2.6 Two general issues in literature
2.6.2 Native phonology vs. loanword phonology
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
80
The difference is statistically significant (χ2 = 33.5, df = 2, p < .0001). We assume that it is the articulatory mechanism of fricatives that gives them a highly audible “hissing”
noise, hence liable to retention in loanword adaptation. Furthermore, the even higher salience of sibilants, a subcategory of fricatives, may exert a “masking effect” on the neighboring stop. Consider (50)14.
(50) Masking effect of sibilants in onset Sequence Item (L2 → L1)
stop-sibilant [.æl.b .t s n.] Albertson → [.ja. .pwo.s n.] 亞爾伯森
[.p i.t s .] Pizza → [.p i.sa.] 披薩
[. .b .t s n.] Robertson → [.lwo.pwo.s n.] 羅伯森 [.%w .t s .n .( .] Schwarzenegger → [. !.wa. in.k .] 史瓦辛格 [.t s .( .] Tsonga → [.so .t$ja.] 宋嘉
[.t s .] Tsar → [. a.xwa .] 沙皇
[.w .t s n.] Watson → [.xwa. .] 華生
For [t s]-onsets, the L1 adapter ignore 87.5% (7/8) of the preceding stops. It is our postulation that it is the extremely high degree of perceptual salience of the sibilant class that weakens the perceptibility of the neighboring stop.
In this subsection, we have shown that the learnability puzzle is chiefly attributed to two linguistic aspects: native phonology and perceptual salience. While in previous studies, they are proven to be true respectively in different languages, the evidence for both can be found in our collection of English loanwords in TM.
2.6.2 Native phonology vs. loanword phonology
The learnability puzzle naturally leads to another long-standing debate: should we treat native words and loanwords on a par, meaning that there is no “loanword
14 The sequence of “sibilant-stop” in onset is a different story. We will leave it to 4.9.1 for detailed discussion.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
81
phonology”, and all foreign inputs are exclusively governed by one single native grammar, or are they subject to two distinct grammatical mechanisms, and foreign inputs should pertain to “loanword phonology” of which the main ingredients are loanword-specific rules and constraints, in addition to the inevitable native structural constraints? As has been discussed in more detail in 2.2.3, Boersma and Hamann (2009) argue that loanword adaptation in Korean can be understood entirely in a single comprehension-production mechanism within the native phonology, and provide Optimality-theoretic rankings of structural and cue constraints for both native processing and loanword adaptation. This strong postulation, however, may find it difficult to analyze loanword adaptations that conflict with native alternations. Refer to the following data from Fula (Paradis and LaCharité 1997) for illustration.
(51) Fula adaptations of French loanwords a.
French Fula
carde [.kard.] [.kar.da.] ‘card (comb)’
force [.f rs.] [.f r.s .] ‘force’
b.
French Fula
table [.tabl.] [.ta .bal.] ‘table’
classe [.klas.] [.ka.la s.] ‘flag’
In Fula, consonant clusters are not allowed in either onset or coda positions. For loanwords from French, an epenthetic vowel is added after the second consonant in the sequence of “liquid-obstruent” (51a), but it is between the two consonants of the
“obstruent-liquid” sequences (51b). In the native phonology, however, the epenthetic vowel is always inserted after the second consonant in both sequences, as shown in (52).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
82
(52) Fula native phonology
a. /.talk.ru./ → [.tal.ku.ru.] ‘amulet’
b. /.sokl.ka./ → [.sok.la.ka.] ‘need’
Such a discrepancy between processes in native and loanword phonology, as can be found also in Korean (Kenstowicz and Sohn 2001) and Lama (Peperkamp 2005), poses a critical problem to works that hold similar stances like Boersma and Hamann’s. The controversy goes on, while current studies seem to lean more to treating processes in native and loanword adaptations as governed by different phonological grammars, or to the stance that at least a few loanword-specific ingredients should be added to the latter, such as Davidson and Noyer’s (1996) constraint Match, Kenstowicz’s (2005) output-output faithfulness constraints, and Yip’s (2006) constraint Mimic.
