Interactions between Markedness and Faithfulness Constraints in Vowel Systems

Viola Giulia Miglio

• Publisher: Routledge
• Year: 2013
• eBook ISBN: 9781135456870
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

Miglio argues that to assess the relative markedness of a segment, frequency of occurrence in vowel inventories is insufficient when considered on its own. In its analysis of the Great Vowel Shift, this book elaborates a more useful model of a unitary change even in a surface-oriented theory such as optimality theory, with the help of local conjunction. Miglio extends the device of local conjunction to model opaque relations, and calls for reranking and lexicon optimization as the means to capture change within optimality theory.

MARKEDNESS AND FAITHFULNESS IN VOWEL SYSTEMS

OUTSTANDING DISSERTATIONS IN LINGUISTICS

Chapter 1

Vowel Systems: Reduction and Change

1.0 Introduction: Themes of the Book

This study is an analysis of phenomena affecting both stressed and unstressed vowel systems. The model for these systems is provided applying methods and concepts from Optimality Theory (Prince and Smolensky 1993, McCarthy and Prince 1993) and its following developments (McCarthy and Prince 1995, Beckman 1998).

1.1 Vowel Reduction

The basic idea for this book surfaced from observations on the effects of stresslessness on the structure of syllables and vowel inventories in some of the world’s languages: it is a well known phenomenon that lack of stress drives some alternations caused by the effect of lack of prominence on human perception. Such alternations seem to affect both vowels and consonants and span from the total loss of unstressed vowels, to the reduction of the vowel inventory that can appear in unstressed syllables, and to the simplification of syllable structure with loss of consonants. Some of the more complex consonant clusters, in fact, tend to be allowed cross-linguistically only in stressed syllables (Ladefoged and Maddieson 1996)¹.

Loss of unstressed vowels is a phenomenon commonly attested historically, both in Germanic languages (Campbell 1959:136 ff.) and in the Romance languages (Rohlfs 1949, Lapesa 1980, Repetti, 1996, forthcoming), to name but two well-studied examples, but the phenomenon is so widespread that it is easy to find examples from non-Indoeuropean languages. Apocope in the Melanesian language Southeast Ambryn, as well as syncope in Lenakel (both languages spoken in Vanuatu, Crowley 1997:4041) and Maltese Arabic (Hume 1994) are cases in point.

Given the common nature of these changes, it is not surprising that one should find the same kind of alternations synchronically: vowel loss, whether word-finally or medially, is in fact attested in many Northern Italian dialects (see chapter 2 below, and Repetti 1995, 1996, and forthcoming), as well as in French (Dell 1980, Tranel 1981), Mexican Spanish (Lope Blanch 1995), Modem Icelandic (Helgason 1991) to name only a few examples.

This study is therefore an attempt at a unifying view of phenomena affecting vowels systems. It is based on case studies of Romance languages and the effect that stresslessness has on the vowel systems in these languages. These case studies are synchronic in nature, i.e. the alternations analysed are still productive in these languages and they vary from complete loss of unstressed vowels to more or less drastic reduction of the vowel qualities allowed in stressless syllables. The book offers a unitary view of these reduction systems and some typological predictions.

The role of OT in modelling these systems is all-important, because it provides the possibility to unify phenomena that had been previously considered related only in so far as they occurred under lack of stress, not because of the way in which they manifested the changes. But lack of stress is merely the cause of these alternations, what needs to be modelled naturally is the effect, i.e. the alternations themselves. OT offers the way to build a unified model of these phenomena by showing that they are a consequence of the interaction between markedness and faithfulness constraints (cf. below, as well as Kirchner 1995, 1996)². Intuitively, this translates into a natural system that shows that only segments which are featurally less marked surface in stressless positions, but that it is important in each different case to evaluate how the featural characteristics of the input are preserved in the output in order to have a complete picture of the vowel system reduction.

Although the main sets of data analysed in the first part of the book are from Romance languages, the universal nature of these phenomena should not be discounted. Most Romance languages undergo some vowel reduction, either segmental or just featural. Moreover, they have reduction both as an active, synchronic process and they almost invariably have undergone reduction historically. These are the reasons why they were chosen as providing the bulk of the evidence.

However, phenomena of complete or partial reduction are extremely common in the world’s languages, independently of language family, the fact that the conclusions reached in the present study are based mostly on Romance languages does not invalidate the typological importance of the findings.

1.1.1 Change in Vowel Features and Opacity

Some languages showing a type of vowel reduction involving the change of vowel quality in stressless syllables show that one of the tendencies in reduction systems consists in allowing only vowels that are maximally distinguishable one from the other. In practice, this translates into having languages with extremely complex vowel systems in stressed environments and a simplified three-vowel system, such as an [i,a,u]-type system, in unstressed positions. This is related to a well-known phonetic phenomenon that is referred to as the “dispersion” of vowels in the phonetic vowel space (Liljenctrants and Lindblom 1972, Lindblom 1986, Flemming 1995).

