From Metrical Theory to Optimality Theory

Contents

Preface

Dedication

Acknoledgements

Phonetic Transcription

Abbreviations

Introduction

Chapter One: Preliminaries
I-0- Introduction
I-1- Word Stress: General Background
I-2 Metrical Analyses of Stress in Berber
I-2-1 Adnor (1995)
I-2-2 Marouane (1997)
I-2-3 Faizi (2002)
I-3 Conclusion

Chapter Two: Distribution of Stress in AWTB
II-1 The Tamazight Dialect of El Ksiba
II-2 The Phoneme Inventory
II-2-1- The Vocalic System
II-2-2 The Consonantal System
II-3 Distribution of Stress in AWTB
II-3-1 Stress Assignment in Nouns and Adjectives
II-3-2 Stress Distribution in Verbs

Chapter Three: An Optimality-Theorectic Analysis of Stress and Extraprosodicity in AWTB
III-1 Optimality Theory
III-1-1 Structure and Principles of OT
III-1-2 The Constraints
III-1-3- The Ranking of Constraints
III-2 The Stress Pattern of Nouns
III-3 The Stress Pattern of Verbs
III- 4 Conclusion

GENERAL CONCLUSION

Bibliography

Preface

The objective of this book is to enrich the research on Berber stress by applying the OT principles to account for stress assignment in AWTB prosodic words. In fact, our aim is to show that the assumptions laid out in the non-linear metrical analyses are not completely wrong, but at the same time, that they do not provide an adequate account of the surface facts of AWTB nouns and verbs. The interaction of different constraints is sufficient to account for the location of stress in these grammatical categories. Our focus will be placed on the prosodic status of the extraprosodic affixes and clitics

Actually, one of the least studied linguistic phenomena in Berber phonology is stress. Apart from the impressionistic and linear treatments conducted by scholars who investigated the metrics of different varieties of Berber (cf. Laoust (1918 –1939), Apllegate (1958), Abdel-Massih (1968), Prasse (1972), Chami (1979), Bounfour (1985), etc.), recent studies of Berber phonology conducted within the non-linear metrical framework include Adnor (1995), Marouane (1997) and Faizi (2002)

Two notions lie at the heart of the last three metrical accounts. The first one concerns syllable weight, a decisive factor in a number of stress systems. In this respect, Marouane (op-cit.) and Faizi (op-cit.) draw a distinction between syllables headed by the epenthetic schwa, and those headed by full vowels. The second notion involves extraprosodicity. Only Adnor and Marouane make use of this concept. However, while Adnor argues that the feminine suffix /t/ and the agentive morphemes /m/ or /n/ is extrametrical, Marouane attributes this feature to all prosodic words’ final consonants

In the literature, three features function as cues for the identification of stress. These are pitch, loudness and duration. However, the presence of these correlates either individually or collectively depends on the language under study

In AWTB, the examination of the native speakers’ realisations shows that stressed syllables tend to be lengthier and louder than the unstressed ones. As a result, duration and loudness are the phonetic correlates of stress in this variety

The remainder of this book is organised as follows. Chapter I reviews the three metrical treatments mentioned above. Inherent deficiencies are pointed out. Chapter II presents a description of the stress facts of AWTB nouns and verbs. Finally, chapter III consists of an OT account of the main issues raised by the facts of AWTB

Acknowledgements

To the many friends and colleagues who have helped me understand the science of linguistics. Special thanks go to Professor Jilali Saib and John McCarthy and all the members of the laboratory “Langage et Sociét é” based in the School of Humanities, Ibn Tofail University, Kenitra

Phonetic Transcription

The transcriptions used in this book conform to the Revised International Phonetic Alphabet (IPA) except for the following symbols:

Abbildung in dieser Leseprobe nicht enthalten

Emphatic sounds are underlined

The following symbols are also used:

Abbildung in dieser Leseprobe nicht enthalten

Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

To my beloved wife Btissam and my sweet daughters Rayhana and Chada, I dedicate this work

Introduction

The main objective of this book is to account for home aspects of the prosodic phonology Ayt Wirra Tamazight Berber within the framework of Optimality Theory as conceived in Prince and Smolensky (1993) and McCarthy and Prince (1993a) and developed in the Correspondence model of McCarthy and Prince (1995, 1999) and other related works. In fact, one of the least studied linguistic phenomena in Berber phonology is stress. Apart from the impressionistic and linear treatments conducted by scholars who investigated the metrics of different varieties of Berber (cf. Laoust (1918 –1939), Apllegate (1958), Abdel-Massih (1968), Prasse (1972), Chami (1979), Bounfour (1985), etc.), recent studies of Berber phonology conducted within the non-linear metrical framework include Adnor (1995), Marouane (1997) and Faizi (2002). For this reason, the present book claims that the stress system and stress assignment are better understood as cases involving interaction between two types of conflicting universal constraints: markedness constraints and faithfulness constraints.

Two notions lie at the heart of the last three metrical accounts. The first one concerns syllable weight, a decisive factor in a number of stress systems. In this respect, Marouane (op-cit.) and Faizi (op-cit.) draw a distinction between syllables headed by the epenthetic schwa, and those headed by full vowels1. The second notion involves extraprosodicity. Only Adnor and Marouane make use of this concept. However, while the former argues that the feminine suffix /t/ and the agentive morpheme /m/ or /n/ are extrametrical, the latter attributes this feature to all prosodic words’ final consonants.

With hindsight, the objective of this book is to enrich the research on Berber stress by applying the principles of Optimality Theory (Prince and Smolensky, 1993; Kager, 1999 and McCarthy, 2000) to account for stress assignment in AWTB prosodic words. In fact, our aim is to show that the assumptions laid out in the non-linear metrical analyses are not completely wrong, but at the same time to argue that they do not provide an adequate account of the surface facts of AWTB nouns and verbs. The interaction of different constraints is sufficient to account for the location of stress in these grammatical categories. The theoretical framework herein conceptualized gives a straightforward answer to the puzzling stress system of the language which shows both trochaic and iambic feet. In particular, it is argued that the foot types that emerge depend on the hierarchical organization of prosodic words. A unitary account of the stress system is offered to the effect that both trochaic and iambic feet occur in the language. Our main focus will be placed on the prosodic status of the extraprosodic affixes and clitics.

The book provides a series of exercises that allows students not only to learn about phonology, but also to do phononological analysis.

