Onset-Driven Processes in OT from the Perspective of Selected Germanic and Slavic Languages

(1)

Onset-Driven Processes in OT from the Perspective of Selected Germanic and Slavic Languages

Piotr Mosionek

Praca doktorska

Wykonana w Instytucie Anglistyki

Wydziału Neofilologii Uniwersytetu Warszawskiego Pod kierunkiem prof. uczelni dr. hab. Bartłomieja Czaplickiego

Oraz promotor pomocniczej dr Joanny Zaleskiej

Warszawa 2022

(2)

Onset-Driven Processes in OT from the Perspective of Selected Germanic and Slavic Languages

Piotr Mosionek

A dissertation submitted

to the Institute of English Studies, University of Warsaw in partial fulfillment of the requirements

for the degree of Doctor of Philosophy in Linguistics Thesis supervisor: Professor Bartłomiej Czaplicki

Assistant thesis supervisor: Dr Joanna Zaleska

Warsaw 2022

(3)

3

Preface

This dissertation addresses the issue of prevocalic consonant epenthesis in selected Germanic and Slavic languages. The Germanic languages discussed are English (Received Pronunciation, American English, and South African English), German, and Dutch. The languages representing the Slavic language family are Polish (alongside its dialectal varieties – rural Polish, Podhale Goralian, and Kurpian Polish), Kashubian, Ukrainian (two varieties – Modern Standard Ukrainian and its south-western dialects), Slovak (and its colloquial, non-regional dialect), Bulgarian and Czech. The dissertation is divided into six chapters.

Chapter 1 introduces the concept of prevocalic consonant insertion and its motivation. The chapter also lists the goals of the dissertation and introduces the theoretical models relevant for the analysis.

Chapter 2 provides a typological outline of prevocalic consonant epenthesis in the languages listed above. The typology is divided into two main parts, discussing the site of epenthesis and the quality of the epenthetic segment, respectively. Concerning the position in which prevocalic consonant insertions occur, the following distinction is adopted: vowel hiatus, and the left edge of a vowel-initial constituent, such as the morphological word, morphological root, or stressed syllable. The typological overview indicates that onsetless syllables are avoided in both environments. The part devoted to the quality of the epenthetic segment shows that epenthetic consonants may be of several types. This part of Chapter 2 aims to demonstrate that the quality of the inserted consonant may vary depending on several factors, including the aforementioned site of epenthesis.

Chapter 3 presents selected formal analyses of prevocalic consonant epenthesis proposed to date. It is shown how different theoretical approaches have captured the process cross-linguistically. The chapter is divided into two main parts, which illustrate generative and listener-oriented approaches, respectively. The discussion of generative models includes SPE Phonology, Autosegmental Phonology, and Optimality Theory, alongside one of its extensions, Derivational Optimality Theory. It is argued that models such as Optimality Theory, which aim at optimizing syllable structure, provide a more insightful perspective on prevocalic consonant epenthesis than SPE-based analyses. In optimality-theoretic accounts, prevocalic consonant

(4)

4

insertion is assumed to be motivated by the need for syllables to have onsets. The section describing the listener-oriented approach to sound change shows that the account sheds new light on certain phenomena, such as intervocalic glide insertion.

The discussed account shows that intervocalic transition from or to high vowels requires a formant transition perceptually similar to a glide. Thus, the listener- oriented account finds an explanation for the emergence of intervocalic glides in phonetics and perception. Consequently, according to some sources, a diachronic, perception-driven explanation is sufficient to account for these processes, with no need to refer to phonological universals (i.e., the need for syllables to have onsets).

Chapter 4 puts forward a theoretical analysis of prevocalic consonant epenthesis, drawing on the assumptions of the aforementioned theoretical models. The proposed model of analysis considers word-initial and intervocalic epenthesis separately. It is argued that word-initial glide epenthesis is onset-driven, as the glides which appear word-initially cannot be a result of an intervocalic transition. On the theoretical side, it is shown that Optimality Theory has the proper tools to account for the process. As for intervocalic glide insertion, which is common cross-linguistically, it is proposed that phonological universals are not necessary to explain the process. Instead, the listener-oriented model is employed, according to which the intervening glides emerge as a natural phonetic effect, and thus intervocalic glide insertion should not be viewed as an onset-driven process. However, there is no consensus in the literature as to the phonetic realization of lexical and transitional (or epenthetic) glides. It remains unclear, whether there is an articulatory and perceptual difference in the realization of those different types of glides. Thus, two experimental studies, described in Chapter 5 are conducted to examine the perception of those glides by users of Polish.

Chapter 5 discusses the methods, goals, procedures, and results of two behavioral studies conducted on users of Polish. In the first study, an experiment was run on a group of phonetically trained speakers of Polish, while the second, main experiment was conducted on users of Polish with no phonetic background. Both studies aimed to determine whether there is a perceptual difference between lexical and transitional/epenthetic glides. In Polish, there are words that end in <uła> in spelling, such as buła [ˈbuwa] ‘big bread roll’. It is universally agreed in the literature that such words are realized with a lexical glide. However, there are also words that end with <ua> in spelling, such as statua [staˈtua]/[staˈtuwa] ‘statue’, which lack a

(5)

5

unanimous interpretation. The results of both studies, in which the participants were exposed to the recordings of nonce words spelled <uła> and <ua>, indicated that there is no perceptual difference between lexical and transitional/epenthetic glides.

Such an outcome suggests that the proposed model of analysis of intervocalic glides as transitional segments, as laid out in Chapter 4, is sufficient to account for the emergence of intervocalic glides.

Chapter 6 presents the conclusions and provides directions for further study.

(6)

6

Acknowledgements

First and foremost, I would like to thank my thesis supervisor, Professor Bartłomiej Czaplicki, for his support and patience in the process of writing this dissertation.

Professor Czaplicki was always helpful and available, even during his one-semester stay in Indiana. The PhD seminar run by Professor Czaplicki and our personal meetings were always very inspiring and motivating.

I would like to express my deepest gratitude to dr Joanna Zaleska, my assistant thesis supervisor, who played a special role in writing this dissertation. Hadn’t it been for Joanna’s encouragement and involvement, this dissertation wouldn’t have been completed. I would like to thank Joanna for the long hours spent meeting me online to discuss my progress.

My thanks are also due to dr Dariusz Zembrzuski, who guided me at the beginning of my PhD studies and helped me to start the whole process. I would like to thank Dariusz for his help at the early stage of writing this dissertation.

My deepest gratitude goes to my wife Eliza for her sacrifice and always being with me, especially in the moments I needed her most. My writing of this dissertation has had a significant impact on the organization of her life as well. Thus, I am very grateful to Eliza for making this possible.

I would like to thank my whole family, especially my parents, Agnieszka and Robert, and my grandparents, Jadwiga and Józef, for their support and unconditional faith in me. I also thank my two younger brothers, Jakub and Michał. Special thanks go to Michał, who apart from being keenly interested in my progress, also helped me with the statistical analysis in Chapter 5. I also thank my parents-in-law, Monika and Maciej, and my sister-in-law Iga, who were always very supportive and always stressed the importance of this project.

I would also like to thank all my friends for their support in difficult moments. At times, writing this dissertation required me to step back from some elements of social life. Nevertheless, my friends showed a great understanding, for which I thank everyone very much. Special thanks are due to Przemek Pstrongowski, who helped me in the preparation of the experimental studies in Chapter 5.