TM and English are two systematically different languages, and comparatively, TM has much stricter limitations on syllable structures. Because of this, when confronted with an English input, the TM adapter is often dealing with structures that lack evidence in the native lexicon. In our bipartite processing model, as shown in (8)/(22), the L1 adapter’s computation of an L2 input in the perceptual processing is based on a perception grammar that is independent of the native phonology, where similar representations to the foreign inputs are absent. In particular, the output of the perceptual processing, i.e. the underlying representation, then serves as the input of the production grammar, in which the interpretation involves semantic factors, which is hardly relevant to the production of a native word. For example, if the L2 input is a family name, the L1 adapter tends to choose a family name (mostly monosyllabic) as the first character, which is equivalent to a syllable in L1. This is often done at the cost of departure from the phonetic structure of the input. Consider (53).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
83
(53) Semantic effect on L1 character selection
English TM
a. [.st k.t n.] Stockton → [. .t a.k .tw n.] 史塔克頓 b. [.st k.holm.] Stockholm → [.s .t .k . .mwo.] 斯德哥爾摩
In (53a), the L2 family name Stockton ([.st k.t n.]) is transliterated as 史塔克 頓 ([. !.t a.k .tw n.]) in L1, where the first character (syllable) 史 is a well-known family name in the Chinese society, though phonetically the best choice may be 斯 ([.s!.]). In contrast, when dealing with an L2 input that is not a person’s name, 斯 becomes the most favorable choice, as in (53b), which is the transliteration of the capital city of Sweden.
English loanwords in TM have provided us with concrete evidence showing that a loanword phonology, one like (8)/(22), that is separate from the native phonology is necessary, especially when the languages involved are two systematically different languages. It is undeniable that a large part of the perception grammar is universal, since the auditory salience induced by certain interior properties and exterior contexts are phonetically based, such as the high friction noise of fricatives and the positional privilege of onsets. However, language-specific native phonology plays a crucial role in perception, as previous empirical results show that L1 speakers do have difficulty identifying structures and contrasts that are foreign to the native sound system, and thus the perceived representations are proven to abstract away from the L2 phonetic form. Moreover, the production grammar of TM loanword adaptation is another special issue. Unlike alphabetic languages, discussed in a large body of literature, adapters of a logographic language such as Mandarin will naturally consider word meaning in choosing characters in addition to phonetic similarity. This is a factor that is hardly considered in producing native words. It therefore follows that construction
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
84
of an independent loanword phonology is necessary.
2.7 Conclusion
This section has provided the background knowledge needed in understanding the focal points that will be discussed in detail in the rest of this dissertation. First, we give an overview of the different views on loanword processing, generally classified into the perception-based, production-based, and perception-production accounts.
Proponents of each stance have shed light on the processing of loanwords to some extent. Considering adapters of the loanword adaptations in our collection are not limited to pure monolinguals or bilinguals, and the fact that we are dealing with two systematically different languages, we believe a perception-production model serves as a more plausible platform for us to develop formal analyses of loanword adaptation, though our focus is laid on the construction of the perception grammar.
Next, employment of cue constraints as the main ingredients in the perception grammar endows us with an in-depth consideration of the physiological and physical properties of speech perception, as well as a more straightforward explanation for the auditory events involved in it.
With respect to theoretical machinery, like most prior literature on loanword phonology, we fundamentally develop the analyses within the OT framework. As pointed out in previous works, the key notions of constraint violability and the interaction between markedness and faithfulness constraints well model the oftentimes conflicting forces of preservation of the input information and conformity with the native phonotactic limitation. However, conventional OT is found to be theoretically untenable in dealing with our loanword data due to its lack of explanatory adequacy with the interaction between traditional markedness and faithfulness constraints and the idea of fixed constraint rankings. What makes our
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
85
analyses different, apart from replacing faithfulness constraints with cue constraints, is the way we treat lexical variation in loanword adaptation, an issue that is hardly taken seriously in literature. As we have seen, a couple of OT-based theories have been developed in order to deal with language variation. Representative works include cophonology, ROE, and stochastic OT. Given the pros and cons of each alternative, the analytical chapter of this dissertation features a stochastic evaluation of our loanword data. With its innovative revision of seeing constraints as ranges of value on a linear scale of strictness, overlapping will be incurred inevitably when the ranking values of two constraints in conflict are close enough. This overlapping area is exactly where constraint dominance can be reversed and hence language variation occurs.
Finally, the two frequently debated mysteries in the literature will be uncovered as the dissertation proceeds. First, the learnability problem is partially a reflection of language-specific facts of the native phonology and partially attributable to the more universal perceptual salience. Second, the necessity for an independent loanword phonology will be verified, as we are dealing with the contact between two systematically different languages.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
86