The remainder of this book is organised as follows. Chapter one provides the unfamiliar reader with a general background on stress. In fact, it offers a definition of the term stress and its phonetic correlates as well as a review of the three metrical treatments mentioned above. Inherent deficiencies are pointed out. Chapter II presents a description of the stress facts of AWTB nouns and verbs. Finally, Chapter III consists of an OT account of the main issues raised by the facts of AWTB.

Chapter One: Preliminaries

I-0- Introduction

The main concern of this chapter is three-fold. First, we offer the reader a summary of the basic notions about stress. Second, we present three main analyses conducted within the metrical framework. Finally, we introduce the reader to Optimality theory (OT), the framework within which our study is couched.

I-1- Word Stress: General Background

In phonetics, stress refers to the degree of force that accompanies the production of a syllable. Scholars draw a distinction between stressed and unstressed syllables. The former are said to be more prominent than the latter. Here prominence may be equated with an increase in length, pitch and loudness.

Stress also plays a phonological role. In fact, it is a means to distinguish “degrees of emphasis or contrast in sentences”, argues Crystal (1985, 289). As a case of contrastive sentence stress we provide the following example: The young lady is coming. In addition, words can be classified in terms of the variation of stress they display. An example of word stress cited by Crystal (op-cit) is the pair ‘increase (noun) and in’crease (verb).

Stress languages share certain properties. Among these four are saliently known. The first one is labelled the culminative property. In fact, in some languages stress falls on a fixed syllable. For instance, in Welsh stress is most of the time assigned to the penultimate. In this respect, Kager (1999, 143) argues that “many languages impose this ‘stressability requirement’on content words (nouns, verbs, adjectives, adverbs) only”. Function words, on the other hand, do not have to be stressed.

Closely related to the notion of culminative stress is the requirement that content words should be of a certain size: two light syllables or a heavy syllable. In the literature on stress, this minimality condition is what is referred to as the foot. Moreover, if a word fails to satisfy this binarity requirement, languages resort to augmentation processes, lengthening an existing full vowel or epenthesising a vowel to the subminimal word. To illustrate, we cite these examples from Levantine Arabic and Iraqi Arabic, respectively.

1) Epenthesis driven by the minimal word

Abbildung in dieser Leseprobe nicht enthalten

The second feature of stress languages is the demarcative property. In some languages stress falls towards the edges of prosodic or morphological constituents-viz., stems, words, phrases, etc. In other words, stress is assigned to the first syllable (e.g. Pintupi, Australian); the penultimate (e.g. Warao, South American) or the ultimate (e.g. Weri, New Guinea). In AWTB, stress falls on different sites, a fact that suggests that stress is neither demarcative nor morphological. Rather, the facts suggest that it is free.

The third characteristic is called the rhythmic property. Rhythmic patterns are observed in many stress systems, with strong and weak syllables alternating at certain intervals. The strong and weak syllables form a foot. Here languages differ from each other. Indeed, some feet have the first syllable as the head, whereas in others it is the second. The former are called trochees and the latter iambs.

The fourth feature of stress languages is the property of quantity-sensitivity. Quantity refers to intrinsic prominence of syllables. If a syllable consists of a long vowel or a diphthong or is closed, it is said to be heavy. On the other hand, if a syllable is composed of a short vowel or a monophtong or is open, it is said to be light. Stress, therefore, tends to be attracted by heavy syllables. Consider the following example from the South American language Hixkaryana cited by Kager (ibid, 146).

2) Heavy syllables attracting stress

Abbildung in dieser Leseprobe nicht enthalten

Worth mentioning is the fact that if a stressed syllable fails to meet the weight requirement, the tendency is to lengthen the vowel of that same syllable. The example below hammers back this point.

3) Stress contrasts supported by lengthening

Abbildung in dieser Leseprobe nicht enthalten

Throughout this book we will show that AWTB has a free stress system that is determined by the phonological and / or the morphosyntactic shape of the formatives on which it operates.

I-2 Metrical Analyses of Stress in Berber

Earlier studies on Berber stress provide no in-depth analyses. These works devote very few pages to this issue. In this section, we discuss the works of Adnor (1995), Marouane (1997) and Faizi (2002), who provide more developed analyses of stress in different dialects of Berber. The first two dissertations deal with two varieties of Tashelhit, whereas the third presents a treatment of a Tamazight dialect. All the three studies are conducted within the Metrical Tree Model.

I-2-1 Adnor (1995)

Adnor (ibid.) is the first full-fledged metrical study devoted to stress in one of the Berber varieties, namely that of Idaw Tanane Tashelhit (henceforth ITT). Given the lack of a profound investigation of linguistic stress, this study aims at finding out the criteria that govern the distribution of this suprasegmental feature. These are: 1) rime projection, 2) voweliness and 3) final position.

The stress domain within which Adnor investigates stress assignment is the lexical word. ITT exhibits a free stress with a culminative function. Moreover, the analysis provided by this author is organised in three parts. First, it accounts for stress in segmental terms, especially the SPE’s model. Adnor quickly abandons this framework because it cannot assign stress to syllables headed by consonants. This is due to the fact that in SPE’s phonology, stress is a feature of vowels only. As a matter of fact, it is impossible to predict stress in words composed solely of consonants1 (e.g.]l.bŕs[‘leprosy’).

Accordingly, given the inadequacy of this segmental model, Adnor opts for a syllabic analysis of stress. In fact, a wide array of data is analysed, the principles governing stress assignment worked out. These principles are: 1) the Stress Scale and 2) the Stress Hiatus Prohibition.

With these constraints at hand, Adnor comes to the conclusion that ITT stress system is quantity-sensitive1. Indeed, the examination of disyllabic words shows that heavy syllables attract stress; in case no heavy syllable is available, stress falls on the last syllable having a vocalic nucleus. If the two syllables of a disyllabic word have equal weight and nuclear vocality, the final syllable gets primary stress. The following data exemplify each of the three cases discussed above (cf. Adnor (op-cit., 118-119).

Abbildung in dieser Leseprobe nicht enthalten

N.B: primary stress is indicated by an accent (‘).

The data in (4-a) show that stress can fall on a heavy syllable, be it initial or final. On the other hand, the items in (4-b) show that in words composed of two light syllables, weight is irrelevant. Instead, voweliness is the decisive factor. In fact, stress falls on the syllable with a vocalic nucleus. In (4-c) since both syllables are equal in weight and nuclear vocality, stress is assigned to the ultimate syllable. To sum up, the Stress Scale Adnor advocates operates through stages: syllable weight takes precedence over syllable vocality and final position. In the absence of syllable weight, vocality overrides final position. This is shown below:

5) Stress Scale (cf. Adnor (op-cit., 120)

Abbildung in dieser Leseprobe nicht enthalten

(Index 3 is the most susceptible for stress location).

