2.2 Diverse views on word-loaning processes
2.2.2 The perception-based account
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31
phonological representation. All the abovementioned authors contend that loanword adaptation is done by language-specific OT production grammars. If this is the case, the fundamental problem remains—how does the L1 speaker learn the loanword production grammar that is specific to a particular language, provided that such foreign input strings are impoverished in L1?
To summarize, authors along this line maintain that phonologically-informed adaptations can only be made in production, and that perception is nothing but a passive process. One premise that they share in common is that the adapter is a proficient bilingual, and correct perception of the L2 morphophonemic representation becomes possible. A possibility that should not be excluded in our investigation, however, is that adaptation of a certain number of loanwords may be done by a monolingual or a bilingual with lower L2 proficiency.
2.2.2 The perception-based account
Authors (Silverman 1992; Yip 1993, 2002, 2006; Peperkamp and Dupoux 2002, 2003; Kenstowicz 2001, 2003ab; Steriade 2001; Kang 2003; Broselow 2004, 2009;
Peperkamp 2005; Peperkamp, Vendelin and Nakamura 2008; Boeorsma and Hamann 2009; Calabrese 2009; Kim 2009) in this account generally adopt the second scenario, but assume that phonological modification occurs early in perception. These recent models of loanword phonology acknowledge the importance of perception in viewing aural similarity/approximation as the basis of handling loanword adaptations. In their models, the input to the adaptation is the surface acoustic representation of L2 and the similarity judgment of the adapter determines the shape of the adaptation form during perception.
As reviewed by Peperkamp and Dupoux (2003), psycholinguists show that all aspects of novel phonological structures and elements, including segments,
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
32
suprasegments and syllable phonotactics, are systematically twisted by the L1 speaker, either a monolingual or bilingual, during speech perception. Indeed, a handful of experimental data point to this striking result. For instance, French listeners have severe difficulty perceiving stress assignments (Dupoux et al. 1997) and it follows that in loanwords, stress is consistently assigned to the word-final syllable, regardless of the stress contrast in L2. Similarly, Korean listeners have trouble discriminating between English consonant liquids [r] and [l] in stimuli with the CV structure (Ingram and See-Gyoon 1998), and this result is further testified in a later observation that English word-initial [l] is adapted as [r] (Kenstowicz and Sohn 2001). Perhaps the most striking experiment result which supports the fundamental role of speech perception in loanword adaptation is the off-line phoneme detection task conducted by Dupoux et al. (1999) on the perception of illusory vowels by Japanese and French speakers. They invented a series of six nonce words that are naturally pronounced (e.g.
[.a.bu.no.], [.a.ku.mo.], [.e.bu.zo.], etc.), in which they gradually reduced the duration of the vowel [u] to zero milliseconds. The result is that Japanese speakers, unlike French speakers, overwhelmingly judged that the vowel was there at all levels of the length of [u]. This was the case 70% of the time even when the vowel is completely removed. The French participants, conversely, judged that the vowel was absent when there was no vowel around 90% of the time. These results have urged authors to conclude that illusory vowels can be introduced during perception by virtue of the influence of L1phonotactic restrictions. Moreover, as cited in Calabrese (2009), a number of researches have shown that children older than 5 or 6 and adults are
“behaviorally deafened” to foreign contrasts. For example, some Spanish dominant Spanish-Catalan bilinguals who did not learn Catalan before 5-6 years of age cannot distinguish Catalan contrasts that are not shared with Spanish (Pallier, Bosch, and Sebastián-Gallés 1997).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
33
Peperkamp, Vendelin and Nakamura (2008) observe an asymmetry in Japanese adaptations of word-final [n] in English and French loanword sources, as given below.
(10) a. Loanwords from English (Peperkamp, Vendelin and Nakamura 2008, revised)
pen → pen
walkman → wōkuman
monsoon → monsūn
b. Loanwords from French
Cannes [.kan.] → kannu
parisienne [.pa.*i.zj n.] ‘Parisian – FEM’ → parijennu terrine [.t .*in.] ‘pâté, terrine’ → terīnu
The syllable structure of Japanese is rather simple. Consonant clusters are strictly banned in syllable margins, and the few possible codas in this language include a moraic nasal consonant and the first half of a geminate. As shown in (10), loanwords from English and French are not treated alike by Japanese adapters: while a word-final nasal is adapted as a moraic nasal consonant in loanwords from English, it is adapted with a following epenthetic vowel in loanwords from French. Their experimental results show that the asymmetry is due to the different phonetic realizations of word-final nasals in these two source languages, which in turn leads to different perceptions of them by Japanese adapters. Specifically, a word-final [n] in French typically has a strong vocalic release, while one in English does not. This is the trait in French that is responsible for the assimilation to the word-final [n] with the epenthetic vowel [+] during phonetic decoding by Japanese adapters. The authors provide two other potential accounts for this phenomenon, i.e. orthographic and phonological, while they are both overruled by the reasoning that follows.
In English, word-final [n] corresponds to the last letter in its written form most often, but it is always followed by the grapheme e in French, which may seem at first
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
34
glance to be a perfect explanation for why loanwords from English are adapted with a final moraic nasal and those from French with an epenthetic vowel. However, assuming that orthography determines the adaptation, an even more suitable choice for the epenthetic vowel is [e] since it is exactly the grapheme of e in the written form, rather than [+]. In terms of this opposition proponents of this view may defend the orthographic effect by saying that a word-final [n] in French is often realized as the vowel schwa, which is phonetically close to [+]. Japanese bilinguals are aware of this correspondence between e and [ ] and thus insert the most similar [+] for the latter.
