國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
plays a major role here is the consideration of semantic factors—as mentioned above, English Spencer is a person’s name, and the adapter chooses the character 史 ([. !.]) as the initial character in place of 斯 or 思 (both are pronounced as [.s!.]), despite the less phonetic closeness of the former. The articulated form is thus 史賓塞 ([. !.pin.s .]). Over time, when it is extensively recognized by the TM-speaking community after it is frequently used in print media or addressed in public, it enters the TM vocabulary. TM speakers thereafter use the nativized form from their mental lexicon, and no adaptation for this word happens ever after.
Of particular interest here is the construction of SPG that governs TM perception of English consonants, which is assumed to be the phase that crucially determines the retention/deletion and segmental change of an L2 segment. What this dissertation features, in comparison with other computer-based works in stochastic-OT, however, is that we figure out the mathematical axioms with which we are able to work out the precise ranking values of the constraints yielding binary variation (two variants), in accordance with the generalized patterned distribution. Within this a theoretical breakthrough is that we take a further step and explicate the mathematical operations that logically fit multiple variation (more than three variants) into the fashion of OT. It turns out to be the case that stochastic OT is no more confined to binary variation, but applicable to phonological processes that involve multiple variable outputs.
1.4 Goals and organization of the dissertation
This dissertation aims to redefine the constraint-based TM loanword phonology from a brand new angle by investigating TM loanwords from English. The focus of this research is laid on the patterned adaptation of consonantal repairs, including syllabification, deletion, and segmental change. We attempt to provide a solid ground for the observed patterns from a wide range of evidence and support from acoustic
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
18
facts, established experimental results, and widely recognized universal principles.
Differing from previous representative studies on Mandarin loanword adaptations from English (Miao 2005, Lin 2007a, 2008ab, Dong 2012), this dissertation features a sizable corpus of English loanwords in the Mandarin spoken in Taiwan, which is believed to have developed an independent loanword system after decades of separation from mainland China. Essentially, this dissertation holds a perception-production view on loanword adaptation, as sketched in (8), where perception and production, each comprising different types of constraints, play equal roles in word-loaning processes.
The traditional markedness-faithfulness interaction in classic OT as the mechanism for evaluation is discarded in the perceptual level. Instead, we formulate a series of cue constraints (Boersma 1997, 1998, 2000, 2006, 2007ab, 2008; Escudero 2005; Boersma and Hamann 2009), which interact with structural constraints (STRUCT) in perception. Viewing cue constraints as the main ingredients in the perceptual grammar may better functionally makes reference to the down-to-earth aural events that occur during each evaluation time.
In addition, the inevitable lexical variations that are widely attested in loanword adaptation but either deemed as minor exceptions or simply ignored in literature, are taken seriously in this dissertation by adopting the theoretical machinery of stochastic OT (Boersma 1997, 1998; Boersma and Hayes 2001), as this OT version is believed to better capture the insights that perception is in its nature full of uncertainty and hence only tendencies rather than absoluteness can be generalized. Employment of the abovementioned cue constraints and stochastic evaluation thus constitute the fundamentals of SPG in this dissertation. Simply put, this research intends to construct a function-oriented loanword grammar that is superior on both explanatory and descriptive grounds—it features comprehensive explanations for the patterned
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
19
consonantal adaptation from the angle of percceptual salience, along with a succinct grammar that suitably formalizes the lexical variation in loanword adaptation.
The dissertation is structured as follows. In Chapter 1, we first address the central issues that this dissertation is concerned with, followed by statements of the fundamental perspectives from which we deal with the generalized adaptation patterns, including a stochastic-OT approach to modeling variation, cue constraints as embedded in the framework of stochastic OT, and a perception-production stance on loanword adaptation.
Chapter 2 is a thorough literature review on loanword phonology and the theoretical background, where we first review a series of representative works on loanword phonology that take different views toward the word-borrowing process.
Subsequently, we give a brief introduction to cue constraints, which functions to provide auditory information in the perception grammar and should be defined as an independent constraint type. The basic ideas of OT are given afterwards, followed by a review on stochastic OT, a revised version of OT that better captures the variability in loanword adaptation.
Chapter 3 gives an overview of the phonotactics of the two languages in contact, providing the background knowledge needed in understanding how the loanword adapter’s compromise is reached by neutralizing the conflicting sounds or structures.
Introduction of the consonants includes the sound inventories and detailed comparisons in terms of voicing, aspiration, and place and manner of articulation.
Subsequently vowels are overviewed with an introduction to the monophthongs and diphthongs, followed by a sequence of comparisons in the order of high, mid, low vowels. This chapter ends with a comparative description on the syllable structures of the two languages in question.