Finally, I would like to thank everyone who wished me well over the four years of preparing this dissertation.

(7)

7

Symbols and abbreviations

acc. – accusative C – consonant cl. – class

CPP – Contextual Predictability Principle dat. – dative

DOT – Derivational Optimality Theory F ‒ feature

f. – feminine freq. – frequency gen. – genitive

GrWd – grammatical word

HGI ‒ Homorganic Glide Insertion inf. – infinitive

ins. – instrumental IP – Intonational Phrase

IPA ‒ International Phonetic Alphabet lar. ‒ laryngeal

loc. – locative m. – masculine N ‒ nucleus

NA ‒ not applicable nas ‒ nasal

n. – neuter

(8)

8 nom. – nominative

NKJP ‒ Narodowy Korpus Języka Polskiego NMP – Northern Mazovia Polish

obs ‒ obstruent

OT – Optimality Theory Part. ‒ participant pl. – plural

PrWd – Prosodic Word PW – Phonological Word RP – Received Pronunciation Rt ‒ root node

SAE – South African English SR – surface representation sg. – singular

SP – Standard Polish

SPE – The Sound Pattern of English Std. ‒ standard

UR – underlying representation V – vowel

voc. – vocative σ ‒ syllable

. ‒ syllable boundary Φ ‒ phonological phrase

* – incorrect form

(9)

9

* – in a tableau: constraint violation / / – intermediate representation

// // – underlying representation, input [ ] – surface representation

<> – spelling

‘ ’ – glosses

’ – palatalization

+ – morpheme boundary

# – word boundary

! – fatal violation

 – winning candidate

 – undesired winning candidate

→ – becomes

 – elimination of the attested surface form

>>– ranked higher than

(10)

10

This dissertation examines prevocalic consonant epenthesis in selected Slavic and Germanic languages with respect to two distinct loci – vowel hiatus (VV) and the left edge of a constituent (__V). The typological overview in Chapter 2 addresses the process with respect to the position in which it takes place, as well as the quality of the segment that is inserted. Chapter 3 provides an outline of the analyses of the phenomenon proposed to date. A distinction is made between two main approaches:

the synchronic model adopted by generative phonology, and the diachronic and listener-oriented model. It will be argued that while the apparatus of generative models such as Optimality Theory (OT henceforth; Prince and Smolensky, 1993;

McCarthy and Prince, 1995) accurately describes epenthesis as an onset-driven process at the left edge of a constituent, it is not necessary to account for certain phenomena, such as intervocalic transitions. The latter can be successfully remodeled within the listener-oriented mechanism of sound change. An experimental account in Chapter 5 aims at corroborating this claim. In addition, the shortcomings of the synchronic models will be exposed against the background of determining the quality of the epenthetic segment.

The remainder of this chapter is structured as follows. Section 1.2. introduces the concept of epenthesis and outlines the motivation for the process. Both generative and listener-oriented approaches and their main assumptions are introduced in section 1.3 of this chapter.

1.2 Motivation for prevocalic consonant epenthesis

In basic terms, epenthesis means the addition of a segment. Both vowels and consonants may be inserted. The function of vowel epenthesis may be, for instance, breaking illicit consonant clusters (see Hall, 2011 for an overview). This dissertation, however, centers on prevocalic insertion of consonants.

The title of the dissertation suggests that prevocalic consonant epenthesis may be motivated by the requirement for syllables to have onsets. The syllable (σ) consists of a Nucleus (N) and may have an onset (the segments to the left of the Nucleus) and a

(14)

14

coda (the segments to the right of the nucleus), as shown in (1) below. The Coda is parsed together with a constituent called the Rhyme. Both onsets and codas may consist of more than one consonant, thus forming a complex onset or a complex coda.

Note that while the representation in (1) shows the widely-attested syllable structure with Onset and Rhyme, there are various proposals regarding the syllable representation. For instance, in some analyses, the Rhyme is not considered a separate constituent (see Bosch, 2011 for an overview).

(1) The internal syllable structure as proposed by Pike and Pike (1947) σ

Rhyme

Onset Nucleus Coda

There are languages which tolerate onsetless syllables and allow heterosyllabification of two adjacent vowels (see, for instance, Senturia, 1998). There are also some unique languages, such as Arrernte (aka Aranda; Breen and Pensalfini, 1999), which are believed not to have syllable onsets (but see Topintzi and Nevins, 2017 for a possible reanalysis). Despite this, however, many authors consider the CV syllable shape the most universal (Jakobson, 1929; Greenberg, 1978; Cairns and Feinstein, 1982; Clements and Keyser, 1983; Clements, 1990; Prince and Smolensky, 1993; Blevins, 1995, to name a few).

The preference for the CV syllable shape over other syllable shapes (V, VC) is correlated with the idea of markedness. This concept has been long known in phonological analyses, as it dates back to the Prague School of linguistics (Jakobson, 1929; Trubetzkoy, 1939). The concept has been developed and adopted to account for cross-linguistic phenomena (Prince and Smolensky 1993, Lombardi, 1995; 1997, among many). In general, the theory of markedness determines natural and common cross-linguistic patterns. The theory assigns the value ‘marked’ to rare, infrequent, and language-specific patterns, and the value ‘unmarked’ to those properties that are cross-

(15)

15

linguistically common and predictable. A list of such patterns, which have been attributed the value ‘unmarked’ or ‘marked’ in the literature is provided in (2) below (after Hume, 2011: 80). Hume observes, however, that the status of those patterns and the concept of markedness itself have been a matter of phonological debate (see Hume, 2011 for the discussion). In addition, as will be argued in chapters 3 and 4 of this dissertation, markedness sometimes makes incorrect predictions with respect to segment insertion.

(2) The unmarked-marked status of linguistic patterns (after Hume, 2011: 80)

UNMARKED MARKED

natural less natural

normal less normal

general specialized

simple complex

inactive active

more frequent less frequent

optimal less optimal

predictable unpredictable

acquired earlier acquired later

more phonetically variable less phonetically variable articulatory simple articulatory difficult perceptually strong perceptually weak perceptually weak perceptually strong

universal language-specific

ubiquitous parochial

There are several arguments which support the cross-linguistic unmarkedness of CV syllables. As Topintzi (2011: 1285‒1286) points out, an argument for the unmarked status of CV syllables may be supported by their presence in every human language. Topintzi observes that in some languages, CV is the only attested syllable shape, as is the case in Hua (Blevins, 1995) or Senoufo (Zec, 2007). Additionally, Topintzi (2011: 1285) observes that in language acquisition, CV syllables are the first syllables to be pronounced. The examples in (3) below illustrate the realization of

(16)

16

some Dutch words by children aged from 1 to 5 (after Topintzi, 2011: 1286, citing Levelt et al., 2000).