The examination of trisyllabic words reveals that the stress assignment mechanism operates iteratively. These words consist of heavy (H) and Light (L) syllables. The classification of trisyllabics on the basis of syllable composition generates eight patterns, of which only six are attested. The remaining two combinations, namely HHH and LHH, are unattested. The data below exhibit the well-formed words in ITT (cf. Adnor (ibid, 121-127).

6)

Abbildung in dieser Leseprobe nicht enthalten

The Stress Scale is responsible for the assignment of stress to the items in (6) above. However, a literal application of this scale yields unattested forms. In fact, a word like a.rri.ráw (sort of door) gets primary stress on the ultimate syllable. Yet, the second pass of the Stress Scale assigns secondary stress on the penultimate syllable (i.e. [rri]), giving rise to a stress clash-viz *a.rrí.ráw. To solve this problem, Adnor makes appeal to the Stress Hiatus Prohibition Principle, whose function is to ban any abutting stressd syllables. This constraint is formulated as follows (cf. Adnor (op-cit, 127)

7) Stress Hiatus Prohibition

Abutting stressed syllables are banned from phonetic representations (abbreviated as…++…).

The application of this principle shifts stress to the first syllable of the word-viz., à-rri-ráw. However, though the syllabic analysis can account for the majority of ITT data, this approach merely achieves descriptive adequacy. For this reason, resort is made to the Metrical Tree model.

Metrical structure is assigned to ITT words via the erection of binary branching trees. This operation is guided by the mechanisms of node labelling, syllable pairing and directionality. Adnors’s analysis is based on the following theoretical assumptions (cf. Adnor (op-cit, 188).

8)

a) The stress pattern of ITT is characterised by binany-branching trees.
b) Stress-bearing units are unstressed underlyingly.
c) Defined relations can only hold between sister nodes.
d) There is no correlation between the s/w and + / - values assigned by stress rules.
e) At the foot level, rightward nodes are always prominent.

The principles of Stress Scale and Stress Hiatus Prohibition, along with the s/w labelling, determine the relative prominence of syllables. Thus, the words am.da “pond” and l.ħlu “apple” would be assigned the following labels (cf. Adnor (op-cit, 189).

Abbildung in dieser Leseprobe nicht enthalten

These prominence values should be preserved in the erection of metrical trees as is shown below:

Abbildung in dieser Leseprobe nicht enthalten

In the examples above, there is a direct correlation between the +/- values and the s/w notations. However, Adnor points out that sometimes there is a mismatch between the (+/- stress) and the s/w labellings. Some [+stress] syllables should be labelled as weak if they are to fit in the metrical structure. Consider the assignment of metrical structure to the following word:

Abbildung in dieser Leseprobe nicht enthalten

In the metrical presentation in (11-a) the labelling of both (+ stress) nodes as strong results in uninterpretable tree structure because the meaning of ‘weaker / stronger than the sister node’ is no longer preserved. So, the second syllable is rendered weak to avoid a stress clash.

Another issue deserving special discussion relates to pairing and the organisation of syllables into higher units. In this respect, Adnor argues that some forms allow two possible syllable pairings. Consider the following example (cf. Adnor (op-cit, 198).

12) á di.nin ‘the year before last’

Abbildung in dieser Leseprobe nicht enthalten

The author also makes a further stipulation: right nodes (at all levels) are strong. According to Adnor (ibid, 199), the following generalisation makes nodes at all levels of metrical structure left-branching.

13) In a sequence (N1; N2), N2 is always strong .

The ill-formed labelling in (12b) is thus accounted for. However, disyllabic words such as ám.da do not observe the requirement of the principle above. The syllable [ ám ] has no left constituent to be paired with. The syllable [ da ] is adjoined. The principle responsible for adjoining leftovers is formulated as follows. (Cf. Adnor (op-cit, 199).

14) Leftover Adjunction

Adjoin a leftover to the constituent on its left as a weak sister.

Therefore, the metrical tree corresponding to ám.da would be as follows :

15)

Abbildung in dieser Leseprobe nicht enthalten

The inadequacy of this account is apparent since the analysis pursued makes reference to information outside the metrical tree to determine levels of relative prominence. As a remedy, Adnor proposes to incorporate prosodic categories (σ = syllable Σ = foot and M = mot). The prosodic foot is integrated in the tree structure to the effect that a primary stressed syllable is the head of a foot. It should also be specified that the foot in this variety consists of a stressed syllable preceded by zero or one unstressed syllable. The metrical tree in (16a) below would be transformed into (16b).

Abbildung in dieser Leseprobe nicht enthalten

The use of prosodic categories instead of segmental properties serves to indicate both constituency and prominence.

As far as the interaction between stress assignment and affixational morphology is concerned, Adnor argues that the feminine suffix [ t ] and the agentive morpheme [ m ] or [ n ] should be treated as stress inert (i.e. extrametrical). For example, the feminine form ti .fi. flt gets the following structure (cf. Adnor (op-cit, 204.)

Abbildung in dieser Leseprobe nicht enthalten

The origin of the ill-formed structure in (17) is traceable to the inability of the feminine suffix to make the ultimate syllable heavy. Thus the form in (17) should be turned into (18) below.

Abbildung in dieser Leseprobe nicht enthalten

As for the agentive morpheme, Adnor resorts to diacritic marking, another instance of extrametricality (cf. Hayes, 1981). The word im.žli ‘the lost’ is assigned the following representation:

Abbildung in dieser Leseprobe nicht enthalten

Before turning to the second contribution to the study of Berber stress, it should be mentioned that Adnor’s analysis has many limitations. To start with, the assumption that vowelless syllables can bear stress is hard to maintain. In fact, given that stress correlates with duration, stress bearing consonants fail to lengthen. According to Saib (1993), this represents a strong argument against their being treated as true nuclei.

Another drawback of this analysis is due to the nature of the framework adopted, namely the Tree-only model. Adnor argues that stress in ITT is analysed in terms of right-headed binany branching (i.e. iambic) feet, which are constructed from right to left. However, in words such as am.da, the final syllable is adjoined to the first syllable, thus creating a left-branching tree. This means that the resulting foot is trochaic. Adnors’s analysis fails to account for the existence of iambic and trochaic feet in ITT phonological system. This apparent paradox is hard to explain within a parameterised theory which forces a language to choose between iambs or trochees. As we will show, an OT approach differs from parameterised theories in allowing for iambs and trochees in the same phonological system. Finally, Adnor’s analysis does not account for the interaction between foot types, the Stress Scale and the Stress Hiatus Prohibition principles. It is our belief that an OT constraint-based analysis can succinctly catch this constraint interplay.

