steriade 1999, Zhang 2002).
Methodologically, this study significantly improves upon Hsieh's study in both data
analysis and the frequency control of the stimuli. It also solidifies a psycholinguistic method for
testing the productivity of complex phonological patterns such as Chinese tone sandhi, which can
yield further insights to both the learnability of such patterns and the relevance of phonetics to
our understanding of these patterns.

Data:

(1) Taiwanese tone sandhi in non-prosodic-final positions:
51 55 33 24

References:

Berko, Jean (1958). The child's learning of English morphology. Word 14: 150-177.
Boersma, Paul (1998). Functional phonology: Formalizing the interaction between articulatory
and perceptual drives. The Hague: Holland Academic Graphics.
Chen, Matthew Y. (2000). Tone sandhi: Patterns across Chinese dialects. Cambridge
University Press, Cambridge, UK.
Chen, Matthew Y. (2004). Changting Hakka tone sandhi: Analytical challenges. Language and
Linguistics 5: 799-820.
Hsieh, Hsin-I (1970). The psychological reality of tone sandhi rules in Taiwanese. In Papers
from the 6th Meeting of the Chicago Linguistic Society. Chicago Linguistic Society,
Chicago. 489-503.
Lin, Hwei-Bing (1988). Contextual stability of Taiwanese tones. Ph.D. dissertation, The
University of Connecticut.
Moreton, Elliot (2004). Non-computable functions in Optimality Theory. In: John J. McCarthy
(ed.), Optimality Theory in Phonology. Blackwell Publishing, Malden, MA. 141-164.
Peng, Shu-Hui (1997). Production and perception of Taiwanese tones in different tonal and
prosodic contexts. Journal of Phonetics 25: 371-400.
Prince, Alan and Paul Smolensky (1993). Optimality Theory: Constraint interaction in
generative grammar. Ms., Rutgers University and University of Colorado, Boulder.
Sanders, Nathan (2001). Preserving synchronic parallelism: Diachrony and opacity in Polish. In
Papers from the 37th Meeting of the Chicago Linguistic Society. Chicago Linguistic
Society, Chicago. 501-515.
Steriade, Donca (1999). Phonetics in phonology: The case of laryngeal neutralization. In
Matthew K. Gordon (ed.), UCLA Working Papers in Linguistics, Papers in Phonology 3:
25-146.
Sumner, Megan (2003). Testing the abstractness of phonological representations in Modern
Hebrew weak verbs. Ph.D. dissertation, State University of New York at Stony Brook.
Tsay, Jane and James Myers (2005). Taiwanese spoken corpus. National Chung Cheng
University, Taiwan.
Zhang, Jie (2002). The effects of duration and sonority on contour tone distribution. Routledge,
New York.