(3) Dutch: acquisition of CV syllables INPUT OUTPUT GLOSS

/pus/ [pu] ‘cat’

/klar/ [ka] ‘finished’

/oto/ [toto] ‘car’

/api/ [tapi] ‘monkey’

The unmarked status of the CV syllables has also been attributed to perceptual distinctiveness. It has been argued in the literature that speech perception may have a bearing on the phonological processing of speech strings (Jakobson and Waugh, 1987; Blevins, 1995; 2004; 2008; Flemming, 1995; Jun, 1995; Steriade, 1995; Gordon, 1999; Hume and Johnson, 2001). Perceptual distinctiveness is required for the correct recognition of the contrastive properties of sounds. An important role in speech perception has been attributed to perceptual cues of each segment. Wright (2004: 36, citing Wright, 2001) defines perceptual cue as ‘the information in the acoustic signal that allows the listener to apprehend the existence of a phonological contrast’. Some sources, such as Steriade (1999; 2001), argue that there is a cross-linguistic tendency to maximize contrast in order to increase the perceptibility of segments. This tendency is related to the sonority of segments. In linguistic terms, sonority may be defined as the prominence with which a given segment is articulated. Some sounds are argued to be more sonorous than others, which is reflected by a sonority hierarchy. The idea of the sonority hierarchy traces back to Sievers (1893) and Jespersen (1904). The scale in (4) (after Prince and Smolensky, 1993: 12) below ranks the main types of speech sounds in order of decreasing sonority (‘>’ stands for ‘is more sonorous than’). Hence, vowels are the most sonorous segments, followed by glides, liquids, nasals, and obstruents.

(4) Sonority Hierarchy

vowels > glides > liquids > nasals > obstruents

(17)

17

According to some sources, such as Steriade (1999; 2001), segments of low sonority (such as stops) are relatively weakly perceptible, and their perceptibility rises if they are adjacent to segments of high perceptibly, such as vowels. When adjacent to another high sonority segment, the perceptibility of the low-sonority segment relies on external cues, which are visible on the neighboring sound. According to Wright (2004), such cues may affect the shape of segmental sequences. In Wright’s (2004: 49) view, the CV syllable shape is optimal as in the transition from consonant to vowel there are numerous cues for consonants (such as place and manner of articulation and voicing) and quality of the vowels. Thus, the contrast in these transitions is maximally perceptible, contrary to, for instance, VC syllables, in which cues for both the vowel and the consonant are weaker.

Thus, in pursuance of the preferred CV syllable shape, languages of the world employ several operations which provide syllables with onsets, including that of consonant epenthesis. The occurrence of adjacent vowels in separate syllables tends to be dispreferred and consequently, numerous strategies are employed to avoid the undesired VV structure (see, for instance, Payne, 1981; Rosenthall, 1994; 1997; Casali, 1996; 1997; Rubach, 2000; 2007, Picard, 2003, to name a few). While cross- linguistically, there are numerous strategies to avoid VV sequences, this dissertation centers on consonant epenthesis. Other strategies, such as vowel deletion, gliding, and diphthongization are briefly discussed in Chapter 2.

1.3 Theoretical background

This section introduces selected theoretical frameworks in which prevocalic consonant epenthesis has been analyzed. The distinction is made between two approaches – the first one being that of generative phonology, described in section 1.3.1, with subsections dedicated to SPE phonology (1.3.1.1), Feature Geometry (1.3.1.2), and Optimality Theory (1.3.1.3). Section 1.3.2 introduces the listener-oriented approach to sound change (1.3.2). An outline of the analyses carried out in those models in the literature will be provided in Chapter 3.

1.3.1 Generative theoretical approaches

While the development of generative phonology began in the 1950s (e.g., Halle, 1959), it had not become an independent field of linguistics until the publication of

(18)

18

The Sound Pattern of English (SPE henceforth; Chomsky and Halle, 1968). The model assumes the existence of two distinct levels of representation: an underlying and a surface level. According to the generative model, every morpheme is represented individually in its Underlying Representation (UR henceforth). Structures composed of those representations, such as words or sentences become the input to phonological rules. The rules modify the input at the underlying level to arrive at the Surface Representation (SR henceforth) or to constitute the input to another rule.

The beginning of the 1990s saw the development of Optimality Theory (Prince and Smolensky, 1993; McCarthy and Prince, 1995), which replaced the rule-based approach with a constraint-based one. According to the latter, phonological processes are no longer formulated in terms of rules. Rather, potential output forms are evaluated based on language-specific rankings of constraints. The candidate which minimally violates the requirements imposed by the ranking is selected as the output.

The aforementioned theoretical models have been used to account for the process of prevocalic consonant epenthesis. In the general spirit of generative models, consonant epenthesis operates in mapping between the input, in which the consonant is absent, and the output, in which the consonant surfaces. This section presents an overview of the analyses of prevocalic consonant epenthesis that have been offered within the generative tradition.

1.3.1.1 SPE Phonology

Chomsky and Halle’s (1968) The Sound Pattern of English was a foundational landmark in the research area of generative phonology. Despite being centered on the phonology of English, the SPE model was later extended to generative descriptions of a considerable number of the world’s languages. The SPE saw each sound (segment) as a bundle of distinctive features. In Chomsky and Halle’s view, the distinctive features were the basic units of phonology, or in other words, building blocks of segments.

Distinctive features are binary classification labels about the properties of sounds (a feature may be of the value ‘+’ or ‘–’). While the concept of distinctive features had been known before the publication of the SPE (Trubetzkoy, 1939; Jakobson, Fant and Halle, 1952), Chomsky and Halle (1968) were responsible for major readjustments, such as introducing the articulatory features [±back], [±high], and [±low]. The

(19)

19

classification of distinctive features proposed by Chomsky and Halle (1968) is provided in (5) below.

(5) SPE’s distinctive features (Chomsky and Halle, 1968: 299) MAJOR CLASS FEATURES

Sonorant Vocalic Consonantal CAVITY FEATURES

Coronal Anterior

T^ONGUE-BODY FEATURES

High Low Back Round Distributed Covered

Glottal constrictions Secondary apertures Nasal

Lateral

MANNER OF ARTICULATION FEATURES

Continuant

RELEASE FEATURES: INSTANTANEOUS AND DELAYED

Primary release Secondary release SUPPLEMENTARY MOVEMENTS

Suction

Velaric suction (clicks) Implosion

Pressure

Velaric pressure

(20)

20 Ejectives

Tense SOURCE FEATURES

Heightened subglottal pressure Voice

Strident

As has been said at the beginning of section 1.3.1, the main assumption of SPE phonology, as well as generative phonology in general, is a recognition of two representation levels: the Underlying Representation (UR) and the Surface Representation (SR). According to the generative model, every morpheme has a single UR. The rules apply to structures composed of those underlying representations, such as words or sentences (that is, the input). The SPE introduced rule-based formalization of phonological processes, in which rules are ordered and apply one after another. The output of all these processes surfaces as the SR. The output of an individual process may also constitute the input to another rule.

The structure of a phonological re-write rule proposed by the SPE is schematized in (6). The rule specifies the input (A), the environment in which the phonological rule applies (X__Y), and the outcome of the transformation (B).

(6) SPE: structure of a rule (Chomsky and Halle, 1968: 14) A → B /X__Y

Selected SPE-style analyses of prevocalic consonant epenthesis will be outlined in section 3.1.2 of Chapter 3. Crucially for the analysis of prevocalic consonant epenthesis, SPE-style analyses did not recognize the syllable as a phonologically relevant component. The syllable was considered by later analyses, such as Optimality Theory.