Next in Chapter 4, we start with an introduction of the loanword corpus from
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20
which a bundle of adaptation patterns are generalized. The main body of this chapter is the presentation of the across-the-board L1 adaptation patterns of L2 consonants.
The phonologically patterned adjustments are arranged in the order of onset and coda, and the discussion within each category proceeds in the order of simplex margins and then complex margins. It will be shown that lexical variation is everywhere and nothing is for certain in the process of perception.
On the basis of the patterns that are generalized in Chapter 4, in Chapter 5 we attribute the revealed adaptation tendencies to a variety of effects that originate in perceptual salience, so as to provide a more functional account for the emergent patterns. The auditory cues that lead to the “gradience” of perceptual salience derive from either a widely approved universal scale (e.g. the sonority scale and the place-markedness hierarchy) or the acoustic qualities of the target segment (the high friction noise of fricatives and the F2 of liquids) per se. There also exist perceptual cues that are external to the target segments, such as effects of syllable positions and the similarity/dissimilarity of the neighboring sound. Essentially, this chapter along with the previous sets the stage for the analytical chapter that follows.
To model the patterned alternations of L2 consonant adaptations, as presented in Chapter 4, and meanwhile respond to the perceptual correlates that are discussed in Chapter 5, Chapter 6 gives a series of formal analyses in an attempt to integrate the perceptually-inspired processes into a single constraint-based phonology. To achieve this, first, by recruiting the fundamentals of stochastic evaluation, a piece of theoretical machinery that is superior on both descriptive and explanatory grounds is designed. The calculation excels in logically determining the ranking values of the involved constraints that are responsible for the widely attested lexical variation in loanword adaptation. Moreover, from a functional viewpoint, the perceptually based aural events are mapped onto a series of cue constraints, a constraint type that should
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
21
be defined independently from markedness and faithfulness constraints. Through the proper arrangements of their ranking values, viz. their relative positions on the continuous scale of strictness, as well as those of the TM-specific structural constraints, different degrees of overlapping are inevitably triggered, providing solid conceptual grounds for the lexical variations that are widely observable in loanword adaptation. The calculated ranking values as a whole constitute a solitary SPG that is both theoretically parsimonious and predictive, devoid of the otherwise multiple grammars that are paradoxical by nature and thus fundamentally at odds.
Finally Chapter 7 brings our research to a conclusion, which summarizes the main findings and addresses the contribution of this dissertation. In particular, a couple of questions that may pose a potential threat to the tenability of SPG are raised and answered. An issue that awaits further investigation will be put forth as the residual problem.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
22
Chapter 2
Loanword Phonology and Theoretical Background
2.1 Introduction
In the area of linguistic investigation, loanword adaptation, i.e. how the nativization of loanwords occurs, has long gained intensive attention among phonologists and sociolinguists for the past few decades. Within the scope of phonology, starting from Hyman (1970), it is generally held that in generative grammar the input that undergoes adaptation processes is constituted by the surface form of L2, and that the output is the product computed by the phonological grammar of L1. A consensus that is reached in literature is that loanword data provide a direct window through which linguists may get a glimpse of how acoustic cues are categorized with respect to the distinctive features of L1, and study the synchronic phonology of L1 by observing their adaptation processes in action. Oftentimes, the emergent patterns from loanword adaptation enable us to probe into aspects of grammatical knowledge in ways that native language data fail to. In TM, for example, dealing with loanwords from English shows aspects of TM grammar that may have remained hidden in the absence of such novel input, e.g. syllables with an obstruent coda.
However, with the theoretical development of loanword adaptation, discrepancies among different stances on a handful of relevant issues have simultaneously arisen. A mainstream issue that has led to intense debates among authors of loanword studies is the source of the adaptation tendencies, if there are any.
Specifically, on condition that most of the emergent patterns are absent in the native phonology, particularly when the languages involved are two systematically different
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
23
ones, such and English and TM, where do the adaptation patterns come from?
Proposals on this are generally divided into three divergent views. In the first, reckoned as the production-based account, a bilingual adapter stores the L2 phonetic form directly as the underlying form and adaptation is done in production with L1 phonology. The second, in contrast, is perception-based account. In this account, a foreign input is adapted early in the perception process of a monolingual adapter, and the adapted form is determined greatly by the auditory judgment of similarity between sounds. The last proposal is the perception-production view, in which both perception and production are involved in loanword adaptation, while each is governed by an independent phonological grammar.