I-2-2 Marouane (1997)

Marouane (1997) is the second study devoted to the stress system of another variety of Tashlhit- namely Ayt Souab Tashelhit Berber (hereafter ASTB). A survey of Marouane’s analysis is necessary if we are to know the peculiarities of this dialect.

The classification of ASTB phonological words is based on the number and type of syllables they comprise. The stress system of this dialect is quantity-sensitive. A heavy syllable would attract stress; in case all the syllables are equal in weight, primary stress falls on the initial position.

Marouane incorporates the mora (u) as a weight unit to capture the quantity of syllables. Thus a light syllable corresponds to one mora while a heavy syllable to two-viz.,:

20) Syllable types in ASTB

Abbildung in dieser Leseprobe nicht enthalten

(Ç is a syllabic consonant and e a schwa.)

In his analysis, Marouane integrates prosodic categories larger than the syllable. Primary stress is represented in terms of a left-headed branching tree. Consider the structure in (21).

21)

Abbildung in dieser Leseprobe nicht enthalten

The same metrical structure is used to represent secondary stress, (i.e. a weak foot). This is shown below:

Abbildung in dieser Leseprobe nicht enthalten

Given that the stress system of ASTB is quantity-sensitive, the stress algorithm needs to have access to the moraic composition of syllables. The rule that assigns a strong foot to a syllable with maximal moraic content is given in (23).

23) Foot-to-Mora Mapping

Abbildung in dieser Leseprobe nicht enthalten

Following the theoretical assumptions laid out in Giegerich (1985), Marouane argues that a monosyllabic word is assigned the following metrical structure:

24) af ‘find’

Abbildung in dieser Leseprobe nicht enthalten

The metrical analysis advocated by Marouane allows only for a disyllabic foot. However, to satisfy this requirement a monosyllabic foot is assigned a structure where the left constituent is a zero syllable.

As far as disyllabic words are concerned, the author argues that if the two syllables are equal in weight, a left-dominant binary branching foot is erected. Consider the following representation:

25) W

Abbildung in dieser Leseprobe nicht enthalten

Two principles are responsible for the foot structure assignment. These are: a) the Strong Foot Assignment Principle and b) the Pairing Principle. These are formulated as follows. (cf. Marouane (op-cit, 245).

26)

a) The Strong Foot Assignment Principle

Assign the head of a foot to the leftmost syllable .

b) The Pairing Principle

Pair the syllable heading a foot with the syllable to the right.

The literal application of the principles above leaves the leftmost weak syllable stray when the ultimate syllable is the heaviest. This state of affairs is exemplified in (27).

27)

Abbildung in dieser Leseprobe nicht enthalten

This led Marouane to revise the principle in (26) to include the provision that heavy syllables attract stress more than the leftmost ones (cf. Marouane (op-cit, 247).

28) The Strong Foot Assignment Principle (revised)

Assign the head of a foot to the heaviest syllable; otherwise assign it to the initial syllable.

To ensure a full parsing of all the syllables into higher prosodic structures, Marouane stipulates that the initial stray syllable should be adjoined to the prosodic word node. This new provision is represented in the form of a principle, as shown below:

29) Stray Weak Syllable Adjunction

A syllable which occurs to the left of a strong foot is adjoined to the W-node.

The form in (27) becomes as in (30) below.

30) W

Abbildung in dieser Leseprobe nicht enthalten

This situation is reminiscent of Adnor’s treatment of disyllabic words where the rightmost stray syllable gets adjoined to the leftmost heavy syllable. This adjunction is dictated by the need to stick to the requirements of the metrical tree model, which allows for a trochaic or an iambic foot only. In addition, since the leftmost stray light syllable is equal in weight to a monosyllabic word (e.g. su ‘drink’), the two get different treatments. Finally, it should be noted that the Stray Weak Syllable Adjunction Principle is in conflict with a well-rooted principle of metrical theory –namely selkirk’s (1984a, 26) Strict Layer Hypothesis2-viz.,:

31) Strict Layer Hypothesis

A category of level I in the hierarchy immediately dominates (a sequence of) categories of level I-1.

Accordingly, Marouane’s analysis suffers from many inconsistencies. The treatment of trisyllabic words is done in the light of the principles outlined above. The data presented below stand for all attested combinations of heavy and light syllables in ASTB-viz.:

32)

Abbildung in dieser Leseprobe nicht enthalten

An LLL item such as í.ni.gi is assigned a foot on the first two syllables. This foot is headed by the initial syllable. This is shown in (33).

33)

Abbildung in dieser Leseprobe nicht enthalten

The last syllable is left astray. It can either adjoin to the metrical structure or it can form an independent foot. Still, as a certain degree of stress is perceived on the ultimate syllable, this one is assigned the head of a weak foot by the following principle:

34) Weak Foot Assignment Principle (cf. Marouane op-cit, 250)

Assign a weak foot to the syllable to the right of a binary branching foot.

This principle allows for the integration of the ultimate syllable in (33) above. See below.

Abbildung in dieser Leseprobe nicht enthalten

In case the trisyllablic item has a HHL structure (e.g. amg. gar. zu), the initial syllable, which is expected to get secondary stress is destressed (i.e. weakened) so as to avoid a stress clash. Marouane (ibid.) stipulates that the following principle militates against the cooccurrence of two adjacent stressed syllables (cf. Marouane (op-cit, 251).

36) Stress Clash Prohibition Constraint

Abbildung in dieser Leseprobe nicht enthalten

The above principle makes the initial syllable in amg.gar.zu weak to avoid clash-viz.:

Abbildung in dieser Leseprobe nicht enthalten

The same structure is assigned to items with a LHL pattern (e.g. a.mud.du). However, words with an LHH shape (like ta. ful.lust) get stress on the final syllable while the first two syllables constitute a weak foot. The metrical tree corresponding to this word is given below. (Cf. Marouane (op-cit, 253).

Abbildung in dieser Leseprobe nicht enthalten

Marouane considers the above foot structure the result of a right-to-left directionality of parsing. On the other hand, the above metrical tree can be the result of a conflict between two constraints, one preferring heavy syllables to head feet and the other prohibiting two abutting stressed syllables. This constraint conflict cannot be expressed in the serial non-linear metrical model adopted by Marouane. Only a constraint-based analysis can account for this constraint interaction.