1.3.1.2 Feature Geometry

Feature Geometry (Sagey, 1986; Halle, 1992; 1995) is a model which builds on the assumption of the distinctive features introduced by the SPE, however, in Feature Geometry the features are organized in a hierarchy which corresponds with their articulatory functions. The novelty of the Feature Geometry representation is that the features are related; they are grouped into nodes, and phonological processes may be

(21)

21

viewed as spreading or movement across the nodes. The model presented in (7) below is that referred to as the Halle-Sagey Articulator Model of Feature Geometry. In Chapter 4, Feature Geometry will be used to account for spreading of glides from vowels.

(7) Halle-Sagey Articulator Model of Feature Geometry (Sagey, 1986: 2)

ROOT TIER

LARYNGEAL TIER SUPRALARYNGEAL TIER

[±stiff][constricted glottis][spread glottis][±slack]

PLACE SOFT PALATE

[±nasal]

LABIAL CORONAL DORSAL

[±round] [±anterior][±strident][±distributed] [±high][±low][±back]

1.3.1.2 Optimality Theory

The 1990s saw the advent of Optimality Theory (OT henceforth; Prince and Smolensky, 1993; McCarthy and Prince, 1995). Classic OT continues the key assumptions of SPE phonology in that it recognizes the input-output mapping, however, the theory rejects the system of the re-write rules and replaces it with a set of violable constraints. The constraints are ordered in a ranking, which is language- specific. In OT, as in the generative model in general, the Underlying Representation of each morpheme is stored in the user’s lexicon. The morphemes provide input, from which output candidates are generated. The evaluation of candidates is carried out based on a constraint ranking. The candidate that minimally violates the restrictions

(22)

22

imposed by the ranking is selected as the output. The process described above is illustrated in (8) below.

(8) Input-output mapping in OT (Kager, 2004: 8)

An assumption concerning a constraint-based analysis is that the same violable constraints are present in every human language. The differences between languages are accounted for by a language-specific ranking. The rankings can either ban or promote certain outputs. For instance, as will be shown in the forthcoming section, some languages require word-initial syllables to have onsets, while other languages prohibit word-initial consonant insertion, which is reflected by different rankings of constraints for such languages.

The constraints build on two major principles: markedness and faithfulness. The concept of markedness refers to the claim that linguistic structures are either

‘unmarked’, that is preferred cross-linguistically, or ‘marked’, that is avoided. For instance, CV syllables, which are widely preferred in many languages are less marked than onsetless syllables. The emergence of marked outputs is militated against by markedness constraints. The principle of faithfulness accounts for the identity of the input and the output. Thus, constraints that prohibit input-output changes, such as segment insertion and deletion are faithfulness constraints.

As has already been mentioned, in OT, potential output candidates are evaluated on the basis of a language-specific ranking. In other words, the more important a given constraint is in a language, the higher it is ranked. An example of an evaluation is provided in (9) below. In (9), constraint A is ranked higher than constraints B and C.

Note that the dashed line means that there is no specified ranking between constraints B and C. They could be re-ranked, but the result of the evaluation would remain unchanged. Both B and C, however, are dominated by A, which is separated in the

(23)

23

ranking by a solid line. The candidates which fail to meet the criteria imposed by a constraint, incur a violation (‘*’ in the tableau). A candidate which incurs the least costly violation of the ranking is selected as an output. In this dissertation, the winning candidate is always listed first and is the one marked with a symbol of a hand with a pointing finger (). An exclamation mark means that a candidate has been eliminated from the evaluation (in other words, the candidate incurred a fatal violation).

(9) OT: a model of evaluation

/input/ CÔNSTRAINTA CÔNSTRAINTB CÔNSTRAINTC

 a. Candidate A *

b. Candidate B *! *

As will be shown in section 3.1.3 in Chapter 3, in the literature, the OT model has been employed to account for the process of prevocalic consonant epenthesis. In OT, however, the process is motivated by the markedness requirement for syllables to have onsets. In other words, the OT account is goal-oriented, as it aims at providing syllables with the requisite onset. Alternative approaches, such as listener-oriented approach to sound change, described in section 1.3.2, seek explanation for the process elsewhere.

1.3.2 Listener-oriented sound change

Section 1.3.1 introduced selected formal models of generative phonology. Some of those models, such as OT, rely on the notion of markedness. Alternative models of analysis, such as the listener-oriented sound change, build on speech perception as a source of sound change. Contrary to the generative models described in section 1.3.1, the listener-oriented model of sound change is not goal-oriented and assumes an important role in speech perception of the listener. Along these lines, consonant epenthesis may be viewed as a perception- rather than an onset-driven process, as will be argued in Chapters 3 and 4.

The mechanism of the listener-oriented sound change (Andersen, 1973; 1978;

Ohala; 1981; Blevins; 2004; 2008; Blevins and Garrett, 1998; 2004; Czaplicki; 2010b;

2013; 2020) ascribes sound change to the perception of ambiguities in the acoustic signal transmitted from the speaker to the listener. The ambiguities which arise during the transmission are subsequently reproduced by the listener when they turn speaker, as shown in (10) below.

(24)

24

(10) A schema of the listener’s sound change (after Ohala, 1981: 187)

SPEAKER LISTENER LISTENER-TURNED-SPEAKER

/yt/ /ut/

| ↑

produced as reconstructed as produced as

↓ | ↓

[yt] ― heard as → [yt] [ut]

The schema of the listener’s sound change in (10) above shows that while the listener correctly hears the utterance [yt] produced by the speaker, they use their phonetic knowledge to reconstruct the information received in the acoustic signal.

The received information, though heard correctly, is parsed according to the knowledge of the listener, and later reconstructed when the listener turns speaker.

Thus, the listener-oriented model seeks an explanation of some processes in phonetic and perceptual rather than phonological factors. The listener-oriented model will be described in detail in section 3.2.1 in Chapter 3, as a model which may account, for instance, for the reinterpretation of intervocalic transitions as glides.

(25)

25

Chapter 2 Typology of prevocalic epenthesis in selected Germanic and Slavic languages

As already mentioned at the beginning of Chapter 1, there are languages which tolerate onsetless syllables, but the CV syllable shape is considered the most universal cross-linguistically. The latter generalization also finds support in selected Germanic and Slavic languages, as they exhibit a strong preference for filled onsets, and for maintaining the canonical syllable shape (Bethin, 1998; Kiparsky, 1998). In order to provide syllable onset, languages employ various phonological processes. One of such onset-driven processes is prevocalic epenthesis, that is, insertion of phonological material which results in an unfaithful mapping between underlying and surface representations.

In this chapter, I present a typological overview of prevocalic epenthesis in selected Germanic and Slavic languages. The main goal of this chapter is to systematize prevocalic consonant insertions in these languages as well as formulate some generalizations with respect to the site of epenthesis and the quality of epenthetic segments. The chapter is divided into two main sections. Section 2.1 offers a systematization of prevocalic epenthesis regarding its site, that is, the environments in which epenthetic consonants are inserted. The two major environments of prevocalic insertions discussed are vowel hiatus (2.1.1) and the left edge of a constituent (2.1.2). In section 2.2, I discuss the quality of epenthetic consonants and the factors that affect it. Sections 2.2.1, 2.2.2, and 2.2.3 discuss the interaction of the epenthetic segment with the site of epenthesis, with segmental context, and the phonetics of the epenthetic segment, respectively. Section 2.3 summarizes and recapitulates the main points of the discussion.