Other complex vowel systems, with many distinctions, retain some of their complexity even in stressless syllables. In such environments, some distinctions are not easily perceived, and they are therefore expected to be lost. Retaining marked segments in environments where they are expected to be lost has sometimes been described in the literature as a case of opacity (Kiparsky 1968, Kenstowicz 1994:98–9 in rule-based theory, Kirchner 1996, McCarthy 1997, 1998, Karvonen and Sherman 1997, and Fukazawa 1999 in OT).

Opacity occurs when the trigger for a certain alternation between the UR and the surface form is no longer present in the surface form itself. In rule-based phonology, opacity is typically found in the surface forms resulting from counterfeeding and counterbleeding rule orders, in Kiparsky’s terms (1968).

In rule-based theory, therefore, opacity was less problematic than in a surface oriented theory such as OT (see below, paragraph 2), since it could be rendered by a rule holding of an intermediate representation that had no visible conditioning environment on the surface (see Dell 1980 for treatments of schwa deletion reminiscent of opacity).

In contemporary phonological theory, however, models or explanations are usually supported on typologically verifiable grounds. In the case of OT, the theory’s technical devices cannot hold of representations other than the surface output.

Some cases of opacity can be modelled in OT in a way that is intuitively and typologically more satisfactory than in preceding theories. If an input has a marked segment that would not normally surface under lack of stress, that segment may change features in order to become “more acceptable” in non-prominent positions (less marked), but it cannot change very radically. The surfacing correspondent of the input cannot be too different from the input itself.

I propose that a segment can change radically to surface as a less marked correspondent. That is, however, not possible in those cases where the input would be “hardly recognizable”. In such cases, the solution is a compromise: the surface output changes features a little and therefore becomes a little less marked than the input itself, but stays marked in that it does not change radically to comply with markedness requirements completely. The technical version of this compromise is provided in OT by a device called Local Conjunction (Smolensky 1993, 1995, 1997 and see some applications below and in Kirchner 1996, Fukazawa and Miglio 1996, Miglio and Fukazawa 1997, see also Kager 1999).

1.1.2 Correlates of Stress

Stress – or lack thereof – is a Leitmotiv of the book. Stress is hard to define because it is a cover term for various phonetic factors that together contribute to the impression of “prominence” that stressed syllables have: duration of the syllable, pitch and loudness are the main correlates of stress. Duration is intended as relative duration of the stressed syllable in comparison to the unstressed ones, pitch is the modification of F₀, and loudness is measured by the amplitude of the waveform: these parameters all increase in prominent syllables, thereby providing the necessary ingredients for the syllable to be set off from the others (abstracting away from secondary stress), and joining into an abstract entity called stress.

Because of the phonological relevance of stress, it has often been considered a unitary entity, and in generative grammar its has been used as a feature part of the feature matrix [+/-stress] (Chomsky and Halle 1968, Kenstowicz and Kisseberth 1979).

The interaction among these phonetic factors is complex, because each one of them can be manipulated independently for phonological or discourse purposes in language. Duration can be phonologically relevant to distinguish phonemic segments: Finnish for instance has long and short vowels as different phonemes. Pitch is used for phonemic distinctions in tone languages, and can be used as focus marker in English. Loudness can be used for emphasis. It is therefore well-known that duration will not be a reliable correlate of stress in languages where it is used to mark phonological distinctions (Lehiste 1970, Fry 1955, 1958); comparatively, pitch is not altered for stress or emphasis in tone languages (Potisuk 1996), and loudness is not always a reliable correlate of stress³.

An acoustic study of Mantuan (see chapter 4 below), one of the languages analysed phonologically provides an idea of the phonetic variations caused by stresslessness. It shows for instance that there is a tendency in the language to manipulate length to convey stress, so that unstressed syllables tend to be shorter than stressed ones, but more significantly it shows that the changes in vowel quality are clearly measurable and not an artifact. This language does display a less marked vowel system in unstressed vowels, but as always “the emergence of the unmarked” (McCarthy and Prince 1993) is not all there is to it. Faithfulness to the input has to be monitored as well, and in this case we also find segments that are highly marked, such as front rounded vowels, that surface even in stressless environments. The acoustic findings, then, not only seem to tend in the expected direction as provided by the literature, they also fit the OT analysis well, providing supporting evidence for its correctness.

1.1.3 Language Change

The final part of the book is devoted to the analysis of a historical change in English. The change analysed is known as the Great Vowel Shift (GVS) and it affected the quality of stressed long vowels towards the end of the Middle English period. The change itself can be represented as a “chain-shift”. The relevance of this last study is ...

Table of contents

• Cover
• Title Page
• Copyright Page
• Dedication
• Table of Contents
• Acknowledgements
• Chapter 1: Vowel Systems: Reduction and Change
• Chapter 2: Mantuan Vowel Deletion
• Chapter 3: Romance Vowel Reduction Phenomena
• Chapter 4: Mantuan Chain Shifts
• Chapter 5: Diachronic Chain Shifts – The Great Vowel Shift
• Chapter 6: Conclusions
• Bibliography
• Index