The other main contribution of Marouane’s analysis is the incorporation of the extrametricality mechanism to account for why certain word-final syllables do not exhibit equal weight as their word-internal counterparts, even if the two may have the same segmental composition. Marouane considers word-final consonants in nouns and verbs alike to be extraprosodic (Cf. Marouane (ibid, 257).

39)

Abbildung in dieser Leseprobe nicht enthalten

The items above suggest that a (C) VC syllable is less prominent when occurring in final position. If the ultimate syllable in am.zil were heavy, it would get default stress, i.e. *am.zíl. The account offered by Marouane is that the last consonant should be considered extraprosodic. However, extrametricality is blocked in the case of monosyllabic words such as fel ‘leave’ and zer ‘see’.

With hindsight, Marouane argues that if the last consonant of a monosyllabic were rendered extrametrical, schwa insertion would be blocked, since schwa cannot occur in an open syllable. Yet, it should be mentioned that Marouane draws no difference between extrasyllabicity and extrametricality. Following Goldsmith (1990), a consonant can be parsed within a syllable, but behaves as extrametrical for the sake of stress assignment. Moreover, nothing is said about the adjunction site of the extrametrical consonant.

In addition, as far as the interaction between stress and affixational morphology is concerned, Marouane argues that the feminine suffix is extrametrical. In fact, it fails to shift stress when attached to a light syllable. On the other hand, it can cause a stress movement when it is adjoined to a heavy syllable. Compare íf.ri/ tífri (t) to áħru(y)/ta.ħrúy(t). Recall that Marouane considers Ç, CÇ, C eC and CV light syllables. As for heavy syllables, they are of the shape CeCC, CVC and CVCC.

Similarly, the adjunction of the plural morphemes and the demonstrative clitics may cause a stress shift. Consider the data below. (Cf. Marouane , ibid. 269-271).

Abbildung in dieser Leseprobe nicht enthalten

A close observation of the plural forms in (40a) shows that stress rules apply iteratively after each affixation process. In other words, the affixed segmental material is integrated into the prosodic structure.

To sum up, having provided a fairly detailed review of Marouane’s analysis of ATSB stress system, we should point out that Marouane considers the stress assignment in nouns and verbs to be the same. We will show that this position is a bit equivocal. In addition, his analysis shows that the different principles worked out conflict with each other, a fact that cannot be explicitly expressed in the adopted tree model. Only a constraint- based approach can.

In the next subsection, we move to pinning down Faizi’s metrical analysis of a Tamazight variety, namely that of Goulmima Tamazight Berber (GTB).

I-2-3 Faizi (2002)

Based on the location of primary stress in GTB nouns and adjectives, Faizi (ibid, 216) argues that two criteria are responsible for the assignment of this linguistic feature in this variety. These are: 1) the weight criterion, which assigns primary stress to a heavy syllable and 2) the default stress assignment, which assigns stress to the initial syllable.

According to the first criterion, a heavy syllable is of the shape CVCC or CVC, where V is a full vowel. Light syllables, on the other hand, include V.CV, CeC or CeCC. The combination of light and heavy syllables yields the following patterns:

41)

Abbildung in dieser Leseprobe nicht enthalten

A close observation of the items above suggests that heavy syllables tend to attract stress. In case the word contains two or more heavy syllables, it is the rightmost one headed by a full vowel that gets primary stress. When only light syllables are available, the initial one is stressed by default. In addition, Faizi points out that though a CeC(C) is heavy, it behaves as a light syllable.

As mentioned before, stress can fall on different positions: penultimate, ultimate, antepenultimate and initial. This means that the stress system of GTB is unbounded and quantity-sensitive. Accordingly, penultimate stress is assigned to disyllabic items with the shape LL or HL and to trisyllabic words of the form LHL or HHL. Further, Faizi (ibid.) stipulates that as syllables should be grouped into larger prosodic categories (e.g. foot), a left-headed binary tree is built. Consider how the foot structure is assigned to the items below (cf. Faizi (ibid, 217).

Abbildung in dieser Leseprobe nicht enthalten

The inclusion of the metrical foot expresses the prominence relations holding between syllables. As such, the rule responsible for the construction of this prosodic category is formulated as follows:

43) Foot Construction Rule (FCR) (Preliminary version)

Assign a binary left-headed foot structure to GTB nouns and adjectives.

On the other hand, trisyllabic words are assigned a metrical structure different from that of disyllabics. Since these items have an odd number of syllables, one syllable would be left over. However, given the right-to-left directionality of parsing attested patterns are yielded. The incorporation of the directionality parameter led Faizi to revise the initial version of the FCR- viz.:

44) Foot construction Rule (Revised) (cf. Faizi (op-cit, 220)

Going from right to left across the word, assign binary left-headed foot structure to GTB nouns and adjectives.

The application of the new version of the FCR to items such as aγuyyi yields the following structure. (Cf. Faizi, op-cit, 220).

Abbildung in dieser Leseprobe nicht enthalten

Further, since all syllables should be integrated into higher constituents, the leftover syllables can either head an independent foot or be adjoined to the prosodic word node (M). The first alternative is to be abandoned since the metrical tree model (i.e. Giegerich (1985) adopted by Faizi allows only for binary branching feet. Therefore, the initial syllables in the words in (45) above cannot constitute independent feet on their own. For this reason, the second option is adopted. The rule responsible for adjoining leftovers is formulated by Faizi (op-cit, 221) as follows:

46) Word Construction Rule (WCR)

Construct a word-level tree to gather up all feet and stray syllables.

The application of the above principle to the forms in (45) yields the following metrical structures:

Abbildung in dieser Leseprobe nicht enthalten

N.B: M stands for Mot.

Faizi (ibid. 222) argues that since metrical trees should express prominence relations that hold between pairs of nodes, he stipulates a Word Rule (WR). This is given below.

48) Word Rule (WR)

In a pair of nodes (N1, N2), N2 is strong iff it branches.

The rule above transforms the structures in (47) into those in (49) below.

Abbildung in dieser Leseprobe nicht enthalten

The provision expressed by the WR is incorporated in the WCR (cf. Faizi (op-cit, 224).

50) WCR (Final version)

Construct a right-dominant word-level tree together up all feet and stray syllables.

On the other hand, the rule above cannot account for the stress patterns of words with ultimate stress (e.g. i.mál ‘next year’). In fact, the construction of a right-dominant foot structure should be abandoned, since the framework adopted by Faizi allows for iambic or trochaic feet only (cf. Angoujard, 1990). The alternative solution advocated in this work is to allow for the erection of a binary foot word-finally, dominating an empty weak node. The tree structure of the word i. mal is given below (cf. Faizi (op-cit, 228).