2.1 Site of epenthesis

In this section, I will argue that prevocalic epenthesis in Slavic and Germanic languages is conditioned by two distinct environments. The first of these, addressed in section 2.1.1, is vowel hiatus, that is, a sequence of adjacent vowels belonging to separate syllables. Hiatus may arise at different morphological levels and thus the data will be presented with reference to such levels. First, in the introductory part, some

(26)

26

other, commonly employed strategies of hiatus resolution are briefly discussed. The morphological levels relevant for prevocalic epenthesis in hiatus position are morpheme-internal position (2.1.1.1), hiatus across morpheme boundaries (2.1.1.2), and hiatus across word boundaries (2.1.1.3). The second site of epenthesis, discussed in section 2.1.2 is the left edge of a constituent. This section will argue that cross- linguistically epenthesis is not only employed to provide syllable onset, but also to emphasize the beginning of a constituent. The section will investigate epenthesis at the left edge of a metrical constituent (2.1.2.1), the left edge of a word (2.1.2.2), the left edge of a morphological root (2.1.2.3), and the left edge of a prosodic constituent (2.1.2.4).

2.1.1 Vowel hiatus

Cross-linguistically, the occurrence of adjacent vowels in separate syllables tends to be dispreferred and consequently avoided (see, for instance, Payne, 1981; Rosenthall, 1994; 1997; Casali, 1996; 1997). Vocalic clusters are often resolved with consonant epenthesis. Apart from epenthesis, languages of the world have been observed to employ multiple strategies aimed at breaking the undesired VV sequences. According to Casali (1996: 1–2), some mechanisms of hiatus resolution include vowel deletion, glide formation, diphthongization, and vowel coalescence. Each of these processes will be briefly illustrated below.

An example of vowel deletion that results in the simplification of underlying vowel sequences is found in a Niger-Congo language Etsako (Elimelech, 1976). As shown in (1) below (after Elimelech, 1976: 37), the language has words that end (Word 1) or begin (Word 2) with a vowel when pronounced in isolation. However, when such words are put together in a phrase, the resulting vowel sequence is simplified by deleting the first vowel.

(1) Etsako: vowel deletion

W^ORD1 G^LOSS W^ORD2 G^LOSS SURFACE FORM G^LOSS [dɛ] ‘buy‘ [akpa] ‘cup‘ [dakpa] ‘buy a cup’

[ukpo] ‘cloth‘ [ɛnodɛ] ‘yesterday‘ [ukpɛnode] ‘yesterday’s cloth’

[owa] ‘house‘ [ɔda] ‘different‘ [owɔda] ‘a different house’

[umhele] ‘salt‘ [ɔtsomhi] ‘some‘ [umhelɔtsomhi] ‘some salt’

Similar processes, removing one of two adjacent vowels from an underlying sequence are robustly attested in many languages from different language families, such as

(27)

27

French (Indo-European; Harris, 2011), Karok (language isolate; Bright, 1957) or Ganda (Bantu; Clements, 1986).

Another process affecting underlying vowel sequences is glide formation. The process may be observed, for instance, in Catalan (Mascaró, 1976; Hualde, 1992;

Cabré and Prieto, 2004, inter alia), where a word-initial high vowel /i/ or /u/ is replaced with a high glide [j] or [w], respectively, when the preceding word ends in a vowel.¹ This is illustrated in (2) (after Hualde, 1992: 375).

(2) Catalan: gliding

SPELLING WORD1 WORD2 SURFACE FORM GLOSS

contará históries

[kuntəˈɾa] [isˈtɔɾjəs] [kuntəˈɾa#ˈjsˈtɔɾjəs] ‘he will tell stories’

catalá universal

[kətəˈla] [uniβəɾˈsal] [kətəˈla#wniβəɾˈsal] ‘universal Catalan’

franco-italià [ˈfranku] [itəˈlja] [ˈfranku#jtəˈlja] ‘Franco- Italian’

le intimitat [lə] [intimiˈtat] [lə#jntimiˈtat] ‘the intimacy’

Gliding is a well-attested strategy of hiatus resolution cross-linguistically, especially if one of the vowels involved is high. The process has been documented in some genetically diverse languages, such as Polish (Indo-European; Rubach, 2000), Finnish (Uralic; Suomi, Toivanen and Ylitato, 2008), Arabic (Afro-Asiatic; Watson, 2002), or Ganda (Bantu; Tucker, 1962; Katamba, 1985).

In some languages, adjacent vowels are both placed in the same syllable, but neither of them occupies the syllable margin, as is the case in glide formation. Rather, both vowels are incorporated into the syllable nucleus, forming a diphthong. This can be observed, for example, in a Sudanic language Ngiti, where, as shown in (3) (after Kutsch Lojenga, 1994: 90–91), the process of diphthongization affects vowel sequences created across word boundaries.²

1I treat the output segments in (2) as glides, following Mascaró (1976) and Hualde (1992). There is, however, a debate whether the output sequences found at the word boundaries in Catalan are the result of glide formation (to [aj], [aw], [uj], [əj], respectively), or should be interpreted as diphthongs (i.e.

[a͜i], [a͜u], [u͜i], [ə͜i]; see, for instance, Cabré and Prieto, 2004). Irrespective of the intepretation of these sequences, however, the data demonstrate that vowel hiatus formed at a word boundary is avoided in Catalan.

2 One may wonder why diphthongisation, rather than gliding, is postulated in (3). Casali justifies this interpretation by citing personal communication with Constance Kutsch Lojenga, who claims that this

(28)

28 (3) Ngiti: diphthongization

W^ORD1 W^ORD2 SURFACE FORMS G^LOSS

/opi/ /àji/ [o.pià.ji] ‘Lendu woman’

/ínɖɽì/ /akpà/ [ín.ɖɽìa.kpà] ‘male goat’

/fà/ /ɔɲʊ/ [fàɔ.ɲʊ] ‘our food’

/fʊkʊ/ /oɓ/ [fɽ.kʊo.ɓi] ‘your (pl) knives’

Diphthongization has also been postulated in languages such as Slovak (Indo- European; Rubach, 1993) or Larike (Austronesian; Rosenthall, 1997). The process itself, however, is not uncontroversial because a surface sequence of a vowel with an on- or off-glide may be interpreted as a combination of a vowel and a glide, rather than a diphthong. Establishing the presence of the latter in a language requires a careful analysis of its phonological system.

Overall, there seems to be ample evidence to indicate that vowel hiatus is cross- linguistically dispreferred and tends to be eliminated by means of various phonological processes, such as vowel deletion, glide formation, or diphthongization.

The focus of the remainder of this section is on yet another hiatus resolution strategy, common in both Germanic and Slavic languages. This strategy involves breaking up an underlying sequence of two adjacent vowels by inserting another segment in between.

Vowel hiatus may arise at various levels of morphosyntactic structure: inside a morpheme, across morpheme boundaries, or across word boundaries. While hiatus- driven epenthesis is attested in all these domains in some Slavic or Germanic languages, the details of the hiatus resolution strategy may differ depending on the morphosyntactic conditioning. In some cases, one may observe parallelism between hiatus resolution in different morphosyntactic structures, however, sometimes a given process is limited to only some morphosyntactic conditions. Therefore, in the following sections, these domains will be considered separately, with morpheme- internal hiatus examined in 2.1.1.1, followed by a discussion of hiatus arising at morpheme boundaries in 2.1.1.2 and hiatus at word boundaries in 2.1.1.3.

is due to the acoustics of the resulting segments. The first vowel in the output is always shorter than the second one. This, however, does not cause auditory neutralisation in V1 position.