Authors against this view, however, reject this explanation since the vowel [+] in fact does not correspond to any French vowel in the spelling. In a majority of examples, it can follow a coda consonant or appear within a consonant cluster to break a “CC”
sequence. In that case, it makes little sense to attribute a single mapping between e and [+] to their correspondence while the latter is overwhelmingly used as an epenthetic vowel in a diversity of other contexts.
An alternative viewpoint is to attribute the observed asymmetry to a result of phonological mapping. As previously discussed, the adaptation of an L2 input is based on a more abstract representation by a highly proficient bilingual who has access to the underlying form of the target source word. A handful of persuasive reasons against this view on the word-final [n] asymmetry are given. A straightforward one is reviewed as follows. Previous studies indeed show that words ending in a consonant in French originate from underlying representations where this consonant precedes a schwa, whereas it is not the case in English. Differences in underlying forms of word-final [n] may again serve as a tenable account for this asymmetry. However, the authors reject this hypothesis by providing more counterexamples that are not adapted this way. For example, French words grand [.g*a,.] and petit [.p .ti.] both contain a
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
35
world-final [t] underlyingly. The “liaison” consonant should surface before vowel initial words only, as in grand arbre [.g*a,.ta*.b*.] and petit arbre [.p ti.ta*.b*.]
respectively (‘big/small tree’). Provided that it is correct that Japanese adaptations of foreign words are based on the underlying form in preference to the surface form, then we would expect the underlying consonant to be present even before a consonant-initial word. This is contrary to the fact, as examples of guranpuri from grand prix and puchipuru from petit borgeois are found.
In cases of suprasegmental adaptation, Broselow (2009) gives a perception-oriented analysis, couched within the OT model, of the stress mapping from Spanish to Huave to respond to the learnability problem of the ranking MatchStress >> TrochaicFeet and Align-R >> Free-V, as proposed earlier in Davidson and Noyer (1997), where the segmental faithfulness constraint Max is arranged in different positions in the ranking hierarchy with respect to the “stratum”
of the borrowed word, namely the most nativized, less nativized, and least nativized.
The otherwise ‘unlearnable’ constraint rankings for Huave stress adaptations from Spanish in different strata of loanwords are claimed to be a reflection of the native perception grammar, as elaborated below.
Huave has no vowel length contrast, and hence syllable weight relies completely on the presence/absence of a coda consonant. In the native vocabulary of Huave, the final syllable is stressed if it is closed whereas the penultimate is stressed when it is light. Since all stems of the major lexical categories and all suffixes end in a consonant, the predominant stress pattern in Huave is one that falls on the final syllable. For loanword adaptations in Davidson and Noyer’s (1997) demarcation of Stratum 1 (most nativized), Max ranks lowest, because in order to maintain the native final-stress pattern, segment deletion of the source is inevitable when stress of the source does not fall on the final, as illustrated in (11).
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
36
(11) Davidson and Noyer’s (1997) production grammar, Stratum 1 (revised) /.i-.(a.do./ ‘liver’ MatchStress TrochaicFeet,
Align-R
Free-V Max
a. ☞[.i-k.] ***
b. [(.i-.ka.).do.] *W L
c. [(.i.-ka.)] *W L
Broselow claims that the demarcative function of stress in Huave should provide little or no knowledge of Spanish for the L1 adapters to assume that stressed syllables are final. Rather than attribute the preservation of source stress over that of segments to the ranking of MatchStress over Max in the production grammar, she postulates a high-ranked constraint pertaining to perception grammar in (12).
(12) Perception Grammar Constraint:
AssumeLexWordEdge-V’C#: in mapping the acoustic signal to phonological representations, assume a word edge following each consonant preceded by a stressed vowel.
It is assumed in her analysis that cases like [.i-k.] do not emerge as a result of segment deletion in the production grammar; instead, they result from a “misanalysis”
of the lexical representation. That is, the first Huave adapter that heard Spanish words [.i-.(a.do.] would assume a boundary after the consonant following a stressed vowel (/.i-(./#ado), which then serve as the input to the production grammar, as sequentially illustrated in (13-14).
‧
(13) Input to perception grammar: ‘liver’ (Broselow 2009, revised) [.i-.(a.do.]
(14) Input to production grammar (Broselow 2009, revised) /.i-(./ TrochaicFeet, perceptual phase, an effect of the constraint AssumeLexWordEdge-V’C#. According to Broselow, the degree of misparsing of an L2 sound string by the L1 adapter will be reduced (entering the next stratum) and a wider range of word shapes will be allowed as more frequent contact with L2 happens.
In summary, proponents of the perceptual stance typically hold that the L1 adapter has no access to the phonological representation of L2 and that adaptation is mostly the result of the adapter’s misperception of the L2 word. It is noted that most authors holding the perceptual stance do not overrule the existence of the production level, which therefore follows that the perception-based account is not mutually exclusive with the perception-production account, as will be discussed in the following section. On the whole, authors arguing for the perceptual account do not reach a consensus on the language background of the adapter, though more work in this vein leans towards the view that the adapter knows little about L2 or is simply monolingual.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
38