Abbildung in dieser Leseprobe nicht enthalten

The incorporation of the weak empty node is required by the fact that the smallest foot made available by the framework advocated is one that dominates at least two syllables.

Returning to words with antepenultimate stress, Faizi argues that a trisyllabic foot is assigned to words comprising three syllables. Consider the metrical tree of the word ta.der.si ‘scarcity’ (cf. Faizi (op-cit, 232)

Abbildung in dieser Leseprobe nicht enthalten

The FCR and the WCR do not apply at this stage. For him, the structure above is sufficient to account for the stress pattern attested in this position. Moreover, a quadrisyllabic foot is assigned to words that are composed of four syllables– see below:

Abbildung in dieser Leseprobe nicht enthalten

As far as the stress assignment in morphologically complex words is concerned, Faizi argues that if the possessive demonstrative clitic corresponds to a heavy syllable, it attracts stress. However, if it constitutes a light syllable, no stress shift is witnessed. The rules developed so far are able to derive the metrical structure of these morphologically-complex words-viz.:

Abbildung in dieser Leseprobe nicht enthalten

Before evaluating Faizi’s work, it should be noted that the assignment of stress to morphologically complex words takes place after the affixational processes. In addition, this author handles cases of phrasal stress. These shall not be reviewed here since the purpose of the present chapter is to see how stress is assigned to lexical words which may or may not contain bound morphemes (i.e. affixes).

To sum up, Faizi is credited for his analysis of stress in a Tamazight dialect close to ours. However, though his analysis can assign stress to GTB nouns and adjectives, he is forced in integrate empty constituents to preserve the FCR, the rule which builds trochaic feet. Further, initial light and heavy syllables in the forms in (47) above are adjoined as weak constituents to the prosodic word. No full-fledged account is put forward to explain the non-suitability of the initial syllables to head an independent foot. In other words, unlike Adnor or Marouane who attribute this failure (in the case of the presence of two contiguous heavy syllables) to the avoidance of stress clash, Faizi makes no such provision. Moreover, unlike the first two authors who maintain that the feminine suffix is extraposodic, Faizi argues that this morpheme contributes to the weight content of the syllables it adjoins to.

I-3 Conclusion

The review of the literature on Berber stress within the Metrical Tree model shows that the proposed analyses suffer from many inadequacies that are specific to standard non-linear approaches. The conflicting relationships that exist between the different principles at play are not given expression. It is our belief that only an OT constraint-based analysis can catch these constraint interactions.

In the next chapter, we turn to providing a description of AWTB and the distribution of stress in nouns and verbs, focusing on the status of extraposodic suffixes.

Practice

1- Apply a metrical analysis to the following data from Ait Wirra Tamazight Berber.

Abbildung in dieser Leseprobe nicht enthalten

2- Syllable weight influences how stress feet are assigned. Quantity-sensitivity is the term used to refer to this criterion of stress distribution. Languages vary in relation to the syllable weight.

a) Q-sentisitive lanaguages: heavy syllables get stressed. English is Q-sensitive: Light penult: stress goes to preceding syllable: Ex. 'Canada, 'metrical, 'visible, 'ultimate.

b) Heavy penult: gets the stress: Ex: A'genda, ho'rizon, de'cided, 'mango Q-insensitive language: heavy syllables may occur in stressless position. French is Q-insensitive: Ex. Che'val, carna'val, Es'pagne

Find examples of languages that belong to the two systems and then provide a metrical analysis.

3-English Words are built from three types of feet.

- binary (trochaic) containing a strong then a weak syllable, eg "elbow"

-ternary containing a strong followed by two weak syllables, eg "oxygen"

-non-branching containing a single strong syllable, eg "cat"

Words can be built by combining sequentially the above feet, or indeed the feet with themselves.

Examine the following data from English and decide the number and types of feet:

- economics
-abracadabra
- matador
-abolishment
-antidisestablishmentarianism

4- Draw a prosodic word tree for the following words from English:

-pyrotechnic
-razor
-psychometric
-cardinal
-hydrogene
- nomenclature

5- Foot Types

To find out if you can remember the types of feet, Seamus Cooney suggests Coleridge's mnemonic lines to help you with the names of the feet.The main ones are listed below, using a hyphen to show a weakly stressed syllable and a slash to show a strongly stressed syllable:

Abbildung in dieser Leseprobe nicht enthalten

Trochee trips from long to short.

From long to long in solemn sort.

Slow Spondee stalks, strong foot!, yea ill-able.

Ever to keep up with Dactyl's trisyllable.

Iambics march from short to long,

With a leap and a bound the swift Anapests throng.

In the following lines which kind of foot predominates? (I am assuming that the opening words are stressed "Do you" and not "Do you.)".

Do you remember an inn, Miranda?

Do you remember an inn?

Iambs- trochees- anapest- spondees- dactyls

When you combine three iambs, you can create in poetry iambic trimeter. Parse the examples below into iambic trimester (http://examples.yourdictionary.com/examples-of-iambic.html#yDXDHBXkSVrPGBlR.99)

- Emily Dickinson

When I was one-and-twenty

I heard a wise man say,

'Give crowns and pounds and guineas

But not your heart away;

When you combine three trochees, you can create in poetry a trochaic trimeter. Parse the poem from Edna St. Vincent Millay into trochaic trimeters.

Sorrow like a ceaseless rain

Beats upon my heart.

People twist and scream in pain,—

Dawn will find them still again;

This has neither wax nor wane,

Neither stop nor start.

Chapter Two: Distribution of Stress in AWTB

II-1 The Tamazight Dialect of El Ksiba

The aim of this subsection is to present a geographical and primarily a linguistic delimitation of AWTB. This dialect is spoken in and around El Ksiba, a small town fifty kilometers away at the north of Beni-Mellal. If we take into consideration the sociolinguistic classification provided by Boukous (1972, 11), AWTB will be arranged within the Northern Berber subgroup characterised by the existence of spirants, dentals and affricates in its consonantal system (cf. Oussikoum 1990-1995).

It should be pointed out that the variety of AWTB dealt with in this thesis is the one spoken in El Ksiba and referred to by the common people as ‘Tamazight El Qsebt’ (i.e. the Tamazight of El Ksiba). AWTB is situated on the borderline between two Berber dialect clusters- viz.,: The Northern and the Southern. Therefore, AWTB is expected to manifest traits of both groups. In fact, as far as the consonantal system is concerned, the focal dialect, like dialects of the south, lacks affricates2 and like dialects of the north, it has spirants and dentals. Still,, we consider AWTB a northern dialect because it realises schwa in its phonetic representations.