(29)

29 2.1.1.1 Morpheme-internal hiatus resolution

In numerous Slavic and Germanic languages, underlying morpheme-internal vowel sequences are broken up at the surface level by epenthetic phonological material.

However, in all these languages, hiatus resolution is restricted by the quality of the vowels in the underlying VV sequence. Simply put, only vowels of a certain quality can trigger consonant insertion.

A language in which this restriction is minimal is Dutch. The process of glide insertion itself in Dutch has been well documented in the literature (for instance, Zonneveld, 1978; Gussenhoven, 1980; Berendsen and den Os, 1987; Booij, 1999;

Rubach, 2002). As shown in the dataset below (4) (after Booij, 1999: 66), underlying vowel sequences involving vowels of different height and backness are all separated by a glide.^3,4

(4) Dutch: glide insertion

SPELLING SURFACE FORM GLOSS

bioscoop [bijɔˈskop] ‘cinema’

vijand [ˈvɛ͜ijɑnt] ‘enemy’

diet [ˈdijet] ‘diet’

Ruanda [ˈruʋanda] ‘Rwanda’

Boaz [ˈboʋɑs] ‘Boaz’

januari [janyˈɥari] ‘January’

duo [ˈdyɥo] ‘duo’

fluor [ˈflyɥɔr] ‘fluor’

In some languages, epenthetic glides may only be spawned by vowels of a certain quality. In Polish, for instance, glides are inserted in the context of high vowels exclusively. Thus, a glide is inserted only if at least one vowel forming a hiatus is high, as indicated by the examples in (5a) below. The glides /j/ and /w/ are homorganic with the source vowels /i/ and /u/, respectively. Examples in (5b) show that the vowel

3A careful reader may notice that the quality of the epenthetic glide varies between the cited words.

The issue is set aside in this section to focus on the factors affecting the presence vs absence of epenthesis. The factors affecting the quality of epenthetic segments will be discussed in detail in Section 2.2.

4Both voiced labiodental approximant [ʋ] and the voiced labio–palatal approximant [ɥ] are interpreted as intervocalic glides by Booij (1999: 65‒66) who builds on work of Gussenhoven (1980).

(30)

30

hiatus of non-high vowels is tolerated, which is a characteristic feature of Slavic languages, such as Polish, Czech, Bulgarian, and Slovak (see Rubach, 2000).⁵

(5) Polish: morpheme-internal glide hiatus SPELLING SURFACE FORM GLOSS

a. trio [ˈtrijɔ] 'trio' boisko [boˈjiskɔ] ‘pitch’

Hanoi [ˈxanɔji] ‘Hanoi’

Kair [ˈkajir] ‘Cairo’

Suez [ˈsuwɛs] ‘Suez’

duet [ˈduwɛt] ‘duo’

b. paella [paˈɛja] ‘paella’

maestro [maˈɛstrɔ] ‘maestro’

aorta [aˈɔrta] ‘aorta’

faraon [faˈraɔn] ‘pharaoh’

poeta [pɔˈɛta] ‘poet’

teatr [ˈtɛatr] ‘theater’

There is no consensus regarding the interpretation of the phonological status of the glides which appear in the surface representations of the words in (4) and (5a) above and in the forthcoming data illustrating glide insertion. The nature of intervocalic glides cross-linguistically definitely merits more attention and will be addressed in the remainder of this dissertation. For the clarity of exposition, a distinction of three types of glides is used. The first category of glides are lexical glides, which are present underlyingly in words such as tuja [ˈtuja] ‘thuya’ or szkoła [ˈškɔwa]

‘school’ where the glide is phonological and it appears outside of the context of

5The hiatus of morpheme-internal non-high vowels in Polish has found several ways of interpretation in the literature. While Rubach (2000) argues that the morpheme-internal hiatus of non-high vowels remains unresolved, as shown in (5), other sources (for example, Dukiewicz, 1995) claim that morpheme-internal hiatus may be resolved with a glottal stop (e.g., nauka [naˈɁuka] ‘science’]. Recent sources, such as Osowicka-Kondratowicz (2018) report an ongoing linguistic change. According to this view, glottal stop insertion has extended the domain of application from word-initial position (as in Ala [ˈɁala] ‘proper name’) to word-internal vowel hiatus of non-high vowels and hence, one may expect realization of words such as poeta ‘poet’ with an intervening glottal stop, i.e., [pɔˈɁɛta]. Moreover, in words such as nauka [naˈɁuka] ‘science’ and poeta [pɔˈɁɛta] ‘poet’ , glottal stop insertion may be caused by the beginning of a stressed syllable. The issue of glottal stop insertion in Polish is going to be investigated in section 2.1.2.2.

(31)

31

insertion.⁶ The second subset of glides are epenthetic (phonological) glides (see, for instance, Rubach, 2000). In view of Rubach’s (2000) analysis, the glides which separate vowels in hiatus should be interpreted as the output of phonological insertion, as in for instance trio [ˈtrijɔ] ‘trio’ or sytuacja [sɨtuˈwaʦja] ‘situation’. The glide is inserted to provide onsetless syllables with the requisite onset. The third type of interpretation of an intervocalic transition has been proposed by Blevins (2008) who argues that intervocalic glides are in fact phonetic transition effects which give the perceptual impression of a separate segment between the first and the second vowel of the hiatus. Under this view, transitional elements should be treated as purely phonetic effects rather than phonological segments. Moreover, one may assume that transitional glides always appear in transition between vowels of a certain quality (that is, in hiati where one vowel is high). According to Blevins (2008: 5) reinterpretation of vocalic transitions as glides is more common if glides pre-exist in a given language in different contexts. This view is in line with Ohala’s (1981) listener- oriented change. It remains unclear, however, at which stage a transitional segment should be reinterpreted as a phonological component. This problem, alongside the nature of the glides, is going to be investigated in detail in Chapters 3 and 4.

Data from other languages indicate that the context of morpheme-internal consonant epenthesis may be narrowed further. For instance, in Czech, in contrast to the Polish and Dutch data, the process is limited to hiatus positions in which one of the involved vowels is /i/, as indicated by the examples in (6a) below. Unlike, for instance, Polish, Czech does not insert a high back glide /w/ to break the hiatus which involves /u/, as indicated by the examples in (6b). Consider the data in (6c) below from Rubach (2000: 298) and de Bray (1980: 43).

6There is a debate regarding the nature of glides in general. It remains unclear whether glides should be treated as distinct phonemes, or rather as allophones of vowels (see, for instance, Hayes, 1989;

Rosenthall, 1994; Rubach, 2002; Levi, 2004; 2011).

(32)

32 (6) Czech: morpheme-internal hiatus

While for some languages, such as Czech, glide epenthesis in hiatus is limited to the environment in which one of the vowels is /i/, some languages impose additional restrictions on the position of the vowel in the hiatus. One of such languages is Ukrainian. The data in (7a) below illustrate /j/ insertion in the context of a neighboring /i/. However, /i/ must crucially occupy the second position in the hiatus (i.e., the vowel configuration must be /Vi/). If the vowel /i/ occupies the first position in the hiatus, epenthesis does not occur (7b).