What comes along with the primary controversy is a series of issues that are related to the above approaches. For example, the abovementioned “learnability puzzle” plausibly leads to another: should we treat native words and loanword on a par, saying that there is no “loanword phonology”, and all foreign inputs are exclusively bound by one single native grammar, or assume that they are 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 rooted native structural constraints? Another critical debate concerns the authentic form of the representations that the adaptation makes reference to: are they abstract phonological representations, fine-grained phonetic representations, or mixtures of the two? Still another question is, from a typological point of view, whether there exist any universal principles in loanword adaptation regardless of the languages involved, as those explored in L1 acquisition.
A solid construction of grammars relies heavily on a firm theoretical framework.
Due to the fundamental notions of violable constraints that suitably model the oftentimes conflicting forces of preservation of L2 input information and obedience to
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24
the sound system of L1, Optimality Theory (OT, Prince and Smolensky 1993/2004) has served as a dominant framework in formal phonological studies on loanword adaptation (Yip 1993, 2002, 2006, Paradis 1995, 1996, Kenstowicz 2003ab, Shinohara 2004ab, Labrune 2002, Kang 2003, Shih 2004, Miao 2005, Lu 2006, Lin 2007a, 2008ab, among many others). In OT, markedness and faithfulness constraints are inherently conflicting, which makes direct reference to the dilemma that the adapter may be faced with when borrowing a foreign word—compliance with the L1 sound regulations can only be done at the expense of phonetic deviation from the L2 input.
What factors may cast their influence on the retention/deletion patterns of a certain feature, segment, or in some cases, prosodic property, serves as one of the primary concerns of this dissertation. These factors should either be translated into constraint interactions or, from a more functional point of view, be straightforwardly interpreted as constraints themselves to account for phonological processes. In this dissertation, however, we side more with the latter approach for its greater explanatory power in accounting for quite a few phenomena that have more to do with the adapter’s judgment of phonetic salience and similarity.
The interaction between markedness and faithfulness may soon be considered untenable, or at least inadequate, as more loanword investigations are done. As we will see, assuming that the adapter is monolingual, adaptation is conditioned largely by perceptual cues of sounds, e.g. intensity, pitch, duration, etc. That is, under the government of L1 phonology, whether and how a sound is perceived is determined by its auditory cues, i.e. information stemming from either the external context or the internal acoustic properties of its very own. Given formalization of grammars is to interpret the speaker’s innate cognitive mechanism, we are led to replace faithfulness constraints with cue constraints (CUE), which should be defined independently from markedness and faithfulness constraints and play a role with structural constraints
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
(STRUCT) in the phase of perception. This perception grammar genuinely reflects the aural events that take place in real world and, meanwhile, better corresponds to the
“instability” during perceptual processing of sounds.
Languages are naturally variable, both diachronically and synchronically. In our investigation of TM loanwords from English, unexceptionally, variation is observable in both free variations of a single word and lexical variations across loanwords with similar input structures. A key notion of conventional OT is that while constraints are violable, constraint rankings are nevertheless fixed. Previous loanword studies in this vein either leave variations unanswered or view them as exceptional to the pattern that is under investigation, as confined by the core idea of fixed ranking. In our corpus, however, the proportion of the “irrelevant” variations is too significant to be deemed peripheral, most of which are attributable to the unstable status of perceptual processes. The need is thus felt for a feasible constraint-based framework that properly locates constraints and defines the variation in a plausible manner.
Responding to this demand, this dissertation recruits the central ideas of stochastic OT (Boersma 1997, 1998; Boersma and Hayes 2001). In this revised version of OT, universal constraints are viewed as ranges of value along a linear scale, and variation in input-output mapping is meant to happen insofar as the ranking values of two contradictory constraints are close enough to incur an overlapping area. The dominance between the two constraints is not defined and may alternate in this area.
Variation is bound to happen on condition that the two constraints militate against one another. Under this rationale, unnecessary constraint re-ranking becomes redundant and theoretical simplicity can be achieved. Meanwhile, probability of a certain adaptation strategy is well predicted based on the relative positions of the constraints at work.
The remainder of this chapter is organized as follows. Section 2.2 elaborates on
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
the diverse proposals of word-loaning processes. Section 2.3 is an introduction to cue constraints, a constraint type that should be defined independently from the traditional faithfulness and markedness constraints. Section 2.4 provides an introduction to the basic concepts of OT, and is followed by an overview of the later OT versions proposed to embed language variations in the constraint-based framework, inclusive of stochastic evaluation. Next, Section 2.6 deals with two general issues that are under heated debate in recent works on loanword adaptation. Finally, Section 2.7 concludes this chapter.