In comparison to the other subvarieties spoken in the neighbouring areas (e.g. Aflla Nifran, Imhiwash, Sarif, etc.), AWTB emerges as the central system. Indeed, affricates, the distinguishing phonetic feature between AWTB and the surrounding subdialects, can be derived via the affrication of geminates. Examples are supplied for illustration:

1) AWTB Other surrounding subdialects

[asekka] ‘tomorrow’ [asetša]

[eggeεδ] ‘climb’ [edžeεδ]

Moreover, AWTB is seen as being the ‘legitimate norm’ (cf. Boukous (1985, p 86) (Translation mine). Indeed, the neighbouring subvarieties are considered by AWTB speakers as rural. Affrication is considered a pertinent sociolinguistic feature that helps localise the place of origin of its users. Affrication is then a socially-relevant linguistic idiosyncrasy (Boukous, ibid; Labov, 1972).

Being the main commercial center in the area of Ait Wirra, El Ksiba has attracted people from Arabic-speaking areas. Characteristic of all contact situations, mutual borrowing between Moroccan Arabic (henceforth MA) and AWTB has resulted in the enrichment of the vocabulary as well as the consonantal system of the variety under study. The loanwords from Arabic have undergone changes in accordance with the morphonological rules of AWTB. Consider the following examples:

2) MA AWTB

[nežžar] ‘carpenter’ [anežžar]

[serwal] ‘Trousers’ [asrawel]

The Arabic loanwords in (2) start with the vowel [a], a characteristic landmark of Tamazight masculine nouns.

Like all Berber dialects, AWTB consonantal system has borrowed MA emphatic consonants-namely /ss/ /l/ and /ll/. Likewise, recent loandwords contain stops that escape spirantisation (e.g. [elkas] ‘cup’; [elkar] ‘bus’).

To sum up, the Tamazight dialect of El Ksiba is an amalgamated variety characterised by numerous borrowings from MA. Nonetheless, we conceive of it as being a homogeneous entity.

II-2 The Phoneme Inventory

Being a member of the Hamito-Semitic cluster, Berber has a poor vocalic system and a rich consonantal one (see Cohen (1988); Louali (1999)). In the following subsection, we try to uncover the peculiarities of AWTB segmental inventory.

II-2-1- The Vocalic System

As indicated above, Tamazight vocalic system is one of the simplest. In fact, like all Berber dialects, AWTB comprises three full vowels - viz.: a i and u. These vowels have phonetic variants attested exclusively in the vicinity of emphatic consonants. The [+constricted pharynx] feature is not to be included in the underlying representation of vowels, for the rule to derive the emphatic counterparts is simple and possible. To explain, /a/ becomes more back [æ] while /i/ and /u/ become more open [ε] and [ɔ] respectively. Illustrative examples are given below. (Underlining indicates emphasis).

3) [attæn] ‘meningitis’

Abbildung in dieser Leseprobe nicht enthalten

It should be pointed out that no length contrasts are witnessed except in relation to stress or intonation. In heavy stressed syllables, these full vowels would be perceived as being longer than those occurring in unstressed syllables. Thus, stress may determine what can be called ‘expressive vowel length’.

Like all Northern Tamazight dialects, a fourth vowel, schwa, is frequently attested in the pronunciations of native speakers. This schwa vowel (transcribed throughout this book as [e]), has been the subject of controversy between Berberists from different trends: Traditionalists (Renisio, 1932; Basset 1946; 1952); Functionalists (Chaker, 1983; Chami, 1979); Generativists (Saib, 1976- 1978-1993); Boukous, 1987); Dell and El Medlaoui, 1985-1988) (see Idrissi for a review of all the approaches).

To sum up, we have seen that the vocalic system of AWTB is similar to that of many Berber dialects. It comprises three full vowels and a reduced one, schwa. It will be shown in subsequent sections that this distinction is relevant to our discussion of stress distribution.

II-2-2 The Consonantal System

Tamazight has a rich consonantal system. The consonantal chart of AWTB is given below.

4) Labial : b, f, m, w

Abbildung in dieser Leseprobe nicht enthalten

Unlike the vocalic system, the AWTB consonantal inventory is richer. Only stops are posited as underlying. The reason for this is that the spirant counterparts can be derived by rule. This is supported by the claim that native speakers can make generalisations about the grammar of their language. As for the loanwords that resist spirantisation, these can be marked as [-spirantisation rule].

This would give the impression that the grammar would look more complex as exceptions are marked by a diacritic feature. However, taking stops as underlying and spirants as derived simplifies the statement of many rules of the grammar.

Moreover, the inventory in (4) includes sounds that are not native to AWTB –viz.: h, ss and l. These have been introduced into AWTB through borrowings from MA (e.g. [essehδ]‘ heat’; [ullah] ‘I swear in the name of Allah’). However, as these sounds occur in a substantial vocabulary, native speakers are unable to draw a distinction between native and non-native phonemes. Thus, we would consider them as making part of AWTB consonantal system. Finally, it should be mentioned that all the consonants have a geminate counterpart except the laryngeal /h/.

In the next section, we move to pinning down the stress facts of AWTB. This step is necessary if we are to understand the peculiarities of the stress system of the focal dialect.

II-3 Distribution of Stress in AWTB

This section is intended to lay down the background for the analysis of AWTB stress to be presented within the OT framework. To achieve this goal, this section is divided into two subsections, one for nouns (and adjectives) and the other for verbs. This organisation is required because it will help the reader recognise easily the effects of morphosyntactic factors on the distribution of stress in different grammatical categories.

Worth mentioning is the fact that a rigorous study of Berber stress system and extraprosodicity rests on the collection of a credible corpus and the elaboration of a methodology aiming at providing bonafide data. In this respect, the first step of data collection consisted in collecting all the feminine forms and the verb-clitic combinations we know of, being quasi-native speaker of AWTB. Desirous of collecting as many forms as possible, we consulted Oussikoum’s “Tamazight-French Dictionary: The Dialect of Ait Wirra-Middle Atlas (Morocco). Through this second step of the process of data collection we had access to a large body of data, meaning and phonetic transcriptions included. This dictionary supplies feminine forms with three consonants word-finally. Oussikoum’s transcriptions have a schwa between the feminine suffix and the preceding consonantal cluster. Being suspicious of the transcriptions provided by this author, we presented native speakers of AWTB with a list of the feminine forms as well as verb-clitic constructions. The native speakers, with different ages and sex and born and grown up in El Ksiba, were asked to pronounce the words in question with a slow (normal or careful) rate of speech. The recorded pronounced forms had no schwa inserted between the feminine suffix or clitic and the preceeding consonantal clusters. This was the third step in the process of data collection.