(7) Ukrainian: morpheme-internal glide insertion (Rubach, 2002: 675) SPELLING SURFACE FORM GLOSS

a. Stojik [ˈstɔjik] ‘Stoic’

Izrajil [ˈizrajil] ‘Israel’

intujityvnyj [intujiˈtɪvnɪj] ‘intuitive’

b. dialekt [diaˈlɛkt] ‘dialect’

biolog [biˈɔlɔɣ] ‘biologist’

Irrespective of the quality of the vowels forming hiatus and their position, some languages never resolve the morpheme-internal hiatus at the phonological level. The intervocalic transition, that is the gliding movement of the tongue which is necessary in the transition from or to a high vowel is not considered an output of a phonological process (see Blevins, 2008). Such languages are, for instance, English, Slovak, and

S^PELLING SURFACE FORM G^LOSS a. dialekt [ˈdɪjalɛkt] ‘dialect’

patriot [ˈpatrɪjɔt] ‘patriot’

idiot [ˈɁidijɔt] ‘idiot’

Anglie [ˈɁaŋglijɛ] ‘England’

fialka [ˈfijalka] ‘violet’

Haiti [ˈxajiti] ‘Haiti’

Oneidas [ˈɔnɛjidas] ‘Oneida man’

b. virtuos [ˈvirtuɔs] ‘virtuoso’

fluorid [ˈfluɔrit] ‘fluoride’

duo [ˈduɔ] ‘duo’

duel [ˈduɛl] ‘duel’

(33)

33

Bulgarian. In the latter two, a hiatus formed by non-high vowels is tolerated, as it commonly is in other languages discussed above. However, hiatus involving high vowels is also argued to remain unresolved. Rubach (2000: 284) notes, however, that in colloquial Slovak speech, morpheme-internal glide insertion is attested in the context of high vowels. Thus, one arrives at surface forms such as trio [trijo] ‘trio’ or diéta [dijeːta] ‘diet’. The latter pronunciation is likely evidence for the existence of intervening transitional glides in colloquial Slovak.

(8) Slovak: no hiatus resolution (cf. Rubach, 2000: 274)⁷ SPELLING SURFACE FORM GLOSS

patriot [io] ‘patriot’

január [uaː] ‘January’

Neandertálec [ea] ‘Neanderthal man’

poeta [oe] ‘poet’

One may wonder why epenthesis is postulated morpheme-internally in the examples (1‒7) cited above, given that in most cases, the morpheme internal glides never alternate with zero. It could therefore be argued that they are present at the underlying level. However, alternations are not the sole criterion for establishing underlying representations. Another important principle guiding the determination of an underlying representation of a morpheme is what Rubach (2019: 195) dubs the Contextual Predictability Principle, cited in (9).

(9) Contextual Predictability Principle (CPP)

Phonological information that is predictable from context is not encoded in the underlying representation.

For the languages described above, a following phonotactic generalization can be postulated: two vowels of a certain quality cannot occur next to each other. Hence, the occurrence of an intervening glide is fully predictable in this context. Apart from being predictable, the process of glide insertion is also exceptionless, as it extends to non-native vocabulary and acronyms (see, for instance, Rubach, 2000: 273). Thus,

7In transcription, I follow Rubach’s (2000: 274) convention of transcribing only the relevant, unresolved vocalic cluster.

(34)

34

according to the CPP, the presence of the glide should not be encoded at the underlying level but should rather be derived by a phonological process. As will be shown in the following section, in most of the languages mentioned above, the presence of such a process is further supported by alternations. In Rubach’s (2019) view, the output of the phonological rule of insertion is not stored in the users’ mental lexicon. This view is in line with the CPP, i.e., if the output is possible to derive with phonological rules, it is absent from the mental lexicon. It should be noted that this view has been challenged in the literature. For example, Steriade (2000) and Booij (2008) claim that predictable phonetic information, such as intervocalic transitional glides, can be encoded in speakers’ mental lexicon. Moreover, the CPP is in conflict with the tenet of OT, that is Lexicon Optimization (Prince and Smolensky, 1993), which holds that the contrast between the input and the output should be minimal. In other words, the output should be derived with a minimal number of constraint violations.

2.1.1.2 Hiatus resolution across morpheme boundaries

Hiatus may also arise across morphological boundaries. For the vast majority of languages discussed in the previous section, there is a parallelism between morpheme- internal and heteromorphemic hiatus, i.e., in both environments, the hiatus is resolved with glide insertion, and the process itself is conditioned by the quality of the neighboring vowels. Additionally, in the case of heteromorphemic hiatus resolution, there are alternations that support the interpretation of the glides as epenthetic. There are, however, languages in which epenthesis operates only across word boundaries, but is never found inside morphemes.

The heteromorphemic hiatus may arise as an upshot of affixation or compounding, that is processes which are common in Slavic and Germanic languages (see, for instance, Booij, 2020 for Germanic languages; Manova, 2015, for Slavic languages).

Illicit vocalic clusters may arise, for instance, at the edges of prefixes and morphological roots, or at the boundaries of roots and affixes. The following section outlines the issue of heteromorphemic consonant insertion and provides illustrative data documenting the process.

Most of the languages described in section 2.1.1.1 show parallelism between hiatus resolution inside and across morphemes – the illicit vocalic cluster is separated with a glide. In Dutch, heteromorphemic hiatus, which may arise at the edge of root

(35)

35

and suffix (10a), at the word-clitic boundary (10b)⁸, and in compounds (10c). In all of these instances, the hiatus is eliminated with a glide. The restriction for the glide to agree in backness and roundness with the source vowel is parallel to the restriction found in the morpheme-internal vowel hiatus.

(10) Dutch: hiatus resolution at root-suffix boundary (after Booij, 1999: 66, 150, 166) S^PELLING UNDERLYING FORM SURFACE FORM G^LOSS

a. úi-en /ˈɶy/+/ən/ [ˈɶyɥən] ‘onions’

réu-en /ˈrø/+/ən/ [ˈrøɥən] ‘male dogs’

zee-en /ˈze/+/ən/ [ˈzejən] ‘seas’

ree-en /ˈre/+/ən/ [ˈrejən] ‘deer’

hou-en /hɔu/+/ən/ [ˈhɔuʋən] ‘to hold’

b. Ik zie ‘t boek

/ɪk/+ /si/+/ət /+/buk/ [ɪk sijət buk] ‘I saw the book’

Ik doe ‘t werk

/ɪg/+/du/+/ət/+/ʋɛrk/ [ɪg duʋət ʋɛrk] ‘I do the chores’

c. zee arend /ze/+/arənt/ [ˈzejarənt] ‘sea eagle’

toe eigenen

/tu/+/ɛiɣənən/ [ˈtuʋɛiɣənən] ‘to

appropriate’

sherry achtig

/ˈʃeri/+/ɑxtəx/ [ˈʃerijɑxtəx] ‘cherry-like’

twee-en- twintig

/twe/+/ən/+/twɪntəx/ [twejəntwɪntəx] ‘twenty-two’

die avond /di/+/aːvɔnt/ [ˈdijaːvɔnt] ‘that evening’

The parallelism in hiatus resolution is also found in Polish. Hiatus and its subsequent resolution with glide insertion may be found at the following morphological junctures:

8In Dutch, the singular definite article het [ˈhɛt] ‘the’ which has a weak form [ət], is interpreted as a clitic (Booij, 1999: 165).

(36)

36

that of a prefix and a root (11a)^9,10, a root and a derivational suffix (11b), and a root and an inflectional suffix (11c). Heteromorphemic hiatus involving two non-high vowels remains unresolved, as it is inside morphemes (11d). The primary source of data was Rubach (2000: 291), and some examples were added by the author of the present dissertation.

(11) Polish: epenthesis at root-derivational suffix boundary

SPELLING UNDERLYING FORM SURFACE FORM GLOSS

a. do-inwestować /dɔ/+/invɛstɔvaʨ/ [dɔjinvɛsˈtɔvaʨ] ‘to invest’

do-uczyć /dɔ/+/uʧɨʨ/ [dɔˈwuʧɨʨ] ‘to learn’

b. mao-ist-a /maɔ/+/ista/ [maɔˈjista] ‘Maoist’

sytu-acja /sɨtu/+/aʦja/ [sɨtuˈwaʦja] ‘situation’

c. Salome-i /salɔmɛ/+/i/ [salɔˈmɛji] ‘Salomea (gen.

sg.)’

Erytre-i /ɛrɨtrɛ/+/i/ [ɛrɨˈtrɛji] ‘Erythrea (gen.

sg.)’

d. muze-alny /muzɛ/+/alnɨ/ [muzɛˈalnɨ] ‘of a museum’

lice-alny /liʦɛ/+/alnɨ/ [liʦɛˈalnɨ] ‘of a high school’

The parallelism between morpheme-internal and heteromorphemic hiatus is also found in Czech. The heteromorphemic vocalic cluster is eliminated with glide

9While Rubach (2000:291) postulates glide insertion at the morphological boundary of a root and a suffix, for instance in words such as koka+ina [kɔkajina] ‘cocaine’, he does not discuss prefixed words such as po+instruować ‘to instruct’

10Glide insertion at the prefix-root boundary in Polish has been a matter of contention. Some sources, such as for instance Karpowicz (2009: 56), claim that the insertion of /j/ between a vowel-final prefix and a root beginning with /i/ is incorrect in terms of prescriptive rules of Polish pronunciation. Other sources, such as Rubach (2007: 332), and Rubach and Booij (1990) argue that glide epenthesis in VV environment operates only in the domain of the Phonological Word (henceforth PW). According to Rubach (2007:332), prefixes have a status like that of actual words, and hence they are not prosodified into the following PW. Thus, according to the author, words such as po-instruować ‘to instruct’ are realized with a hiatus, i.e., [pɔ.in.ˈstru.wɔ.vaʨ]. It remains unclear, however, how one should interpret the phonological status of the glide which is seemingly absent (i.e., as a component of phonological domain, or rather, as a transition effect). If there is no transitional glide, one should expect an intervocalic glottal stop.

(37)

37

insertion, if one of the vowels is /i/, as shown in (12a) below. Hiatus in words in (12b), which do not contain /i/, remains unresolved.

(12) Czech: glide insertion in heteromorphemic hiatus (Rubach, 2000: 298)¹¹ SPELLING SURFACE FORM GLOSS

a. koka-in [aji] ‘cocaine’

hindu-ista [uji] ‘Hinduist’

b. muze-um [eu] ‘museum’

lice-um [eu] ‘high school’

Ukrainian, like Dutch, Polish and Czech shows parallelism in resolving hiatus arising at morpheme boundaries. As in the case of morpheme-internal hiatus resolution, the inserted segment is /j/. The illustrative data in (13) below are drawn from Rubach (2002: 675).

(13) Ukrainian: glide insertion in heteromorphemic hiatus¹² SPELLING SURFACE FORM GLOSS

mao-jist [jist] ‘Maoist’

indu-jist [jist] ‘Hinduist’

Slovak and Bulgarian also maintain parallelism between hiatus inside and across morphemes by tolerating the hiatus irrespective of the quality of the vowels, as shown in (14) below. However, recall (8) and morpheme-internal vowel hiatus in Slovak and Bulgarian. While there is no postulated phonological glide insertion, one may expect a transitional glide or a glottal stop.

11I follow Rubach (2000: 298) in transcribing only the relevant part of the surface form.

12Note that Ukrainian and Bulgarian use the Cyrillic alphabet. Thus, the column S^PELLING in the examples from Ukrainian and Bulgarian below in fact shows transliterations. In Ukrainian, the glide is marked by<ї> in spelling as in маоїст ‘Maoist’ and індуїст ‘Hinduist’. Thus, the transliterations include <ji>.

(38)

38

(14) Slovak: no epenthesis in heteromorphemic hiatus (Rubach, 2000: 274) S^PELLING SURFACE FORM G^LOSS

eventu-álne [u.aː] ‘perhaps’

altru-izmus [u.i] ‘altruism’

múze-um [e.u] ‘museum’

In some languages, phonological prevocalic epenthesis is not found to operate inside morphemes. An example of such a language is English, which contrasts with the languages discussed above in two ways: it inserts a consonant in heteromorphemic hiatus only and employs /r/ rather than a glide to break illicit vocalic clusters. This process may be documented in the non-rhotic varieties of English (for instance, the Received Pronunciation (henceforth RP), Australian English, and Welsh English). The non-rhotic varieties are those in which the approximant /r/ does not appear in surface representations word-finally and in preconsonantal position. In the non-rhotic varieties of English, the words in (15a) below are realized without /r/. However, in morphologically complex words consisting of a root and the <-ing> suffix, /r/ is realized in an intervocalic position, as shown in (15b).

(15) English: linking /r/ across morpheme boundaries (Giegerich, 1992: 281) SPELLING UNDERLYING FORM SURFACE FORM

a. bar /ˈbɑ/ [ˈbɑː]

bore /ˈbɔ/ [ˈbɔː]

hear /ˈhɪə/ [ˈhɪə]

cure /ˈkjʊə/ [ˈkjʊə]

b. barring /ˈbɑ/+/ɪŋ/ [ˈbɑːrɪŋ]

boring /ˈbɔ/+/ɪŋ/ [ˈbɔːrɪŋ]

hearing /ˈhɪə/+/ɪŋ/ [ˈhɪərɪŋ]

curing /ˈkjʊə/+/ɪŋ/ [ˈkjʊərɪŋ]

The nature of the so-called linking /r/ in English has been a matter of debate. Whether the /r/ in non-rhotic accents is inserted in an intervocalic position or is simply deleted word-finally and before consonants remains a matter of contention (see, for instance, Rubach, 1997; Gick, 1999 for an overview). However, given its presence on the surface in the prevocalic position exclusively, one may assume that the linking /r/

Onset-Driven Processes in OT from the Perspective of Selected Germanic and Slavic Languages

Onset-Driven Processes in OT from the Perspective of Selected Germanic and Slavic Languages

Piotr Mosionek

Onset-Driven Processes in OT from the Perspective of Selected Germanic and Slavic Languages

Piotr Mosionek

Preface

Acknowledgements

Symbols and abbreviations

Table of Contents

Chapter 1 Introduction

Chapter 2

Typology of prevocalic epenthesis in selected Germanic and Slavic languages

2.1 Site of epenthesis