The methodology pursued in this book has at least three advantages: First, to get a recorded data of a high-enough quality while at the same time tapping the speakers’ most natural (careful) speech; second, to verify the existence or the absence of schwa in the forms that contain sequences of three consonants word-finally; and third, to limit any pronunciation variation that could arise as the result of interference from Moroccan Arabic or the surrounding dialects.

Since one of the ambitious goals of the present study is to determine the metrical status of extrasyllabic affixes and clitics, the recorded data provide a means to extract the stress patterning of AWTB. Furthermore, we distributed lists of the forms containing the feminine suffix and the verbal clitics to university students living in El Ksiba and speaking AWTB. These students were asked to assign stress to the forms in the lists.

In short, the choice of this particular socio-linguistic milieu has the advantage of restricting the field of research, hence reaching a certain degree of homogeneity.

II-3-1 Stress Assignment in Nouns and Adjectives

It should be noted that the items were chosen in such a way as to include all syllable types-namely V, CV, CVC, VCC, CVCC, eC(C), CeC and CeCC. The words in question contain the different syllables in different positions. The reason for this is to see whether stress is sensitive to weight and if it is prominence-driven, as claimed by Adnor (1995), Marouane (1997) and Faizi (2002). The last term, prominence-driven, simply means that stress is sensitive to the nature and quality of the vowel that constitutes the nucleus of the syllable.

The list of items includes monosyllabic, disyllabic, trisyllabic and quadrisyllabic lexical words. The letters H and L are used to indicate heavy and light syllables, respectively. For reasons that will be presented later, a heavy syllable corresponds to CVC, CVCC and VC(C), where V is a full vowel (i.e. a, u or i), whereas a light syllable comprises V, CeC, (C)eC(C) and eC.

II-3-1-1 Monosyllabic words

Abbildung in dieser Leseprobe nicht enthalten

N.B.:

a) The accent (‘) indicates primary stress
b) Only broad transcriptions are supplied.

Given that the items in (5) are monosyllabic, stress falls on the only syllable that is available.

II-3-1-2 Disyllabic words

The disyllabic collection presents a richer syllabic parttern because all possible combinations of H and L syllables are attested. The data below exemplify each combination.

Abbildung in dieser Leseprobe nicht enthalten

In the items above, stress is assigned to the initial syllable. This amounts to saying that when the two syllables of a disyllabic word are light, stress is assigned to the initial one.

As opposed to items in (6a), the ones given in (6b) below contain two syllables with distinct weight contents-viz.:

Abbildung in dieser Leseprobe nicht enthalten

The items above show that stress falls on the heavy syllable. This means that AWTB makes recourse to syllabic weight in the assignment of stress. This fact is further corroborated by the stress parttern of the items in (6c) below.

Abbildung in dieser Leseprobe nicht enthalten

Since the initial syllable in the words above is more prominent than the ultimate one, it attracts stress.

In strings made up of two heavy syllables, the rightmost one bears stress-viz.:

Abbildung in dieser Leseprobe nicht enthalten

In the items above, stress falls on the final heavy syllable, counting from the right. This means that the directionality of parsing is from right to left. The opposite directionality yields unattested stress patterns. This is shown below:

7) Left to right directionality

*án.zar

*ár.jaz

As the asterisks show, the stress patterns above are unattested. This constitutes a cogent argument in favour of the right-to-left directionality (Cf. Adnor, 1995; Marouane, 1997; and Faizi, 2002 for a similar position).

In the next subsection we move to considering the stress patterns of trisyllabic words.

II-3-1-3- Trisyllabic words

The examination of AWTB trisyllabic words reveals that seven patterns are permissible. The first set to be considered is the one containing a series of light syllables.

Abbildung in dieser Leseprobe nicht enthalten

In these items, stress falls on the initial syllable. The syllables that make up these words are all light. At this point, it should be pointed out that the second syllable (sek) in a.sek.ka is light. In my book (2008) we show that schwa is epenthesised to a consonantal cluster (here /sk/) to head the mora that serves as a licenser for the consonantal sequence.

On the other hand, trisyllabic items corresponding to the shape LLH have primary stress on the final syllable. This is exemplified in (8b) below.

Abbildung in dieser Leseprobe nicht enthalten

Again, the data above show that syllable weight is a decisive factor in the assignment of stress.

Trisyllabic nouns of the shape LHL have the following stress pattern.

Abbildung in dieser Leseprobe nicht enthalten

Since the penultimate syllables in the words above are heavy, they host stress.

Concerning the trisyllabic words composed of a combination of one light and two heavy syllables, stress is assigned to the rightmost branching syllable. Consider the following data:

Abbildung in dieser Leseprobe nicht enthalten

The consideration of the HLL items reveals that stress falls on the initial heavy syllable- viz.,:

Abbildung in dieser Leseprobe nicht enthalten

As for HLH items, stress shows up on the last syllable since it is the rightmost heavy syllable. The following words are given for the sake of illustration.

Abbildung in dieser Leseprobe nicht enthalten

The items corresponding to the HHL pattern have stress on the heavy syllable near the right edge (i.e. the one in the middle). This is shown below:

Abbildung in dieser Leseprobe nicht enthalten

II-3-1-4 Quadrisyllabic words

The last set of items to be considered is the one that consists of four syllables. The corpus shows that these combinations are attested in words inflected for number. Let us consider the LLLH pattern first.

Abbildung in dieser Leseprobe nicht enthalten

Again, these words are stressed on the final syllable since it is heavy. This shows that the distinction heavy/light syllable determines the stress placement in AWTB.

As far as words with a LHLL sequence are concerned, we provide the following data for the sake of illustration.

Abbildung in dieser Leseprobe nicht enthalten

In the words above, primary stress falls on the rightmost heavy syllable. This means that syllable weight and the syllable’s position condition the assignment of stress.

The quadrisyllabic set to be considered next corresponds to the pattern LLHH. Syllable weight and the right-to-left directionality of parsing assign stress to the rightmost heavy syllable-viz:

Abbildung in dieser Leseprobe nicht enthalten

The LLHH items are limited in number.

The last items to be examined are of the shape LHLL. Here again, stress falls on the heavy syllable of the word.

[...]